Re: [squid-users] Evaluating SQUID performance
On 25/07/2013 6:30 p.m., John Joseph wrote: HI All My Sincere thanks to Amos,Eliezer,Firas,Henrik Based on your advice, I have decided to try out squidclient calamaris msar web-polygraph Right now trying with squidclient, trying out its options and trying to understand. FYI: there is not much squidclient options related to what you are needing. It is just a fetch client like wget - but able to pull out Squid cachemgr self-reports of the performance. I have another idea which sprang up, I plan to document my test and write a How to do, aiming not for squid proffesionals, but aiming for people like me, who had just tried out squid That would be a great addition to our benchmarking pages thank you. All the proper external performance benchmarking references we have so far are so old they are almost embarassing to point people at. Amos
Re: [squid-users] Evaluating SQUID performance
HI All My Sincere thanks to Amos,Eliezer,Firas,Henrik Based on your advice, I have decided to try out squidclient calamaris msar web-polygraph Right now trying with squidclient, trying out its options and trying to understand. I have another idea which sprang up, I plan to document my test and write a How to do, aiming not for squid proffesionals, but aiming for people like me, who had just tried out squid Thanks to all for the Advice and tips Great mailing list thanks Joseph John
Re: [squid-users] Evaluating SQUID performance
On 25/07/2013 9:07 a.m., Golden Shadow wrote: Hi Amos, Sorry if I provided inaccurate information, I just based my answer on the following: http://wiki.squid-cache.org/SquidFaq/InnerWorkings#Why_is_my_cache.27s_inbound_traffic_equal_to_the_outbound_traffic.3F Perhaps I just got that wrong! Ah. That article is about measurement mistakes identifying inbound/outbound making traffic counters look different from the saving HIT ratio records. Or mistaking the request hit ratio for the byte ratio. Amos
Re: [squid-users] Evaluating SQUID performance
On 07/25/2013 12:07 AM, Golden Shadow wrote: > Hi Amos, > > Sorry if I provided inaccurate information, I just based my answer on the > following: > > http://wiki.squid-cache.org/SquidFaq/InnerWorkings#Why_is_my_cache.27s_inbound_traffic_equal_to_the_outbound_traffic.3F > > Perhaps I just got that wrong! Bandwidth from the admin point of view is in the logs but from the OS level it's including DNS HTTP RPC others... if you will run a TCPDUMP dump you will see what I am talking about. But then again if you analyse the logs you see that it saves you traffic even if you see on the network interface other traffic try to rethink it since you can try squid by caching one file but then understand that in a 200Mbit per sec traffic you would not see the bytes that squid saves and saves the server from leaving too many connections open with no reason. It's ok since you and me are right it's just that I described it in a more detailed way just in case. Eliezer > > > Best regards, > Firas > > > > - Original Message - > From: Amos Jeffries > To: squid-users@squid-cache.org > Cc: > Sent: Wednesday, July 24, 2013 10:02 PM > Subject: Re: [squid-users] Evaluating SQUID performance > > On 25/07/2013 1:44 a.m., Golden Shadow wrote: >> Hi, >> >> You can show them the hits statistics you get from the cache manager using: >> squidclient mgr:info >> >> However, having a hit ratio of 30% let's say does not necessarily mean that >> squid would save you 30% of bandwidth. > > On the contrary. That is exactly what "Byte HIT-ratio" means: that in > the last 5 min or 60 min (whichever it appears in) it has *already* > saved that much % of upstream bandwidth. > > Amos >
Re: [squid-users] Evaluating SQUID performance
Hi Amos, Sorry if I provided inaccurate information, I just based my answer on the following: http://wiki.squid-cache.org/SquidFaq/InnerWorkings#Why_is_my_cache.27s_inbound_traffic_equal_to_the_outbound_traffic.3F Perhaps I just got that wrong! Best regards, Firas - Original Message - From: Amos Jeffries To: squid-users@squid-cache.org Cc: Sent: Wednesday, July 24, 2013 10:02 PM Subject: Re: [squid-users] Evaluating SQUID performance On 25/07/2013 1:44 a.m., Golden Shadow wrote: > Hi, > > You can show them the hits statistics you get from the cache manager using: > squidclient mgr:info > > However, having a hit ratio of 30% let's say does not necessarily mean that > squid would save you 30% of bandwidth. On the contrary. That is exactly what "Byte HIT-ratio" means: that in the last 5 min or 60 min (whichever it appears in) it has *already* saved that much % of upstream bandwidth. Amos
