Re: [hlds_linux] Dual Xeon 2.8GHz and FPS Drops
This is a multi-part message in MIME format. -- [ Picked text/plain from multipart/alternative ] It is simpler, don't turn off HT. Run your cs instances and watch your cpu utilization with top (after starting top, type 1 (one) into the window, that will show utilization for all cpu's.) Ignore the idle time and realize that the total usage numbers for us and sy will only ever add to 200% because there are only two execution units between the four "cpus" listed. William Warren wrote: > It's simple. Turn off HT. Then run your cs instances and watch > your cpu utilization with tip(load average is not really relevant > here..look at the percent of utilization) if you have more than > 50% idle you can run a second instance. Also make sure you have > the ram for this. Check how much ram the cs server are running. > They tend to be leaky. If you have enough free cpu AND you > have enough RAM then try a third instance. if it works ok then > the issue was HT. > > Matt Donnon wrote: > >> ok, so what's the verdict? >> has there been any changes or experiments happening with this that have >> any further relevance to the list? >> >> >> ___ >> To unsubscribe, edit your list preferences, or view the list archives, >> please visit: >> http://list.valvesoftware.com/mailman/listinfo/hlds_linux >> > > -- > Computer House Calls, Networks, Security, Web Design: > http://www.emmanuelcomputerconsulting.com > What businesses are in Brunswick, Maryland? Check Brunswick First! > http://www.checkbrunswickfirst.com > My "Foundation" verse: > Isa 54:17 No weapon that is formed against thee shall prosper; > and every tongue that shall rise against thee in judgment thou > shalt condemn. This is the heritage of the servants of the LORD, > and their righteousness is of me, saith the LORD. > > -- carpe ductum -- "Grab the tape" > CDTT (Certified Duct Tape Technician) > > Linux user #322099 > Machines: > 206822 > 256638 > 276825 > http://counter.li.org/ > > ___ > To unsubscribe, edit your list preferences, or view the list archives, > please visit: > http://list.valvesoftware.com/mailman/listinfo/hlds_linux > > > -- /Sid Stuart Games Engineering / Yahoo! Games 701 First Avenue Sunnyvale, CA 94089 (408) 349-7575 office -- ___ To unsubscribe, edit your list preferences, or view the list archives, please visit: http://list.valvesoftware.com/mailman/listinfo/hlds_linux
Re: [hlds_linux] Dual Xeon 2.8GHz and FPS Drops
It's simple. Turn off HT. Then run your cs instances and watch your cpu utilization with tip(load average is not really relevant here..look at the percent of utilization) if you have more than 50% idle you can run a second instance. Also make sure you have the ram for this. Check how much ram the cs server are running. They tend to be leaky. If you have enough free cpu AND you have enough RAM then try a third instance. if it works ok then the issue was HT. Matt Donnon wrote: ok, so what's the verdict? has there been any changes or experiments happening with this that have any further relevance to the list? ___ To unsubscribe, edit your list preferences, or view the list archives, please visit: http://list.valvesoftware.com/mailman/listinfo/hlds_linux -- Computer House Calls, Networks, Security, Web Design: http://www.emmanuelcomputerconsulting.com What businesses are in Brunswick, Maryland? Check Brunswick First! http://www.checkbrunswickfirst.com My "Foundation" verse: Isa 54:17 No weapon that is formed against thee shall prosper; and every tongue that shall rise against thee in judgment thou shalt condemn. This is the heritage of the servants of the LORD, and their righteousness is of me, saith the LORD. -- carpe ductum -- "Grab the tape" CDTT (Certified Duct Tape Technician) Linux user #322099 Machines: 206822 256638 276825 http://counter.li.org/ ___ To unsubscribe, edit your list preferences, or view the list archives, please visit: http://list.valvesoftware.com/mailman/listinfo/hlds_linux
Re: [hlds_linux] Dual Xeon 2.8GHz and FPS Drops
ok, so what's the verdict? has there been any changes or experiments happening with this that have any further relevance to the list? ___ To unsubscribe, edit your list preferences, or view the list archives, please visit: http://list.valvesoftware.com/mailman/listinfo/hlds_linux
Re: [hlds_linux] Dual Xeon 2.8GHz and FPS Drops
well, if you think load average is more important then where I'd rank it, that's up to you. I think it's just one of a handful of things to watch, and only carries so much weight. In and of itself it doesn't mean much to me, only when combined with other factors do I pay it much mind. On 6/26/05, Ian mu <[EMAIL PROTECTED]> wrote: > Ok, think we're getting somewhere a bit better now. I will disagree > though, load should be on the list of concerns. If you have a load of > 5 on a gameserver box, you should be looking as to whats happening as > they may be having problems. > > I.e you don't use load to tell you about your cpu, you use load to > tell you about how your system is currently working, and you have to > take into consideration how many active processes you have when > looking at the value. > > I.e two scenarios, lets say we have 4 gameservers on a single cpu > machine, and currently is using 30% cpu in total and a load of 4. I'd > be worried as the servers may not be performing that well, chances are > some would be complaining of lag and hitching. At the same time you > could have 4 servers using 75% cpu and a load of 1. This I'd be a lot > less worried about (but would be watching closely) and gameservers > would generally be performing better than the previous scenario. -- Clayton Macleod ___ To unsubscribe, edit your list preferences, or view the list archives, please visit: http://list.valvesoftware.com/mailman/listinfo/hlds_linux
Re: [hlds_linux] Dual Xeon 2.8GHz and FPS Drops
Ok, think we're getting somewhere a bit better now. I will disagree though, load should be on the list of concerns. If you have a load of 5 on a gameserver box, you should be looking as to whats happening as they may be having problems. I.e you don't use load to tell you about your cpu, you use load to tell you about how your system is currently working, and you have to take into consideration how many active processes you have when looking at the value. I.e two scenarios, lets say we have 4 gameservers on a single cpu machine, and currently is using 30% cpu in total and a load of 4. I'd be worried as the servers may not be performing that well, chances are some would be complaining of lag and hitching. At the same time you could have 4 servers using 75% cpu and a load of 1. This I'd be a lot less worried about (but would be watching closely) and gameservers would generally be performing better than the previous scenario. ___ To unsubscribe, edit your list preferences, or view the list archives, please visit: http://list.valvesoftware.com/mailman/listinfo/hlds_linux
Re: [hlds_linux] Dual Xeon 2.8GHz and FPS Drops
I think you're still missing the point I'm trying to make. Which is, load average doesn't really tell you much of anything as far as percentage of CPU used goes. Load average really shouldn't be high on your list of concerns when deciding if your server is up to all of its tasks. Because even this ludicrous example of five processes all using as much CPU as they can get still results in a machine that behaves just fine for all the other processes that are running on it. And it has a 5.0+ load average. So saying that a server should have a load average below 1.0 in order for it to be 'powerful enough' for everything it's running makes no sense at all. And yes, I'm well aware that the fact that a process exists on the machine doesn't mean that the process is constantly eating CPU. Don't know why you keep saying that I do. I never said any existing process results in an additional 1.0 load average. Otherwise, why didn't I act surprised that the load average in the top screen I just pasted didn't say ~130.0? heh. I know what you're saying. You just seem to be discounting what I'm saying. Which, boiled down is, don't pay too much attention to load average cuz it doesn't tell you much. On 6/26/05, Ian mu <[EMAIL PROTECTED]> wrote: > Ok cool, now try it with 5 processes that don't use the cpu constantly > (because most gameservers don't..I think the misleading term is "using > the cpu") and we'll get somewhere in terms of explaining it properly. > I think I know what you are trying to say (or maybe not if you think > it would still be 5) but you're just wording it so it sounds like > something else and we're arguing about semantics. > > I.e why is the load of 12 gaming servers on a box currently 0.5, and > later on it will be 1? Then we'll be getting somewhere. -- Clayton Macleod ___ To unsubscribe, edit your list preferences, or view the list archives, please visit: http://list.valvesoftware.com/mailman/listinfo/hlds_linux
Re: [hlds_linux] Dual Xeon 2.8GHz and FPS Drops
Ok cool, now try it with 5 processes that don't use the cpu constantly (because most gameservers don't..I think the misleading term is "using the cpu") and we'll get somewhere in terms of explaining it properly. I think I know what you are trying to say (or maybe not if you think it would still be 5) but you're just wording it so it sounds like something else and we're arguing about semantics. I.e why is the load of 12 gaming servers on a box currently 0.5, and later on it will be 1? Then we'll be getting somewhere. On 6/26/05, Clayton Macleod <[EMAIL PROTECTED]> wrote: > gee, what a surprise. Five processes all using the CPU, and what's the > load average say? heh. Like I said, try it yourself if you don't > believe me. Since obviously you don't, and you obviously don't think > five processes all competing for the CPU is going to give you a load > average of 5.xx, since you said as much. > > top - 22:32:20 up 1 day, 9:13, 1 user, load average: 5.00, 5.03, 4.20 > Tasks: 130 total, 7 running, 123 sleeping, 0 stopped, 0 zombie > Cpu(s): 1.3% us, 1.0% sy, 97.7% ni, 0.0% id, 0.0% wa, 0.0% hi, 0.0% si > Mem:255692k total, 251644k used, 4048k free,22280k buffers > Swap: 522072k total,24640k used, 497432k free,59384k cached > > PID USER PR NI VIRT RES SHR S %CPU %MEMTIME+ COMMAND > 4547 root 39 19 25772 3244 904 R 19.8 1.3 4:56.91 > FahCore_82.exe > 4612 root 39 19 25772 3244 904 R 19.8 1.3 4:56.21 > FahCore_82.exe > 4430 root 39 19 25768 3240 904 R 19.5 1.3 5:22.33 > FahCore_82.exe > 4507 root 39 19 25728 3128 892 R 19.5 1.2 4:59.86 > FahCore_82.exe > 4634 root 39 19 25728 3124 892 R 19.5 1.2 4:54.65 > FahCore_82.exe > 4636 root 16 0 2020 1012 784 R 1.3 0.4 0:17.24 top > 2084 root 18 0 274m 36m 6420 S 0.3 14.5 69:49.81 java > 2301 root 15 0 19896 4900 2064 S 0.3 1.9 6:37.68 X > ___ To unsubscribe, edit your list preferences, or view the list archives, please visit: http://list.valvesoftware.com/mailman/listinfo/hlds_linux
Re: [hlds_linux] Dual Xeon 2.8GHz and FPS Drops