Re: [squid-users] Evaluating SQUID performance
On 07/24/2013 10:11 PM, Amos Jeffries wrote: > On 25/07/2013 7:08 a.m., Eliezer Croitoru wrote: >> On 07/24/2013 10:02 PM, Amos Jeffries wrote: >>> On 25/07/2013 1:44 a.m., Golden Shadow wrote: Hi, You can show them the hits statistics you get from the cache manager using: squidclient mgr:info However, having a hit ratio of 30% let's say does not necessarily mean that squid would save you 30% of bandwidth. >>> On the contrary. That is exactly what "Byte HIT-ratio" means: that in >>> the last 5 min or 60 min (whichever it appears in) it has *already* >>> saved that much % of upstream bandwidth. >>> >>> Amos >>> >> Would squid change the expiration headers? >> it means that IMS requests will accrue more frequently and it means that >> the bandwidth saving is much more then most logs will show. > > Squid should not be, unless the refresh responded with new ones. In > which case those new ones get passed on and if possible the existign > cache updated (see bug #7 though). > > Amos Have seen and responded. Eliezer
Re: [squid-users] Evaluating SQUID performance
On 25/07/2013 7:08 a.m., Eliezer Croitoru wrote: On 07/24/2013 10:02 PM, Amos Jeffries wrote: On 25/07/2013 1:44 a.m., Golden Shadow wrote: Hi, You can show them the hits statistics you get from the cache manager using: squidclient mgr:info However, having a hit ratio of 30% let's say does not necessarily mean that squid would save you 30% of bandwidth. On the contrary. That is exactly what "Byte HIT-ratio" means: that in the last 5 min or 60 min (whichever it appears in) it has *already* saved that much % of upstream bandwidth. Amos Would squid change the expiration headers? it means that IMS requests will accrue more frequently and it means that the bandwidth saving is much more then most logs will show. Squid should not be, unless the refresh responded with new ones. In which case those new ones get passed on and if possible the existign cache updated (see bug #7 though). Amos
Re: [squid-users] Evaluating SQUID performance
On 07/24/2013 10:02 PM, Amos Jeffries wrote: > On 25/07/2013 1:44 a.m., Golden Shadow wrote: >> Hi, >> >> You can show them the hits statistics you get from the cache manager >> using: >> squidclient mgr:info >> >> However, having a hit ratio of 30% let's say does not necessarily mean >> that squid would save you 30% of bandwidth. > > On the contrary. That is exactly what "Byte HIT-ratio" means: that in > the last 5 min or 60 min (whichever it appears in) it has *already* > saved that much % of upstream bandwidth. > > Amos > Would squid change the expiration headers? it means that IMS requests will accrue more frequently and it means that the bandwidth saving is much more then most logs will show. Eliezer
Re: [squid-users] Evaluating SQUID performance
On 25/07/2013 1:44 a.m., Golden Shadow wrote: Hi, You can show them the hits statistics you get from the cache manager using: squidclient mgr:info However, having a hit ratio of 30% let's say does not necessarily mean that squid would save you 30% of bandwidth. On the contrary. That is exactly what "Byte HIT-ratio" means: that in the last 5 min or 60 min (whichever it appears in) it has *already* saved that much % of upstream bandwidth. Amos
Re: [squid-users] Evaluating SQUID performance
Squid dosn't just offer bandwidth saving if you ask me. Squid can sit between two network interfaces or the net and the client and be a "proxy" which can handle much traffic and packets then other network equipment exists. if it works on LINUX or any other OS affect it in many ways that normally you would not notice unless you will do a tcpdump and see how much traffic get drops in the kernel level on a loaded system.