gee, what a surprise. Five processes all using the CPU, and what's the load average say? heh. Like I said, try it yourself if you don't believe me. Since obviously you don't, and you obviously don't think five processes all competing for the CPU is going to give you a load average of 5.xx, since you said as much. top - 22:32:20 up 1 day, 9:13, 1 user, load average: 5.00, 5.03, 4.20 Tasks: 130 total, 7 running, 123 sleeping, 0 stopped, 0 zombie Cpu(s): 1.3% us, 1.0% sy, 97.7% ni, 0.0% id, 0.0% wa, 0.0% hi, 0.0% si Mem:255692k total, 251644k used, 4048k free,22280k buffers Swap: 522072k total,24640k used, 497432k free,59384k cached PID USER PR NI VIRT RES SHR S %CPU %MEMTIME+ COMMAND 4547 root 39 19 25772 3244 904 R 19.8 1.3 4:56.91 FahCore_82.exe 4612 root 39 19 25772 3244 904 R 19.8 1.3 4:56.21 FahCore_82.exe 4430 root 39 19 25768 3240 904 R 19.5 1.3 5:22.33 FahCore_82.exe 4507 root 39 19 25728 3128 892 R 19.5 1.2 4:59.86 FahCore_82.exe 4634 root 39 19 25728 3124 892 R 19.5 1.2 4:54.65 FahCore_82.exe 4636 root 16 0 2020 1012 784 R 1.3 0.4 0:17.24 top 2084 root 18 0 274m 36m 6420 S 0.3 14.5 69:49.81 java 2301 root 15 0 19896 4900 2064 S 0.3 1.9 6:37.68 X On 6/25/05, Ian mu <[EMAIL PROTECTED]> wrote: > Just ignore Clayton for the time. > > To reiterate...he has tweaked the Hz on the server... > > "Load average only tells you how many active jobs there are." > > Incorrect (unless you want to refine it as active plus queue..queue > being the operative word). > > "It does not tell you if your CPU is being underworked/overworked." > > I'll go along with that, but it can give an indication of your > "system" being overworked which is just as/more important. > > "It has nothing to do with how much idle time the CPU has left over." > > Correct (to a point). > > You still seem to be saying if you have 5 processes running there will > be a load of 5. I really don't think you mean this in truth just bad > explanation, but everything you are saying seems to point to this. > Maybe paste in here the manpage part you are seeing as its clearly > different to other peoples. > > Bottom line for me is ignore all the garbage (including my posts), try > both and ignore stats as they can be misleading, go for what you see > and feel on the game itself (and server fps if necessary) and not on > the physical machine cpu usage. > > ___ > To unsubscribe, edit your list preferences, or view the list archives, please > visit: > http://list.valvesoftware.com/mailman/listinfo/hlds_linux > -- Clayton Macleod ___ To unsubscribe, edit your list preferences, or view the list archives, please visit: http://list.valvesoftware.com/mailman/listinfo/hlds_linux
Re: [hlds_linux] Dual Xeon 2.8GHz and FPS Drops
Just as a p.s to highlight the issue and confusion (not trying to criticize, just hoping it will explain better as the terminology I think is confusing some when arguing), I have one dual cpu machine running 12 dedicated serverswhat do you think the load should be? 12+/6+ or something else? (answer is the latter "something else", when you explain that maybe others will understand what you are really trying to say). On 6/26/05, Ian mu <[EMAIL PROTECTED]> wrote: > Just ignore Clayton for the time. > > To reiterate...he has tweaked the Hz on the server... > > "Load average only tells you how many active jobs there are." > > Incorrect (unless you want to refine it as active plus queue..queue > being the operative word). > > "It does not tell you if your CPU is being underworked/overworked." > > I'll go along with that, but it can give an indication of your > "system" being overworked which is just as/more important. > > "It has nothing to do with how much idle time the CPU has left over." > > Correct (to a point). > > You still seem to be saying if you have 5 processes running there will > be a load of 5. I really don't think you mean this in truth just bad > explanation, but everything you are saying seems to point to this. > Maybe paste in here the manpage part you are seeing as its clearly > different to other peoples. > > Bottom line for me is ignore all the garbage (including my posts), try > both and ignore stats as they can be misleading, go for what you see > and feel on the game itself (and server fps if necessary) and not on > the physical machine cpu usage. > ___ To unsubscribe, edit your list preferences, or view the list archives, please visit: http://list.valvesoftware.com/mailman/listinfo/hlds_linux
Re: [hlds_linux] Dual Xeon 2.8GHz and FPS Drops
Just ignore Clayton for the time. To reiterate...he has tweaked the Hz on the server... "Load average only tells you how many active jobs there are." Incorrect (unless you want to refine it as active plus queue..queue being the operative word). "It does not tell you if your CPU is being underworked/overworked." I'll go along with that, but it can give an indication of your "system" being overworked which is just as/more important. "It has nothing to do with how much idle time the CPU has left over." Correct (to a point). You still seem to be saying if you have 5 processes running there will be a load of 5. I really don't think you mean this in truth just bad explanation, but everything you are saying seems to point to this. Maybe paste in here the manpage part you are seeing as its clearly different to other peoples. Bottom line for me is ignore all the garbage (including my posts), try both and ignore stats as they can be misleading, go for what you see and feel on the game itself (and server fps if necessary) and not on the physical machine cpu usage. ___ To unsubscribe, edit your list preferences, or view the list archives, please visit: http://list.valvesoftware.com/mailman/listinfo/hlds_linux
Re: [hlds_linux] Dual Xeon 2.8GHz and FPS Drops
nopers..two isntances of hlds..i said in my last post without folding..the confusion is nt mine..:) Clayton Macleod wrote: you must be confused, because if you have two folding clients running, each using 50% CPU obviously, then you will have a load average of 2.0 On 6/25/05, William Warren <[EMAIL PROTECTED]> wrote: I can actually run two..i have on another box(minus folding). As long as they did not total 100% cpu usage the load never went to one. -- Clayton Macleod ___ To unsubscribe, edit your list preferences, or view the list archives, please visit: http://list.valvesoftware.com/mailman/listinfo/hlds_linux -- Computer House Calls, Networks, Security, Web Design: http://www.emmanuelcomputerconsulting.com What businesses are in Brunswick, Maryland? Check Brunswick First! http://www.checkbrunswickfirst.com My "Foundation" verse: Isa 54:17 No weapon that is formed against thee shall prosper; and every tongue that shall rise against thee in judgment thou shalt condemn. This is the heritage of the servants of the LORD, and their righteousness is of me, saith the LORD. -- carpe ductum -- "Grab the tape" CDTT (Certified Duct Tape Technician) Linux user #322099 Machines: 206822 256638 276825 http://counter.li.org/ ___ To unsubscribe, edit your list preferences, or view the list archives, please visit: http://list.valvesoftware.com/mailman/listinfo/hlds_linux
Re: [hlds_linux] Dual Xeon 2.8GHz and FPS Drops
you must be confused, because if you have two folding clients running, each using 50% CPU obviously, then you will have a load average of 2.0 On 6/25/05, William Warren <[EMAIL PROTECTED]> wrote: > I can actually run two..i have on another box(minus folding). As > long as they did not total 100% cpu usage the load never went to > one. -- Clayton Macleod ___ To unsubscribe, edit your list preferences, or view the list archives, please visit: http://list.valvesoftware.com/mailman/listinfo/hlds_linux
Re: [hlds_linux] Dual Xeon 2.8GHz and FPS Drops
I can actually run two..i have on another box(minus folding). As long as they did not total 100% cpu usage the load never went to one. Clayton Macleod wrote: only to test/witness what is being discussed here, I'm not suggesting you run that way in general. Only to see how 'load average' and the system behaves with two instances running. That's why. On 6/25/05, William Warren <[EMAIL PROTECTED]> wrote: why would i run two instances when i have one cpu? That makes no sense. I still reccomend the disabling of HT and seeing what the actual usage per cpu is. probably the best thing he can do is remove pingboost..:) -- Clayton Macleod ___ To unsubscribe, edit your list preferences, or view the list archives, please visit: http://list.valvesoftware.com/mailman/listinfo/hlds_linux -- Computer House Calls, Networks, Security, Web Design: http://www.emmanuelcomputerconsulting.com What businesses are in Brunswick, Maryland? Check Brunswick First! http://www.checkbrunswickfirst.com My "Foundation" verse: Isa 54:17 No weapon that is formed against thee shall prosper; and every tongue that shall rise against thee in judgment thou shalt condemn. This is the heritage of the servants of the LORD, and their righteousness is of me, saith the LORD. -- carpe ductum -- "Grab the tape" CDTT (Certified Duct Tape Technician) Linux user #322099 Machines: 206822 256638 276825 http://counter.li.org/ ___ To unsubscribe, edit your list preferences, or view the list archives, please visit: http://list.valvesoftware.com/mailman/listinfo/hlds_linux
Re: [hlds_linux] Dual Xeon 2.8GHz and FPS Drops
only to test/witness what is being discussed here, I'm not suggesting you run that way in general. Only to see how 'load average' and the system behaves with two instances running. That's why. On 6/25/05, William Warren <[EMAIL PROTECTED]> wrote: > why would i run two instances when i have one cpu? That makes no > sense. I still reccomend the disabling of HT and seeing what the > actual usage per cpu is. probably the best thing he can do is > remove pingboost..:) -- Clayton Macleod ___ To unsubscribe, edit your list preferences, or view the list archives, please visit: http://list.valvesoftware.com/mailman/listinfo/hlds_linux
Re: [hlds_linux] Dual Xeon 2.8GHz and FPS Drops
why would i run two instances when i have one cpu? That makes no sense. I still reccomend the disabling of HT and seeing what the actual usage per cpu is. probably the best thing he can do is remove pingboost..:) Clayton Macleod wrote: run two instances of the folding client. Your load average will be 2.0, give or take depending on the other processes. Run three instances of the client, your load average will be 3.0, etc. And as long as ram isn't a problem you'll not notice any difference in responsiveness of the box, because the priority of those folding clients will make them yield the CPU to the other processes that aren't niced at 19. On 6/25/05, William Warren <[EMAIL PROTECTED]> wrote: the two are related though. I run [EMAIL PROTECTED] on my machines. cpu usage is pegged @100% and my load is always at 1.0. Now folding @ home is run at the lowest priority but if anything else starts using the cpu then the load start creeping above 1.x. -- Clayton Macleod ___ To unsubscribe, edit your list preferences, or view the list archives, please visit: http://list.valvesoftware.com/mailman/listinfo/hlds_linux -- Computer House Calls, Networks, Security, Web Design: http://www.emmanuelcomputerconsulting.com What businesses are in Brunswick, Maryland? Check Brunswick First! http://www.checkbrunswickfirst.com My "Foundation" verse: Isa 54:17 No weapon that is formed against thee shall prosper; and every tongue that shall rise against thee in judgment thou shalt condemn. This is the heritage of the servants of the LORD, and their righteousness is of me, saith the LORD. -- carpe ductum -- "Grab the tape" CDTT (Certified Duct Tape Technician) Linux user #322099 Machines: 206822 256638 276825 http://counter.li.org/ ___ To unsubscribe, edit your list preferences, or view the list archives, please visit: http://list.valvesoftware.com/mailman/listinfo/hlds_linux
Re: [hlds_linux] Dual Xeon 2.8GHz and FPS Drops
I'm well aware that a CPU is more or less a serial device and things run through it more or less single file. You know what I meant, you just wish to argue for no reason. It's kinda funny that no matter how many times you read something you just don't understand it. Maybe five times will do it. Load average only tells you how many active jobs there are. Load average only tells you how many active jobs there are. Load average only tells you how many active jobs there are. Load average only tells you how many active jobs there are. Load average only tells you how many active jobs there are. It does not tell you if your CPU is being underworked/overworked. It does not tell you if your CPU is being underworked/overworked. It does not tell you if your CPU is being underworked/overworked. It does not tell you if your CPU is being underworked/overworked. It does not tell you if your CPU is being underworked/overworked. It has nothing to do with how much idle time the CPU has left over. It has nothing to do with how much idle time the CPU has left over. It has nothing to do with how much idle time the CPU has left over. It has nothing to do with how much idle time the CPU has left over. It has nothing to do with how much idle time the CPU has left over. Someone just mentioned using his idle time for a [EMAIL PROTECTED] client. Fire up five of them, even with just one CPU, and you shouldn't notice any difference in the box's performance at all. Even though your load average will be at least 5.0, and likely higher because of the other processes. Unless of course the folding clients use a significant portion of your RAM and this hinders the other processes' RAM usage. Very easy way to prove your theory of a high load average making the box unusable being wrong. But, go ahead, tell me it's not wrong, because a load average of 5.0 is insane. I'll believe you. Honest. On 6/25/05, kama <[EMAIL PROTECTED]> wrote: > > You are not really getting the whole truth. Its not how many processes are > using the CPU at once. Read the first lines on the link I posted earlier. > > "What is Server Load Average? > > Server Load tries to measure the number of active processes - taking into > account waiting processes in the queue to access the processors, and also > the current running processes. > > The Server Load Average gives the sum of the average number of jobs in the > queue over the last 1, 5, and 15 minutes. Load average is not a UNIX > command - it is an embedded metric that appears in the output of other > UNIX commands such as uptime and procinfo. " > > Note the word 'queue' in the first sentence. > > If you still are not convinced try this link instead: > > http://www.teamquest.com/resources/gunther/ldavg1.shtml > > The summary of that link specifies: > > 1. The "load" is not the utilization but the total queue length. > 2. They are point samples of three different time series. > 3. They are exponentially-damped moving averages. > 4. They are in the wrong order to represent trend information. > > I cant find any information about the load in the top manpage. But I do > have the description from the manpage from the function that top, w, > uptime and such uses. > > "The getloadavg() function returns the number of processes in the system > run queue averaged over various periods of time. Up to nelem samples are > retrieved and assigned to successive elements of loadavg[]. The system > imposes a maximum of 3 samples, representing averages over the last 1, 5, > and 15 minutes, respectively." > > Maybe you are trying to say the same thing, but you are writing it in a > way that makes it quite wrong. Btw, there is no such thing as "using the > CPU at once". Each process must wait its turn to get kernel resources. The > schedular is trying to do this the best way it can, but its not always > enough, hence the load gets higher. -- Clayton Macleod ___ To unsubscribe, edit your list preferences, or view the list archives, please visit: http://list.valvesoftware.com/mailman/listinfo/hlds_linux
Re: [hlds_linux] Dual Xeon 2.8GHz and FPS Drops
run two instances of the folding client. Your load average will be 2.0, give or take depending on the other processes. Run three instances of the client, your load average will be 3.0, etc. And as long as ram isn't a problem you'll not notice any difference in responsiveness of the box, because the priority of those folding clients will make them yield the CPU to the other processes that aren't niced at 19. On 6/25/05, William Warren <[EMAIL PROTECTED]> wrote: > the two are related though. I run [EMAIL PROTECTED] on my machines. > cpu usage is pegged @100% and my load is always at 1.0. Now > folding @ home is run at the lowest priority but if anything else > starts using the cpu then the load start creeping above 1.x. http://list.valvesoftware.com/mailman/listinfo/hlds_linux
Re: [hlds_linux] Dual Xeon 2.8GHz and FPS Drops
On Sun, 26 Jun 2005, Matt Donnon wrote: > > Don't take my word for it, run a benchmark and see for yourself. > > I have. Somethings go faster on HT, somethings dont, and occassional > things run like shite because HT confuses them into thinking they have two > real CPUs when they dont. > > With the introduction of dual-core (and in the case of intels extreme dual > core HT) there's been quite a lot of discussion around the hardware review > sites about how to properly benchmark such beasties. It makes for > interesting reading. > > Have you tried disabling HT for a day/hour yet? > an alternative experiment would be to drop the FPS from 500 to 333 for a > day also to see if that will allow more servers to be run. > Both of these experiments are also quite a bit cheaper than option 1; > which is more memory option 2; which is the biggest clockspeed xeons > around, or option 3; which is ditch the box and go opteron and see if that > helps. Btw, He probably will never see all three servers running at 500 fps until he tweaks up the HZ-value. In FreeBSD 5.x its easy. you just bump it with kern.hz in the loader.conf. In linux you probably need to recompile the kernel. I still dont understand why he should go for 500 anyhow, anything over 100 should be enough. /Bjorn -- The devil is in my pants and he wants to say hello. ___ To unsubscribe, edit your list preferences, or view the list archives, please visit: http://list.valvesoftware.com/mailman/listinfo/hlds_linux
Re: [hlds_linux] Dual Xeon 2.8GHz and FPS Drops
You are not really getting the whole truth. Its not how many processes are using the CPU at once. Read the first lines on the link I posted earlier. "What is Server Load Average? Server Load tries to measure the number of active processes - taking into account waiting processes in the queue to access the processors, and also the current running processes. The Server Load Average gives the sum of the average number of jobs in the queue over the last 1, 5, and 15 minutes. Load average is not a UNIX command - it is an embedded metric that appears in the output of other UNIX commands such as uptime and procinfo. " Note the word 'queue' in the first sentence. If you still are not convinced try this link instead: http://www.teamquest.com/resources/gunther/ldavg1.shtml The summary of that link specifies: 1. The "load" is not the utilization but the total queue length. 2. They are point samples of three different time series. 3. They are exponentially-damped moving averages. 4. They are in the wrong order to represent trend information. I cant find any information about the load in the top manpage. But I do have the description from the manpage from the function that top, w, uptime and such uses. "The getloadavg() function returns the number of processes in the system run queue averaged over various periods of time. Up to nelem samples are retrieved and assigned to successive elements of loadavg[]. The system imposes a maximum of 3 samples, representing averages over the last 1, 5, and 15 minutes, respectively." Maybe you are trying to say the same thing, but you are writing it in a way that makes it quite wrong. Btw, there is no such thing as "using the CPU at once". Each process must wait its turn to get kernel resources. The schedular is trying to do this the best way it can, but its not always enough, hence the load gets higher. /Bjorn On Sat, 25 Jun 2005, Clayton Macleod wrote: > argh. It's not hard to understand. It's the average of how many > processes are using the CPU *at once*, over a given time period. I > truly don't give a rat's ass how you *guess* how your servers are > performing. But giving load average such a high priority in your > decision clearly makes it a guess. Doowutchyalike. What makes you > think there are 9 million different man pages for a given app? heh. > /me shakes head > > On 6/25/05, Ian mu <[EMAIL PROTECTED]> wrote: > > I don't think it means 100% at all. No idea where you get that from. > > Its just your explanation doesn't make sense. How can it be how many > > processes have used the cpu in the last minute? Are you saying when > > load is 0.5 half a process has used the cpu in the last 5 mins? > > Basically you're not making sense is the problem. > > > > Load is the amount of processes on average "waiting". Thats VERY > > different. I.e if your load is 1, it means on average there is always > > 1 process waiting to grab some cpu time. If load is 2, there are 2 > > processes waiting on average to keep it pretty simplified. > > > > Load doesn't directly relate to cpu "usage", no one has said that > > iirc. I would say people need to watch both. A load > 2 on a single > > cpu will start to give bad performance on games even if cpu usage is > > 30% (you can get this on certain maps/games for example). Load for me > > is "generally" a better indicator than cpu usage as it gives a better > > viewpoint of responsiveness of servers (imo). But I wouldn't let a > > game machine go over load 2 or load 70% as a rule of thumb for me. > > > > Not sure what man page you're looking at btw. > > > -- > Clayton Macleod > > ___ > To unsubscribe, edit your list preferences, or view the list archives, please > visit: > http://list.valvesoftware.com/mailman/listinfo/hlds_linux > -- Users are like bacteria - each one causes a thousand tiny crises until the host finally gives up and dies. ___ To unsubscribe, edit your list preferences, or view the list archives, please visit: http://list.valvesoftware.com/mailman/listinfo/hlds_linux
Re: [hlds_linux] Dual Xeon 2.8GHz and FPS Drops
Don't take my word for it, run a benchmark and see for yourself. I have. Somethings go faster on HT, somethings dont, and occassional things run like shite because HT confuses them into thinking they have two real CPUs when they dont. With the introduction of dual-core (and in the case of intels extreme dual core HT) there's been quite a lot of discussion around the hardware review sites about how to properly benchmark such beasties. It makes for interesting reading. Have you tried disabling HT for a day/hour yet? an alternative experiment would be to drop the FPS from 500 to 333 for a day also to see if that will allow more servers to be run. Both of these experiments are also quite a bit cheaper than option 1; which is more memory option 2; which is the biggest clockspeed xeons around, or option 3; which is ditch the box and go opteron and see if that helps. Matt ___ To unsubscribe, edit your list preferences, or view the list archives, please visit: http://list.valvesoftware.com/mailman/listinfo/hlds_linux
Re: [hlds_linux] Dual Xeon 2.8GHz and FPS Drops