(try it let say on a small cisco that do not fit the network traffic load). The hit ratio is one of the aims of a proxy and specially a cache proxy server. You can use calamaris to analyse the log in a schematic way but since there is a traffic flow, once you understand the traffic flow you will see different things that happen which you would normally wouldn't. A real-world scenario is that the client has good network traffic flow from his machine to the proxy but the network from the proxy can handle more traffic. In the above case once you will see in tcpudmp how the traffic goes from one side to the other and some if not many network issues are being proccessed and handled by the proxy server and not the client. Which means the proxy coordinate the client and the site networks which for example The Linux kernel well known to handle better then couple alternatives. Eliezer On 07/24/2013 05:19 PM, Henrik Lidström wrote: > On 07/24/13 12:48, John Joseph wrote: >> >> Hi >> How could I do a test on squid server and check the performance on the >> bandwidth saved. >> Is there any tool for the same. >> I know squid can save bandwith, but I want to convince others with proof >> Guidance and Advice requested >> thanks >> Joseph John >> > You could test calamaris V.2.99 logparser if that fits your need? > http://cord.de/tools/squid/calamaris/ > > Example from my home: > > # Summary > Calamaris statistics > - -- > -- > lines parsed: lines 157608 > invalid lines: lines 1 > parse time: sec 6 > parse speed: lines/sec 26268 > - -- > -- > Proxy statistics > - -- > -- > Total amount: requests 157608 > unique hosts/users: hosts 24 > Total Bandwidth: Byte 14063M > Proxy efficiency (HIT [kB/sec] / DIRECT [kB/sec]): factor 9.15 > Average speed increase: > % 29.71 > TCP response time of 100% requests: msec322 > - -- > -- > Cache statistics > - -- > -- > Total amount cached: requests 85761 > Request hit rate: > % 54.41 > Bandwidth savings: Byte 3616M > Bandwidth savings in Percent (Byte hit rate): > % 25.71 > Average cached object size: Byte 44214 > Average direct object size: Byte 152472 > Average object size: Byte 93564 > - -- > -- > > > /Henrik
Re: [squid-users] Evaluating SQUID performance
On 07/24/13 12:48, John Joseph wrote: Hi How could I do a test on squid server and check the performance on the bandwidth saved. Is there any tool for the same. I know squid can save bandwith, but I want to convince others with proof Guidance and Advice requested thanks Joseph John You could test calamaris V.2.99 logparser if that fits your need? http://cord.de/tools/squid/calamaris/ Example from my home: # Summary Calamaris statistics - -- -- lines parsed: lines 157608 invalid lines: lines 1 parse time: sec 6 parse speed: lines/sec 26268 - -- -- Proxy statistics - -- -- Total amount: requests 157608 unique hosts/users: hosts 24 Total Bandwidth: Byte 14063M Proxy efficiency (HIT [kB/sec] / DIRECT [kB/sec]): factor 9.15 Average speed increase: % 29.71 TCP response time of 100% requests: msec322 - -- -- Cache statistics - -- -- Total amount cached: requests 85761 Request hit rate: % 54.41 Bandwidth savings: Byte 3616M Bandwidth savings in Percent (Byte hit rate): % 25.71 Average cached object size: Byte 44214 Average direct object size: Byte 152472 Average object size: Byte 93564 - -- -- /Henrik
Re: [squid-users] Evaluating SQUID performance
Hi, You can show them the hits statistics you get from the cache manager using: squidclient mgr:info However, having a hit ratio of 30% let's say does not necessarily mean that squid would save you 30% of bandwidth. Regards, Firas - Original Message - From: John Joseph To: "squid-users@squid-cache.org" Cc: Sent: Wednesday, July 24, 2013 1:48 PM Subject: [squid-users] Evaluating SQUID performance Hi How could I do a test on squid server and check the performance on the bandwidth saved. Is there any tool for the same. I know squid can save bandwith, but I want to convince others with proof Guidance and Advice requested thanks Joseph John
Re: [squid-users] Evaluating SQUID performance
You could go and do a simple demonstration, download a 300mb+ file one one machine, then download the same file on a machine sitting next to it, let em see how fast it is coming from the cache. - Original Message - From: "John Joseph" To: Sent: Wednesday, July 24, 2013 3:48 AM Subject: [squid-users] Evaluating SQUID performance Hi How could I do a test on squid server and check the performance on the bandwidth saved. Is there any tool for the same. I know squid can save bandwith, but I want to convince others with proof Guidance and Advice requested thanks Joseph John