the two are related though. I run [EMAIL PROTECTED] on my machines. cpu usage is pegged @100% and my load is always at 1.0. Now folding @ home is run at the lowest priority but if anything else starts using the cpu then the load start creeping above 1.x. it's spelled out rather clearly in the man pages. It's how many processes used the CPU during the last 1 minute, 5 minutes, and 15 minutes. My wicked old 500MHz P3 is sitting in the closet running a handful of stuff for the rest of the computers in the house. It routinely has a load average of between 1.5 and 2.5, with a handful of java apps that use about 65-70% CPU. That machine pretty much always has about 25% idle CPU, it's never really maxed out. Yet the load average sits between 1.5 and 2.5. There's still CPU left over for other things. The load average definitely does not tell you anything directly about active/idle CPU cycles. It only deals with *how many processes* are running and actively using CPU. Man pages clearly state this. Experience clearly shows this. For some reason people seem to think that a 1.0 load average means your CPU is being used 100%, when this couldn't be more wrong. If that were true you could never have a load average greater than 1.0, now could you? With that line of thinking you would need to have a dual CPU system to ever see a load average of 2.0, and this is obviously not the case. Load average is only one factor that should govern your choice in whether or not your servers need upgrading. The actual amount of idle CPU time should be a bigger factor than load average. As should memory usage. Load average only tells you so much, it's still useful but it definitely is not the only thing you should look at. It is way down the list of what you should consider, actually. On 6/25/05, Ian mu <[EMAIL PROTECTED]> wrote: Not quite sure where you get that from (if I'm reading it right, apologies if not). Load is previously stated is the amount of processes on average that are waiting. For decent server performance (i.e gaming am referring to, webservers for example you can get away with a lot more), I'd tend to try and keep down to a load of 1, dual cpu's to a load of 2. I'd say you start to get more noticable problems when you get to about double that, so you can just about get away with a load of 2 on a single cpu system, 4 on a dual, but wouldn't want to run a box at those loads. -- Clayton Macleod ___ To unsubscribe, edit your list preferences, or view the list archives, please visit: http://list.valvesoftware.com/mailman/listinfo/hlds_linux -- Computer House Calls, Networks, Security, Web Design: http://www.emmanuelcomputerconsulting.com What businesses are in Brunswick, Maryland? Check Brunswick First! http://www.checkbrunswickfirst.com My "Foundation" verse: Isa 54:17 No weapon that is formed against thee shall prosper; and every tongue that shall rise against thee in judgment thou shalt condemn. This is the heritage of the servants of the LORD, and their righteousness is of me, saith the LORD. -- carpe ductum -- "Grab the tape" CDTT (Certified Duct Tape Technician) Linux user #322099 Machines: 206822 256638 276825 http://counter.li.org/ ___ To unsubscribe, edit your list preferences, or view the list archives, please visit: http://list.valvesoftware.com/mailman/listinfo/hlds_linux
Re: [hlds_linux] Dual Xeon 2.8GHz and FPS Drops
argh. It's not hard to understand. It's the average of how many processes are using the CPU *at once*, over a given time period. I truly don't give a rat's ass how you *guess* how your servers are performing. But giving load average such a high priority in your decision clearly makes it a guess. Doowutchyalike. What makes you think there are 9 million different man pages for a given app? heh. /me shakes head On 6/25/05, Ian mu <[EMAIL PROTECTED]> wrote: > I don't think it means 100% at all. No idea where you get that from. > Its just your explanation doesn't make sense. How can it be how many > processes have used the cpu in the last minute? Are you saying when > load is 0.5 half a process has used the cpu in the last 5 mins? > Basically you're not making sense is the problem. > > Load is the amount of processes on average "waiting". Thats VERY > different. I.e if your load is 1, it means on average there is always > 1 process waiting to grab some cpu time. If load is 2, there are 2 > processes waiting on average to keep it pretty simplified. > > Load doesn't directly relate to cpu "usage", no one has said that > iirc. I would say people need to watch both. A load > 2 on a single > cpu will start to give bad performance on games even if cpu usage is > 30% (you can get this on certain maps/games for example). Load for me > is "generally" a better indicator than cpu usage as it gives a better > viewpoint of responsiveness of servers (imo). But I wouldn't let a > game machine go over load 2 or load 70% as a rule of thumb for me. > > Not sure what man page you're looking at btw. -- Clayton Macleod ___ To unsubscribe, edit your list preferences, or view the list archives, please visit: http://list.valvesoftware.com/mailman/listinfo/hlds_linux
Re: [hlds_linux] Dual Xeon 2.8GHz and FPS Drops
I don't think it means 100% at all. No idea where you get that from. Its just your explanation doesn't make sense. How can it be how many processes have used the cpu in the last minute? Are you saying when load is 0.5 half a process has used the cpu in the last 5 mins? Basically you're not making sense is the problem. Load is the amount of processes on average "waiting". Thats VERY different. I.e if your load is 1, it means on average there is always 1 process waiting to grab some cpu time. If load is 2, there are 2 processes waiting on average to keep it pretty simplified. Load doesn't directly relate to cpu "usage", no one has said that iirc. I would say people need to watch both. A load > 2 on a single cpu will start to give bad performance on games even if cpu usage is 30% (you can get this on certain maps/games for example). Load for me is "generally" a better indicator than cpu usage as it gives a better viewpoint of responsiveness of servers (imo). But I wouldn't let a game machine go over load 2 or load 70% as a rule of thumb for me. Not sure what man page you're looking at btw. On 6/25/05, Clayton Macleod <[EMAIL PROTECTED]> wrote: > it's spelled out rather clearly in the man pages. It's how many > processes used the CPU during the last 1 minute, 5 minutes, and 15 > minutes. My wicked old 500MHz P3 is sitting in the closet running a > handful of stuff for the rest of the computers in the house. It > routinely has a load average of between 1.5 and 2.5, with a handful of > java apps that use about 65-70% CPU. That machine pretty much always > has about 25% idle CPU, it's never really maxed out. Yet the load > average sits between 1.5 and 2.5. There's still CPU left over for > other things. The load average definitely does not tell you anything > directly about active/idle CPU cycles. It only deals with *how many > processes* are running and actively using CPU. Man pages clearly state > this. Experience clearly shows this. > ___ To unsubscribe, edit your list preferences, or view the list archives, please visit: http://list.valvesoftware.com/mailman/listinfo/hlds_linux
Re: [hlds_linux] Dual Xeon 2.8GHz and FPS Drops
it's spelled out rather clearly in the man pages. It's how many processes used the CPU during the last 1 minute, 5 minutes, and 15 minutes. My wicked old 500MHz P3 is sitting in the closet running a handful of stuff for the rest of the computers in the house. It routinely has a load average of between 1.5 and 2.5, with a handful of java apps that use about 65-70% CPU. That machine pretty much always has about 25% idle CPU, it's never really maxed out. Yet the load average sits between 1.5 and 2.5. There's still CPU left over for other things. The load average definitely does not tell you anything directly about active/idle CPU cycles. It only deals with *how many processes* are running and actively using CPU. Man pages clearly state this. Experience clearly shows this. For some reason people seem to think that a 1.0 load average means your CPU is being used 100%, when this couldn't be more wrong. If that were true you could never have a load average greater than 1.0, now could you? With that line of thinking you would need to have a dual CPU system to ever see a load average of 2.0, and this is obviously not the case. Load average is only one factor that should govern your choice in whether or not your servers need upgrading. The actual amount of idle CPU time should be a bigger factor than load average. As should memory usage. Load average only tells you so much, it's still useful but it definitely is not the only thing you should look at. It is way down the list of what you should consider, actually. On 6/25/05, Ian mu <[EMAIL PROTECTED]> wrote: > Not quite sure where you get that from (if I'm reading it right, > apologies if not). Load is previously stated is the amount of > processes on average that are waiting. > > For decent server performance (i.e gaming am referring to, webservers > for example you can get away with a lot more), I'd tend to try and > keep down to a load of 1, dual cpu's to a load of 2. I'd say you start > to get more noticable problems when you get to about double that, so > you can just about get away with a load of 2 on a single cpu system, 4 > on a dual, but wouldn't want to run a box at those loads. -- Clayton Macleod ___ To unsubscribe, edit your list preferences, or view the list archives, please visit: http://list.valvesoftware.com/mailman/listinfo/hlds_linux
Re: [hlds_linux] Dual Xeon 2.8GHz and FPS Drops
>> If you had 25 processes that each constantly used up 2% of your CPU then you would get a load average of 25. 25 processes, 25 load average. >> Not quite sure where you get that from (if I'm reading it right, apologies if not). Load is previously stated is the amount of processes on average that are waiting. For decent server performance (i.e gaming am referring to, webservers for example you can get away with a lot more), I'd tend to try and keep down to a load of 1, dual cpu's to a load of 2. I'd say you start to get more noticable problems when you get to about double that, so you can just about get away with a load of 2 on a single cpu system, 4 on a dual, but wouldn't want to run a box at those loads. ___ To unsubscribe, edit your list preferences, or view the list archives, please visit: http://list.valvesoftware.com/mailman/listinfo/hlds_linux
Re: [hlds_linux] Dual Xeon 2.8GHz and FPS Drops
heh, sorry, but your article just says exactly what I just said. So apparently you haven't even read it. Load average does NOT tell you anything about CPU usage, not as far as amount used / amount idle is concerned. As I just said. Also as I just said, even a load average as high as 5 on a single CPU doesn't mean you need a faster CPU, because a load average as high as 5 on a single CPU can still have 80% idle CPU time. Load average doesn't tell you anything about percentage of CPU usage. Only tells you how many active processes are actively using up *some* CPU time, without actually telling you *how much* CPU time is getting used. Your article states the exact same thing. As does the man pages for 'top' if you bothered to read them. If you had 25 processes that each constantly used up 2% of your CPU then you would get a load average of 25. 25 processes, 25 load average. And you'd still have 50% of your CPU time left over for additional workload. Load average is a good detail to know about your machine, but it definitely doesn't tell you much about CPU usage. On 6/25/05, kama <[EMAIL PROTECTED]> wrote: > On Fri, 24 Jun 2005, Clayton Macleod wrote: > > > You are contradicting yourself or you have a scew picture of what load is. > You are first talking about "how many processes were competing" and then > you write "all that tells you is that in the last minute there was an > average of 5.2 processes each using *some* CPU time at once." Which is not > the same thing and is not really the truth. > > The correct desription of load is more like: > > Load: An average of how many processes that where in queue for getting > kernel resources. > > Have a look at http://www.hostpronto.com/article/36 for more info on UNIX > load. > > > This is also not really true. You will get a good look how much each > virtual processor uses. At least in FreeBSD. In FreeBSD 5.x you can do a > simple 'ps aux | grep idle' and get how much idle each virtual processor > has. Also all the %CPU usage in top and ps for a process only shows on a > virtual CPU, so in theory you can have 4 processes using 99.999% CPU > without hitting the roof. > > But I agree, he should try to disable HT. In theory you will get an > performance boost, since the SMP will now only need to have to know about > 2 instead of 4 CPU's. You also is not bound to half the CPU power for the > process. So if a process needs to have more than what a HT CPU can offer, > it will then get that without hitting the roof. > > The only reason I can see why you should have HT enabled is that if one > process hangs and start using up 100% of CPU, you still have 75% of the > total cpu power left to work with instead of 50%. But if that happens on a > gameserver, you probably jump to the server directly to fix it anyway. > > There is also a bug in the HT that allows in certain circumstances to read > data from other processes. Due to this bug, FreeBSD now have HT disabled > by default. You need to switch it on with a sysctl variable. > > If I could I would have switched of HT on all my servers. But due to a > bug in FreeBSD, it will not boot up. I get some strange IDE timeouts at > bootup. Perhaps upgrading from 5.3 to 5.4 will solve it. > > /Bjorn -- Clayton Macleod ___ To unsubscribe, edit your list preferences, or view the list archives, please visit: http://list.valvesoftware.com/mailman/listinfo/hlds_linux
Re: [hlds_linux] Dual Xeon 2.8GHz and FPS Drops
This is a multi-part message in MIME format. -- [ Picked text/plain from multipart/alternative ] X-MIME-Autoconverted: from 8bit to quoted-printable by mail-relay1.yahoo.com id j5PEaRMS057260 Sorry for using less than precise wording in my previous letter. To make amends, I quote from the top manual page for the interactive '1' (one) command, ´1´ :Toggle_Single/Separate_Cpu_States -- On/Off This command affects how the 't' command's Cpu States portion is shown. Although this toggle exists primarily to serve mas- sively-parallel SMP machines, it is not restricted to solely SMP environments. When you see 'Cpu(s):' in the summary area, the '1' toggle is On and all cpu information is gathered in a single line. Oth- erwise, each cpu is displayed separately as: 'Cpu0, Cpu1, ... The trick in looking at the output from above is to realize that for CPU pairs (not sure which ones, but one would hope it is 0/1 and 2/3) the percentages reported for us, sy and ni will add, to at most, 100%. (The other trick is not to pay attention to the reported numbers for id and wa when calculating total CPU utiltztion.) As for vmstat, we can remove the myth from the math by thinking about how the average CPU utilization is calculated. The data for the calculation is read from /proc/stat, which has a separate line for each real/virtual cpu. My guess for the vmstat algorithm is to read each line, add the numbers together and then divide by the total number of CPU's. Since an HT system reports four cpus but only has two execution units, the real average is cut in half (divided by four instead of two.) So the us and sy times should be multiplied by two to get a clear reading. For the computer whose problems started this thread, that would give a system average cpu utilization of 40%. This still looks like a low enough number that it can support double the number of servers at 80% CPU utilization. So, there is no need to turn HT off to get better statistics, one just needs to understand what is going on. (Which, I have to admit, I didn't when this thread started.) As for the turning HT off to make multitasking OS's run faster, the argument I have seen so far "since the SMP will now only need to have to know about 2 instead of 4 CPU's" is poorly informed. HT reduces the amount of overhead associated with a context switch.On a system with multiple active processes, this more than makes up for the additional OS overhead of tracking 2 additional CPU's. Don't take my word for it, run a benchmark and see for yourself. Sid Clayton Macleod wrote: >actually, top's "load average" has very little to do with CPU usage. >All it tells you is how many processes were competing for / using *any >amount* of CPU in a given time period. It doesn't tell you what >percentage of CPU is being used, you can have a load average in the >5.x range and still have the CPU sitting idle 80% of the time. Load >average does help you guage performance, but it's only one factor of >many that tell you how the box is actually performing and whether or >not it is underpowered for the tasks it is doing. If the first of the >three load average numbers is 5.2, for argument's sake, all that tells >you is that in the last minute there was an average of 5.2 processes >each using *some* CPU time at once. Again, it does not tell you how >much CPU was actually used. It only tells you how many processes were >competing for CPU. Some/all of them could be using next to no CPU, or >some/all of them could be eating as much as they possibly could. You >can't tell that from the load average values. > >And I'm afraid everyone here that has said HT skews the CPU usage >tracking are entirely correct, you can't accurately tell how much CPU >is actually being used and how much is actually free/idle when you >have HT turned on. It's just a fact of HT life. It doesn't relate at >all to a true dual CPU machine when it comes to measuring how much CPU >is being used and how much is idle, because it's difficult to actually >measure when you're only using tricks to emulate dual CPUs. HT >definitely helps out *some* in a multi-threaded/multi-tasking >environment, but its very nature means you can't really tell how much >it will help out. Nor can you easily measure exactly how much idle >time you have left, because of the slight of hand that's involved. But >if you turn HT off you can very accurately tell how much CPU is >actually being used. So if you're having performance problems, and you >need to know how much CPU is really being used, that is indeed a very >good step to take. Turning HT off and running the same scenario again. > >On 6/24/05, Sid Stuart <[EMAIL PROTECTED]> wrote: > > >>Actually, the top utility will track utilization across all CPU's quite >>nicely. No reason to turn HT off to understand what the system is doing. >>Start the top program and then type 1 (
Re: [hlds_linux] Dual Xeon 2.8GHz and FPS Drops
On Fri, 24 Jun 2005, Clayton Macleod wrote: > actually, top's "load average" has very little to do with CPU usage. > All it tells you is how many processes were competing for / using *any > amount* of CPU in a given time period. It doesn't tell you what > percentage of CPU is being used, you can have a load average in the > 5.x range and still have the CPU sitting idle 80% of the time. Load > average does help you guage performance, but it's only one factor of > many that tell you how the box is actually performing and whether or > not it is underpowered for the tasks it is doing. If the first of the > three load average numbers is 5.2, for argument's sake, all that tells > you is that in the last minute there was an average of 5.2 processes > each using *some* CPU time at once. You are contradicting yourself or you have a scew picture of what load is. You are first talking about "how many processes were competing" and then you write "all that tells you is that in the last minute there was an average of 5.2 processes each using *some* CPU time at once." Which is not the same thing and is not really the truth. The correct desription of load is more like: Load: An average of how many processes that where in queue for getting kernel resources. Have a look at http://www.hostpronto.com/article/36 for more info on UNIX load. > And I'm afraid everyone here that has said HT skews the CPU usage > tracking are entirely correct, you can't accurately tell how much CPU > is actually being used and how much is actually free/idle when you > have HT turned on. It's just a fact of HT life. It doesn't relate at > all to a true dual CPU machine when it comes to measuring how much CPU > is being used and how much is idle, because it's difficult to actually > measure when you're only using tricks to emulate dual CPUs. > HT definitely helps out *some* in a multi-threaded/multi-tasking > environment, but its very nature means you can't really tell how much > it will help out. Nor can you easily measure exactly how much idle > time you have left, because of the slight of hand that's involved. But > if you turn HT off you can very accurately tell how much CPU is > actually being used. So if you're having performance problems, and you > need to know how much CPU is really being used, that is indeed a very > good step to take. Turning HT off and running the same scenario again. This is also not really true. You will get a good look how much each virtual processor uses. At least in FreeBSD. In FreeBSD 5.x you can do a simple 'ps aux | grep idle' and get how much idle each virtual processor has. Also all the %CPU usage in top and ps for a process only shows on a virtual CPU, so in theory you can have 4 processes using 99.999% CPU without hitting the roof. But I agree, he should try to disable HT. In theory you will get an performance boost, since the SMP will now only need to have to know about 2 instead of 4 CPU's. You also is not bound to half the CPU power for the process. So if a process needs to have more than what a HT CPU can offer, it will then get that without hitting the roof. The only reason I can see why you should have HT enabled is that if one process hangs and start using up 100% of CPU, you still have 75% of the total cpu power left to work with instead of 50%. But if that happens on a gameserver, you probably jump to the server directly to fix it anyway. There is also a bug in the HT that allows in certain circumstances to read data from other processes. Due to this bug, FreeBSD now have HT disabled by default. You need to switch it on with a sysctl variable. If I could I would have switched of HT on all my servers. But due to a bug in FreeBSD, it will not boot up. I get some strange IDE timeouts at bootup. Perhaps upgrading from 5.3 to 5.4 will solve it. /Bjorn -- Sometimes...in the morning...when I haven't slept...the computers talk to me, they say "good morning dave", so I say to myself, "That g**D**N Hal's awake!" ___ To unsubscribe, edit your list preferences, or view the list archives, please visit: http://list.valvesoftware.com/mailman/listinfo/hlds_linux
Re: [hlds_linux] Dual Xeon 2.8GHz and FPS Drops
actually, top's "load average" has very little to do with CPU usage. All it tells you is how many processes were competing for / using *any amount* of CPU in a given time period. It doesn't tell you what percentage of CPU is being used, you can have a load average in the 5.x range and still have the CPU sitting idle 80% of the time. Load average does help you guage performance, but it's only one factor of many that tell you how the box is actually performing and whether or not it is underpowered for the tasks it is doing. If the first of the three load average numbers is 5.2, for argument's sake, all that tells you is that in the last minute there was an average of 5.2 processes each using *some* CPU time at once. Again, it does not tell you how much CPU was actually used. It only tells you how many processes were competing for CPU. Some/all of them could be using next to no CPU, or some/all of them could be eating as much as they possibly could. You can't tell that from the load average values. And I'm afraid everyone here that has said HT skews the CPU usage tracking are entirely correct, you can't accurately tell how much CPU is actually being used and how much is actually free/idle when you have HT turned on. It's just a fact of HT life. It doesn't relate at all to a true dual CPU machine when it comes to measuring how much CPU is being used and how much is idle, because it's difficult to actually measure when you're only using tricks to emulate dual CPUs. HT definitely helps out *some* in a multi-threaded/multi-tasking environment, but its very nature means you can't really tell how much it will help out. Nor can you easily measure exactly how much idle time you have left, because of the slight of hand that's involved. But if you turn HT off you can very accurately tell how much CPU is actually being used. So if you're having performance problems, and you need to know how much CPU is really being used, that is indeed a very good step to take. Turning HT off and running the same scenario again. On 6/24/05, Sid Stuart <[EMAIL PROTECTED]> wrote: > Actually, the top utility will track utilization across all CPU's quite > nicely. No reason to turn HT off to understand what the system is doing. > Start the top program and then type 1 (the number one). This should give > a load average by CPU display in the summary area. -- Clayton Macleod ___ To unsubscribe, edit your list preferences, or view the list archives, please visit: http://list.valvesoftware.com/mailman/listinfo/hlds_linux
Re: [hlds_linux] Dual Xeon 2.8GHz and FPS Drops
Actually, the top utility will track utilization across all CPU's quite nicely. No reason to turn HT off to understand what the system is doing. Start the top program and then type 1 (the number one). This should give a load average by CPU display in the summary area. Matt Donnon wrote: I'm sure you're capable of cramming more game servers on a single box. its perfectly possible, but not with the pingboost and FPS your wanting. These chew up quite a lot of processor time, and the HT is only confusing the system as to the real load level. Try it for a day. Disable HT and see what the system load is like when the PCU's arn't lying about their real utilisation. I'm betting that each of your servers is using more than half of a CPU to run. Thus starting the third instance is causing the scheduler to constantly swap threads amongst CPU's meaning you have one that fully loaded and causing lag, and another thats not (but will be in the next schedule cycle). Lets say for example you have two containers that are 100 units each. A block wants/needs 52 units. Thus two units will fit nicely into two containers, but how do you fit a third block of 52 when its indivisible? the simple answer is you cant. ___ To unsubscribe, edit your list preferences, or view the list archives, please visit: http://list.valvesoftware.com/mailman/listinfo/hlds_linux ___ To unsubscribe, edit your list preferences, or view the list archives, please visit: http://list.valvesoftware.com/mailman/listinfo/hlds_linux
Re: [hlds_linux] Dual Xeon 2.8GHz and FPS Drops
I'm sure you're capable of cramming more game servers on a single box. its perfectly possible, but not with the pingboost and FPS your wanting. These chew up quite a lot of processor time, and the HT is only confusing the system as to the real load level. Try it for a day. Disable HT and see what the system load is like when the PCU's arn't lying about their real utilisation. I'm betting that each of your servers is using more than half of a CPU to run. Thus starting the third instance is causing the scheduler to constantly swap threads amongst CPU's meaning you have one that fully loaded and causing lag, and another thats not (but will be in the next schedule cycle). Lets say for example you have two containers that are 100 units each. A block wants/needs 52 units. Thus two units will fit nicely into two containers, but how do you fit a third block of 52 when its indivisible? the simple answer is you cant. ___ To unsubscribe, edit your list preferences, or view the list archives, please visit: http://list.valvesoftware.com/mailman/listinfo/hlds_linux
Re: [hlds_linux] Dual Xeon 2.8GHz and FPS Drops
So if we jump this up to 2GB of RAM (we have 1GB now), you think we'll be in better shape? Enough to be able to run 3-4 HLDS (pingboost 2 and 500fps)? Thanks Sid! - Matt Sid Stuart wrote: This is a multi-part message in MIME format. -- [ Picked text/plain from multipart/alternative ] The performance problem does not look to be due to high CPU load.The first line of vmstat shows the utilization over the life of the machine. The server averages 13% (sy + id) load. With the two servers running, it shows 22% load, lot's of headroom. The wait state percentage (wa) is 0, so the system is not hanging on I/O. Memory looks more likely to be the problem. The six hlds processes are consuming 66% of memory. My math skills say starting another game would consume 99% of memory and leave little space available for buffering maps and such. Doubling or tripling the amount of memory in the system would let it run four to six servers, given the CPU utilization. sid Matt Savona wrote: At the moment, the 3rd server isnt running (because preformance is too poor for all other HLDS). Here is vmstat 1: procs ---memory-- ---swap-- -io --system--cpu r b swpd free buff cache si sobibo incs us sy id wa 2 0 0 72480 40148 64495200 2 34 0 12 1 87 0 1 0 0 72480 40148 64495200 0 0 3843 1473 20 2 77 0 And ps vax (required processes trimmed for brevity): 2451 ?S 0:00 027 1508 568 0.0 syslogd -m 0 2455 ?S 0:00 020 1467 444 0.0 klogd -x 2465 ?S 0:02 0 6 1473 452 0.0 irqbalance 2482 ?S 0:00 127 1588 572 0.0 portmap 2501 ?S 0:00 033 1586 700 0.0 rpc.statd 2536 ?S 0:00 026 1517 560 0.0 rpc.idmapd 2643 ?S 0:00 015 1460 460 0.0 /usr/sbin/acpid 2654 ?S 0:00 9 237 7402 1996 0.1 cupsd 2818 ?S 0:01 0 262 3469 1456 0.1 /usr/sbin/sshd 2836 ?S 0:00 0 143 1944 896 0.0 xinetd -stayalive -pidfile /var/run/xinetd.pid 2854 ?S 0:00 0 690 6349 2768 0.2 sendmail: accepting connections 2863 ?S 0:00 0 690 5453 2348 0.2 sendmail: Queue [EMAIL PROTECTED]:00:00 for /var/spool/clientmqueue 2878 ?S 0:00 023 1572 632 0.0 crond 2894 ?S 0:00 272 4351 2684 0.2 xfs -droppriv -daemon 2911 ?S 0:00 015 1572 596 0.0 /usr/sbin/atd 2920 ?S 0:00 0 230 1713 812 0.0 dbus-daemon-1 --system 2933 ?S 0:38 159 2136 1116 0.1 mdadm --monitor --scan 2949 tty2 S 0:00 0 8 1463 344 0.0 /sbin/mingetty tty2 2950 tty3 S 0:00 0 8 1463 340 0.0 /sbin/mingetty tty3 2951 tty4 S 0:00 0 8 1459 340 0.0 /sbin/mingetty tty4 2952 tty5 S 0:00 0 8 1463 340 0.0 /sbin/mingetty tty5 2953 tty6 S 0:00 0 8 1459 340 0.0 /sbin/mingetty tty6 3104 tty1 S 0:00 0 8 1459 340 0.0 /sbin/mingetty tty1 3270 ?S 0:10 0 306 4269 1240 0.1 SCREEN -d -m -S 10 ./hlds_run -game cstrike -pingboost 2 +ip 69.93.50.152 +exec server.cfg +port 270 3271 pts/1S 0:00 0 554 3869 1048 0.1 /bin/sh ./hlds_run -game cstrike -pingboost 2 +ip 69.93.50.152 +exec server.cfg +port 27015 +maxplay 4231 pts/1R393:12 044 133683 122860 11.8 ./hlds_i686 -game cstrike -pingboost 2 +ip 69.93.50.152 +exec server.cfg +port 27015 +maxplayers 4232 pts/1S 0:00 044 133683 122860 11.8 ./hlds_i686 -game cstrike -pingboost 2 +ip 69.93.50.152 +exec server.cfg +port 27015 +maxplayers 4233 pts/1S 0:00 044 133683 122860 11.8 ./hlds_i686 -game cstrike -pingboost 2 +ip 69.93.50.152 +exec server.cfg +port 27015 +maxplayers 5092 ?S 0:03 0 306 4265 1232 0.1 SCREEN -d -m -S cstrike-62 ./hlds_run -game cstrike -pingboost 2 +ip 69.93.50.153 +exec server.cfg + 5093 pts/2S 0:00 0 554 3865 1020 0.0 /bin/sh ./hlds_run -game cstrike -pingboost 2 +ip 69.93.50.153 +exec server.cfg +port 27015 +maxplay 5101 pts/2S248:45 244 115999 106120 10.2 ./hlds_i686 -game cstrike -pingboost 2 +ip 69.93.50.153 +exec server.cfg +port 27015 +maxplayers 5102 pts/2S 0:00 044 115999 106120 10.2 ./hlds_i686 -game cstrike -pingboost 2 +ip 69.93.50.153 +exec server.cfg +port 27015 +maxplayers 5103 pts/2S 0:00 044 115999 106120 10.2 ./hlds_i686 -game cstrike -pingboost 2 +ip 69.93.50.153 +exec server.cfg +port 27015 +maxplayers 5453 ?S 0:00 0 262 7121 2080 0.2 sshd: 62 [priv] 5455 ?S 0:00 0 262 7261 2276 0.2 sshd: [EMAIL PROTECTED]/0 5805 ?SN 0:00 093 1706 804 0.0 vsftpd 5807 ?SN 0:01 09
Re: [hlds_linux] Dual Xeon 2.8GHz and FPS Drops
This is a multi-part message in MIME format. -- [ Picked text/plain from multipart/alternative ] The performance problem does not look to be due to high CPU load.The first line of vmstat shows the utilization over the life of the machine. The server averages 13% (sy + id) load. With the two servers running, it shows 22% load, lot's of headroom. The wait state percentage (wa) is 0, so the system is not hanging on I/O. Memory looks more likely to be the problem. The six hlds processes are consuming 66% of memory. My math skills say starting another game would consume 99% of memory and leave little space available for buffering maps and such. Doubling or tripling the amount of memory in the system would let it run four to six servers, given the CPU utilization. sid Matt Savona wrote: > At the moment, the 3rd server isnt running (because preformance is too > poor for all other HLDS). Here is vmstat 1: > > procs ---memory-- ---swap-- -io > --system--cpu > r b swpd free buff cache si sobibo incs us > sy id wa > 2 0 0 72480 40148 64495200 2 34 0 12 > 1 87 0 > 1 0 0 72480 40148 64495200 0 0 3843 1473 20 > 2 77 0 > > > And ps vax (required processes trimmed for brevity): > > 2451 ?S 0:00 027 1508 568 0.0 syslogd -m 0 > 2455 ?S 0:00 020 1467 444 0.0 klogd -x > 2465 ?S 0:02 0 6 1473 452 0.0 irqbalance > 2482 ?S 0:00 127 1588 572 0.0 portmap > 2501 ?S 0:00 033 1586 700 0.0 rpc.statd > 2536 ?S 0:00 026 1517 560 0.0 rpc.idmapd > 2643 ?S 0:00 015 1460 460 0.0 /usr/sbin/acpid > 2654 ?S 0:00 9 237 7402 1996 0.1 cupsd > 2818 ?S 0:01 0 262 3469 1456 0.1 /usr/sbin/sshd > 2836 ?S 0:00 0 143 1944 896 0.0 xinetd > -stayalive -pidfile /var/run/xinetd.pid > 2854 ?S 0:00 0 690 6349 2768 0.2 sendmail: > accepting connections > 2863 ?S 0:00 0 690 5453 2348 0.2 sendmail: Queue > [EMAIL PROTECTED]:00:00 for /var/spool/clientmqueue > 2878 ?S 0:00 023 1572 632 0.0 crond > 2894 ?S 0:00 272 4351 2684 0.2 xfs -droppriv > -daemon > 2911 ?S 0:00 015 1572 596 0.0 /usr/sbin/atd > 2920 ?S 0:00 0 230 1713 812 0.0 dbus-daemon-1 > --system > 2933 ?S 0:38 159 2136 1116 0.1 mdadm --monitor > --scan > 2949 tty2 S 0:00 0 8 1463 344 0.0 /sbin/mingetty > tty2 > 2950 tty3 S 0:00 0 8 1463 340 0.0 /sbin/mingetty > tty3 > 2951 tty4 S 0:00 0 8 1459 340 0.0 /sbin/mingetty > tty4 > 2952 tty5 S 0:00 0 8 1463 340 0.0 /sbin/mingetty > tty5 > 2953 tty6 S 0:00 0 8 1459 340 0.0 /sbin/mingetty > tty6 > 3104 tty1 S 0:00 0 8 1459 340 0.0 /sbin/mingetty > tty1 > 3270 ?S 0:10 0 306 4269 1240 0.1 SCREEN -d -m -S > 10 ./hlds_run -game cstrike -pingboost 2 +ip 69.93.50.152 +exec > server.cfg +port 270 > 3271 pts/1S 0:00 0 554 3869 1048 0.1 /bin/sh > ./hlds_run -game cstrike -pingboost 2 +ip 69.93.50.152 +exec server.cfg > +port 27015 +maxplay > 4231 pts/1R393:12 044 133683 122860 11.8 ./hlds_i686 > -game cstrike -pingboost 2 +ip 69.93.50.152 +exec server.cfg +port 27015 > +maxplayers > 4232 pts/1S 0:00 044 133683 122860 11.8 ./hlds_i686 > -game cstrike -pingboost 2 +ip 69.93.50.152 +exec server.cfg +port 27015 > +maxplayers > 4233 pts/1S 0:00 044 133683 122860 11.8 ./hlds_i686 > -game cstrike -pingboost 2 +ip 69.93.50.152 +exec server.cfg +port 27015 > +maxplayers > 5092 ?S 0:03 0 306 4265 1232 0.1 SCREEN -d -m -S > cstrike-62 ./hlds_run -game cstrike -pingboost 2 +ip 69.93.50.153 +exec > server.cfg + > 5093 pts/2S 0:00 0 554 3865 1020 0.0 /bin/sh > ./hlds_run -game cstrike -pingboost 2 +ip 69.93.50.153 +exec server.cfg > +port 27015 +maxplay > 5101 pts/2S248:45 244 115999 106120 10.2 ./hlds_i686 > -game cstrike -pingboost 2 +ip 69.93.50.153 +exec server.cfg +port 27015 > +maxplayers > 5102 pts/2S 0:00 044 115999 106120 10.2 ./hlds_i686 > -game cstrike -pingboost 2 +ip 69.93.50.153 +exec server.cfg +port 27015 > +maxplayers > 5103 pts/2S 0:00 044 115999 106120 10.2 ./hlds_i686 > -game cstrike -pingboost 2 +ip 69.93.50.153 +exec server.cfg +port 27015 > +maxplayers > 5453 ?S 0:00 0 262 7121 2080 0.2 sshd: 62 [priv] > 5455 ?S 0:00 0 262 7261 2276 0.2 sshd: [EMAIL > PROTECTED]/0 > 5805 ?SN 0:00 093 1706 804 0.0 vsftpd > 5807 ?SN 0:01 093 1814 908 0.0 vsftpd > 591
Re: [hlds_linux] Dual Xeon 2.8GHz and FPS Drops
Exactly my point. :) At 05:29 AM 6/24/2005, William Warren wrote: HT is not going to help in this environment. Turning it off is not going to hurt performance. It will more than likely help a bit since the kernel scheduler is not having to juggle two fake cpu's and can concentrate on the two real cpus. Ian mu wrote: Urgh don't turn off HT with 3 servers, probably the worst thing you can do (with 2 or 2 process is fine without). Thing to remember is there's no accurate way of reading HT cpu usage, so they may look a bit skewed, but check other things like fps and you should fine you'll get more. On 6/24/05, Gary <[EMAIL PROTECTED]> wrote: HTT should be disabled to get better performance metrics on a server as it can inflate values to be off by alot. When you turn it off you might see about a 2 percent drop in some performance but it's worth the tradeoff in my opinion. Go out to Intel's web site's "developer" section, and look for "SMT". There's alot of literature about performance measurements. ___ To unsubscribe, edit your list preferences, or view the list archives, please visit: http://list.valvesoftware.com/mailman/listinfo/hlds_linux -- Computer House Calls, Networks, Security, Web Design: http://www.emmanuelcomputerconsulting.com What businesses are in Brunswick, Maryland? Check Brunswick First! http://www.checkbrunswickfirst.com My "Foundation" verse: Isa 54:17 No weapon that is formed against thee shall prosper; and every tongue that shall rise against thee in judgment thou shalt condemn. This is the heritage of the servants of the LORD, and their righteousness is of me, saith the LORD. -- carpe ductum -- "Grab the tape" CDTT (Certified Duct Tape Technician) Linux user #322099 Machines: 206822 256638 276825 http://counter.li.org/ ___ To unsubscribe, edit your list preferences, or view the list archives, please visit: http://list.valvesoftware.com/mailman/listinfo/hlds_linux ___ To unsubscribe, edit your list preferences, or view the list archives, please visit: http://list.valvesoftware.com/mailman/listinfo/hlds_linux
Re: [hlds_linux] Dual Xeon 2.8GHz and FPS Drops
HT or not - I'd still like to be able to run more than two instances of HLDS on this hardware. If you guys are using similar specs, I'm sure you're capable of cramming more game servers on a single box. 2 instances is pretty unreasonable to be honest. What are these big GSPs doing to get so many instances of HLDS running at high fps on a single server? The issue I have with the hz/user_hz tweak is that when pingboost is not enabled (or pingboost 1), and sys_ticrate 500 - we're not getting a steady 500fps. The only way we can hit 500fps 99% of the time is if pingpoost is 2 and sys_ticrate 500. If I can turn pingboost off completely and get a steady 500fps just by manipulating the ticrate - I'd very much like to do it (I was under the impression this is what the hz/user_hz mod achieved). Does anyone have an advice in that department? - Matt Matt Donnon wrote: We are under the assumption that the processor can handle 3 instances of HLDS without a problem. But we're not sure why this 3rd HLDS process is making such a big difference in game server performance. Does anyone have any ideas? Is it the RAM? Have we configured the kernel wrong? Any suggestions would be very helpful! I'd contest that you dont have four CPU's. You have two real ones and two "fake" HT impressions of CPUs. I'm guessing that with two instances of HLDS with AMX and heavy pingboost with high FPS, you are using almost all of the real processors, yet your OS is counting the fake ones towards the overall load (thus in effect making it seem low). When you fire up the third instance, this is asking more than the CPU's can provide and thus you see the performance issues. I'd bet that you'd see almost identical behaviour with HT off as on. Matt ___ To unsubscribe, edit your list preferences, or view the list archives, please visit: http://list.valvesoftware.com/mailman/listinfo/hlds_linux ___ To unsubscribe, edit your list preferences, or view the list archives, please visit: http://list.valvesoftware.com/mailman/listinfo/hlds_linux
Re: [hlds_linux] Dual Xeon 2.8GHz and FPS Drops
It's not ideal as its a bad scenario with 3 servers on 2 cpu's however you look at it, but I'd be surprised if you get better performance with 3 servers on 2 cpu's with HT off. Not going to get lots of improvement, but possibly a small amount I would guess. I seriously doubt there will be any change in performance by disabling HT, what you will get is more accurate system load indicators. AFIK, HT allows you to make simultaneous execution use of both integer and floating point units (normal cpu's can only run one at a time), this is helpful if you are running two bits of software that make use of different resources, but is totally useless when the two bits of software both want the same resource. Matt ___ To unsubscribe, edit your list preferences, or view the list archives, please visit: http://list.valvesoftware.com/mailman/listinfo/hlds_linux
Re: [hlds_linux] Dual Xeon 2.8GHz and FPS Drops
I agree there's obviously only 2 real processors, but the theory is kind of the same, the big problem is you have 2 processes and 2 processors currently working fine. That situation is ideal. The problem is the 3rd process added as the 2 currently are sucking up the best part of each cpu, and how that is handled It's not ideal as its a bad scenario with 3 servers on 2 cpu's however you look at it, but I'd be surprised if you get better performance with 3 servers on 2 cpu's with HT off. Not going to get lots of improvement, but possibly a small amount I would guess. http://www.hardwareanalysis.com/content/reviews/article/1557.7/ That gives a better type of explanation really of 2 different processes on a HT system, I'd be genuinely interested if it comes out better performance with 3 ded servers with HT disabled as I'm no expert on it. ___ To unsubscribe, edit your list preferences, or view the list archives, please visit: http://list.valvesoftware.com/mailman/listinfo/hlds_linux
Re: [hlds_linux] Dual Xeon 2.8GHz and FPS Drops
the only thing HT might help is offloading other processes (like nic/kernel/ide/scsi reqs). The thing to remember is you only have two processors to use. HT helps you use the resources on them more effectively, but doesn't magically make them a proper SMP solution. - Original Message - From: "William Warren" <[EMAIL PROTECTED]> To: Sent: Friday, June 24, 2005 6:59 PM Subject: Re: [hlds_linux] Dual Xeon 2.8GHz and FPS Drops HT is not going to help in this environment. Turning it off is not going to hurt performance. It will more than likely help a bit since the kernel scheduler is not having to juggle two fake cpu's and can concentrate on the two real cpus. Ian mu wrote: Urgh don't turn off HT with 3 servers, probably the worst thing you can do (with 2 or 2 process is fine without). Thing to remember is there's no accurate way of reading HT cpu usage, so they may look a bit skewed, but check other things like fps and you should fine you'll get more. ___ To unsubscribe, edit your list preferences, or view the list archives, please visit: http://list.valvesoftware.com/mailman/listinfo/hlds_linux
Re: [hlds_linux] Dual Xeon 2.8GHz and FPS Drops
HT is not going to help in this environment. Turning it off is not going to hurt performance. It will more than likely help a bit since the kernel scheduler is not having to juggle two fake cpu's and can concentrate on the two real cpus. Ian mu wrote: Urgh don't turn off HT with 3 servers, probably the worst thing you can do (with 2 or 2 process is fine without). Thing to remember is there's no accurate way of reading HT cpu usage, so they may look a bit skewed, but check other things like fps and you should fine you'll get more. On 6/24/05, Gary <[EMAIL PROTECTED]> wrote: HTT should be disabled to get better performance metrics on a server as it can inflate values to be off by alot. When you turn it off you might see about a 2 percent drop in some performance but it's worth the tradeoff in my opinion. Go out to Intel's web site's "developer" section, and look for "SMT". There's alot of literature about performance measurements. ___ To unsubscribe, edit your list preferences, or view the list archives, please visit: http://list.valvesoftware.com/mailman/listinfo/hlds_linux -- Computer House Calls, Networks, Security, Web Design: http://www.emmanuelcomputerconsulting.com What businesses are in Brunswick, Maryland? Check Brunswick First! http://www.checkbrunswickfirst.com My "Foundation" verse: Isa 54:17 No weapon that is formed against thee shall prosper; and every tongue that shall rise against thee in judgment thou shalt condemn. This is the heritage of the servants of the LORD, and their righteousness is of me, saith the LORD. -- carpe ductum -- "Grab the tape" CDTT (Certified Duct Tape Technician) Linux user #322099 Machines: 206822 256638 276825 http://counter.li.org/ ___ To unsubscribe, edit your list preferences, or view the list archives, please visit: http://list.valvesoftware.com/mailman/listinfo/hlds_linux
Re: [hlds_linux] Dual Xeon 2.8GHz and FPS Drops
Urgh don't turn off HT with 3 servers, probably the worst thing you can do (with 2 or 2 process is fine without). Thing to remember is there's no accurate way of reading HT cpu usage, so they may look a bit skewed, but check other things like fps and you should fine you'll get more. On 6/24/05, Gary <[EMAIL PROTECTED]> wrote: > HTT should be disabled to get better performance metrics on a server as it > can inflate values to be off by alot. > When you turn it off you might see about a 2 percent drop in some > performance but it's worth the tradeoff in my opinion. > Go out to Intel's web site's "developer" section, and look for "SMT". > There's alot of literature about performance measurements. > ___ To unsubscribe, edit your list preferences, or view the list archives, please visit: http://list.valvesoftware.com/mailman/listinfo/hlds_linux
Re: [hlds_linux] Dual Xeon 2.8GHz and FPS Drops
HTT should be disabled to get better performance metrics on a server as it can inflate values to be off by alot. When you turn it off you might see about a 2 percent drop in some performance but it's worth the tradeoff in my opinion. Go out to Intel's web site's "developer" section, and look for "SMT". There's alot of literature about performance measurements. At 01:48 AM 6/24/2005, Matt Donnon wrote: > We are under the assumption that the processor can > handle 3 instances of HLDS without a problem. But we're > not sure why this 3rd HLDS process is making such a big > difference in game server performance. Does anyone have > any ideas? Is it the RAM? Have we configured the kernel > wrong? Any suggestions would be very helpful! I'd contest that you dont have four CPU's. You have two real ones and two "fake" HT impressions of CPUs. I'm guessing that with two instances of HLDS with AMX and heavy pingboost with high FPS, you are using almost all of the real processors, yet your OS is counting the fake ones towards the overall load (thus in effect making it seem low). When you fire up the third instance, this is asking more than the CPU's can provide and thus you see the performance issues. I'd bet that you'd see almost identical behaviour with HT off as on. Matt ___ To unsubscribe, edit your list preferences, or view the list archives, please visit: http://list.valvesoftware.com/mailman/listinfo/hlds_linux ___ To unsubscribe, edit your list preferences, or view the list archives, please visit: http://list.valvesoftware.com/mailman/listinfo/hlds_linux
Re: [hlds_linux] Dual Xeon 2.8GHz and FPS Drops
> We are under the assumption that the processor can > handle 3 instances of HLDS without a problem. But we're > not sure why this 3rd HLDS process is making such a big > difference in game server performance. Does anyone have > any ideas? Is it the RAM? Have we configured the kernel > wrong? Any suggestions would be very helpful! I'd contest that you dont have four CPU's. You have two real ones and two "fake" HT impressions of CPUs. I'm guessing that with two instances of HLDS with AMX and heavy pingboost with high FPS, you are using almost all of the real processors, yet your OS is counting the fake ones towards the overall load (thus in effect making it seem low). When you fire up the third instance, this is asking more than the CPU's can provide and thus you see the performance issues. I'd bet that you'd see almost identical behaviour with HT off as on. Matt ___ To unsubscribe, edit your list preferences, or view the list archives, please visit: http://list.valvesoftware.com/mailman/listinfo/hlds_linux
Re: [hlds_linux] Dual Xeon 2.8GHz and FPS Drops
At the moment, the 3rd server isnt running (because preformance is too poor for all other HLDS). Here is vmstat 1: procs ---memory-- ---swap-- -io --system-- cpu r b swpd free buff cache si sobibo incs us sy id wa 2 0 0 72480 40148 64495200 2 34 0 12 1 87 0 1 0 0 72480 40148 64495200 0 0 3843 1473 20 2 77 0 2 0 0 72480 40148 64495200 0 0 3789 1507 20 3 77 0 2 0 0 72480 40148 64495200 0 0 3794 1489 21 2 77 0 1 0 0 72480 40148 64495200 0 0 3804 1549 21 2 77 0 1 0 0 72480 40152 64494800 056 3801 1577 21 3 77 0 1 0 0 72480 40152 64494800 040 3847 1624 21 2 77 0 2 0 0 72480 40152 64494800 0 0 3793 1651 21 2 77 0 1 0 0 71736 40152 64494800 0 0 3830 1714 21 2 77 0 1 0 0 71612 40152 64494800 0 0 3739 1795 21 2 77 0 2 0 0 71736 40152 64494800 0 0 3854 1729 22 2 76 0 2 0 0 71736 40156 64494400 072 3794 1784 23 2 75 0 2 0 0 71736 40156 64494400 0 0 3809 1782 23 3 74 0 1 0 0 73100 40156 64494400 0 0 3831 1878 23 2 75 0 2 0 0 73828 40156 64494400 0 0 3841 1927 22 2 75 0 2 0 0 73828 40156 64494400 0 0 3889 2006 23 2 76 0 1 0 0 73828 40156 64494400 0 0 3874 2009 22 2 76 0 2 0 0 73828 40156 64494400 028 3863 2059 23 2 75 0 2 0 0 73828 40156 64494400 0 0 3826 2144 23 2 75 0 1 0 0 73828 40156 64494400 0 0 3808 1488 21 3 76 0 1 0 0 73828 40156 64494400 0 0 3848 1266 21 2 77 0 1 0 0 73828 40156 64494400 0 0 3843 1228 23 2 75 0 2 0 0 73828 40156 64494400 0 0 3806 1231 23 2 74 0 2 0 0 73828 40160 64494000 028 3847 1282 24 2 74 0 2 0 0 73828 40160 64494000 0 0 3819 1282 24 3 74 0 2 0 0 73828 40160 64494000 0 0 3872 1260 24 2 74 0 2 0 0 73828 40160 64494000 0 0 3843 1304 24 2 74 0 2 0 0 73828 40160 64494000 0 0 3811 1386 24 3 73 0 0 0 0 73828 40160 64494000 0 0 3739 1402 24 2 74 0 2 0 0 73828 40164 64493600 028 3831 1444 24 2 73 0 2 0 0 73828 40164 64493600 0 0 3844 1473 24 2 74 0 2 0 0 73828 40164 64493600 0 0 3790 1498 23 2 74 0 2 0 0 73828 40164 64493600 0 0 3790 1551 25 2 73 0 2 0 0 73828 40164 64493600 0 0 3822 1624 24 2 74 0 2 0 0 73828 40164 64493600 0 0 3826 1767 24 2 73 0 1 0 0 73768 40168 64493200 048 3855 1723 24 2 74 0 1 0 0 73348 40168 64493200 0 0 3795 1812 25 3 72 0 2 0 0 73348 40168 64493200 0 0 3867 1899 23 2 75 0 2 0 0 73348 40168 64493200 0 0 3799 1936 23 2 75 0 2 0 0 73348 40168 64493200 0 0 3857 2000 23 3 73 0 1 0 0 73468 40168 64493200 0 0 3917 2093 22 2 76 0 2 0 0 73648 40176 64492400 028 3910 2144 23 3 75 0 1 0 0 73648 40176 64492400 0 0 3807 1543 22 2 76 0 1 0 0 73648 40176 64492400 0 0 3822 1225 22 2 76 0 1 0 0 73648 40176 64492400 0 0 3855 1296 21 3 76 0 2 0 0 73648 40176 64492400 0 0 3889 1304 22 2 77 0 2 0 0 73648 40176 64492400 032 3855 1282 23 3 75 0 1 0 0 73648 40184 64491600 064 3833 1355 22 2 76 0 2 0 0 73648 40184 64491600 0 0 3788 1502 21 3 77 0 1 0 0 73648 40184 64491600 0 0 3879 1545 20 2 78 0 1 0 0 73648 40184 64491600 0 0 3909 1622 20 2 78 0 2 0 0 73648 40184 64491600 0 0 3886 1655 20 3 77 0 2 0 0 73648 40184 64491600 0 0 3868 1692 21 2 77 0 1 0 0 73648 40192 64490800 032 3877 1774 20 3 78 0 1 0 0 73648 40192 64490800 0 0 3873 1848 20 2 78 0 1 0 0 73648 40192 64490800 0 0 3929 1862 20 2 78 0 2 0 0 73648 40192 64490800 0 0 3843 1890 20 3 77 0 procs ---memory-- ---swap-- -io --system-- cpu r b swpd free buff cache si sobibo incs us sy id wa 1 0 0 73648 40192 64490800 0 0 3837 2013 19 2 78 0 1 0 0 73648 40192 64490800
Re: [hlds_linux] Dual Xeon 2.8GHz and FPS Drops
Remove AMX mod and try again. At 12:11 AM 6/24/2005, Matt Savona wrote: I forgot to mention: - These are 1.6 servers - All 3 instances of HLDS use AMX mod Any input on this issue is greatly appreciated! - Matt Matt Savona wrote: Hi Guys, At the moment we have a machine with the following specs: Dual Xeon 2.8GHz (HT enabled in the kernel) 1GB RAM 74GB 10k RPM SCSI Hard Drive I'm running 2.6.12 kernel with the hz/user_hz tweak (=1000). At the moment we have 2x 12 man privates that are full 100% of the time. These are running with pingboost 2 and sys_ticrate 500. FPS on these two servers is almost always 500. System load is around 1.0. However, when we attempt to run a 3rd HLDS with the same pingboost/ticrate settings, FPS on all 3 servers drop significantly (and randomly) below 100. Load hits about 2.10. When we kill the 3rd process, load drops back to ~1.0 and we get 500 fps on the other two servers again. We are under the assumption that the processor can handle 3 instances of HLDS without a problem. But we're not sure why this 3rd HLDS process is making such a big difference in game server performance. Does anyone have any ideas? Is it the RAM? Have we configured the kernel wrong? Any suggestions would be very helpful! Thanks! - Matt ___ To unsubscribe, edit your list preferences, or view the list archives, please visit: http://list.valvesoftware.com/mailman/listinfo/hlds_linux ___ To unsubscribe, edit your list preferences, or view the list archives, please visit: http://list.valvesoftware.com/mailman/listinfo/hlds_linux -- G. Stanley IP Engineering VSNX [EMAIL PROTECTED] / [EMAIL PROTECTED] Office: +1 (716) 666-2819 Mobile: +1 (304) 283-1298 ___ To unsubscribe, edit your list preferences, or view the list archives, please visit: http://list.valvesoftware.com/mailman/listinfo/hlds_linux
Re: [hlds_linux] Dual Xeon 2.8GHz and FPS Drops
More information is needed to diagnose the performance problem, Run a "vmstat 1" for about 10 seconds while the three programs are running and post the results. Also do a "ps vax" and post the results from that. With the output from those two programs, I can see whether it is CPU or memory bound. Sid Matt Savona wrote: Hi Guys, At the moment we have a machine with the following specs: Dual Xeon 2.8GHz (HT enabled in the kernel) 1GB RAM 74GB 10k RPM SCSI Hard Drive I'm running 2.6.12 kernel with the hz/user_hz tweak (=1000). At the moment we have 2x 12 man privates that are full 100% of the time. These are running with pingboost 2 and sys_ticrate 500. FPS on these two servers is almost always 500. System load is around 1.0. However, when we attempt to run a 3rd HLDS with the same pingboost/ticrate settings, FPS on all 3 servers drop significantly (and randomly) below 100. Load hits about 2.10. When we kill the 3rd process, load drops back to ~1.0 and we get 500 fps on the other two servers again. We are under the assumption that the processor can handle 3 instances of HLDS without a problem. But we're not sure why this 3rd HLDS process is making such a big difference in game server performance. Does anyone have any ideas? Is it the RAM? Have we configured the kernel wrong? Any suggestions would be very helpful! Thanks! - Matt ___ To unsubscribe, edit your list preferences, or view the list archives, please visit: http://list.valvesoftware.com/mailman/listinfo/hlds_linux ___ To unsubscribe, edit your list preferences, or view the list archives, please visit: http://list.valvesoftware.com/mailman/listinfo/hlds_linux
Re: [hlds_linux] Dual Xeon 2.8GHz and FPS Drops
I forgot to mention: - These are 1.6 servers - All 3 instances of HLDS use AMX mod Any input on this issue is greatly appreciated! - Matt Matt Savona wrote: Hi Guys, At the moment we have a machine with the following specs: Dual Xeon 2.8GHz (HT enabled in the kernel) 1GB RAM 74GB 10k RPM SCSI Hard Drive I'm running 2.6.12 kernel with the hz/user_hz tweak (=1000). At the moment we have 2x 12 man privates that are full 100% of the time. These are running with pingboost 2 and sys_ticrate 500. FPS on these two servers is almost always 500. System load is around 1.0. However, when we attempt to run a 3rd HLDS with the same pingboost/ticrate settings, FPS on all 3 servers drop significantly (and randomly) below 100. Load hits about 2.10. When we kill the 3rd process, load drops back to ~1.0 and we get 500 fps on the other two servers again. We are under the assumption that the processor can handle 3 instances of HLDS without a problem. But we're not sure why this 3rd HLDS process is making such a big difference in game server performance. Does anyone have any ideas? Is it the RAM? Have we configured the kernel wrong? Any suggestions would be very helpful! Thanks! - Matt ___ To unsubscribe, edit your list preferences, or view the list archives, please visit: http://list.valvesoftware.com/mailman/listinfo/hlds_linux ___ To unsubscribe, edit your list preferences, or view the list archives, please visit: http://list.valvesoftware.com/mailman/listinfo/hlds_linux
[hlds_linux] Dual Xeon 2.8GHz and FPS Drops
Hi Guys, At the moment we have a machine with the following specs: Dual Xeon 2.8GHz (HT enabled in the kernel) 1GB RAM 74GB 10k RPM SCSI Hard Drive I'm running 2.6.12 kernel with the hz/user_hz tweak (=1000). At the moment we have 2x 12 man privates that are full 100% of the time. These are running with pingboost 2 and sys_ticrate 500. FPS on these two servers is almost always 500. System load is around 1.0. However, when we attempt to run a 3rd HLDS with the same pingboost/ticrate settings, FPS on all 3 servers drop significantly (and randomly) below 100. Load hits about 2.10. When we kill the 3rd process, load drops back to ~1.0 and we get 500 fps on the other two servers again. We are under the assumption that the processor can handle 3 instances of HLDS without a problem. But we're not sure why this 3rd HLDS process is making such a big difference in game server performance. Does anyone have any ideas? Is it the RAM? Have we configured the kernel wrong? Any suggestions would be very helpful! Thanks! - Matt ___ To unsubscribe, edit your list preferences, or view the list archives, please visit: http://list.valvesoftware.com/mailman/listinfo/hlds_linux
