Re: [PERFORM] Samsung 32GB SATA SSD tested
On Tue, Jul 22, 2008 at 5:32 PM, Scott Marlowe <[EMAIL PROTECTED]> wrote: > On Tue, Jul 22, 2008 at 6:04 PM, Jeffrey W. Baker <[EMAIL PROTECTED]> wrote: > >> Strangely the RAID controller behaves badly on the TPC-B workload. It >> is faster than disk, but not by a lot, and it's much slower than the >> other flash configurations. The read/write benchmark did not vary when >> changing the number of clients between 1 and 8. I suspect this is some >> kind of problem with Areca's kernel driver or firmware. > > Are you still using the 2.6.18 kernel for testing, or have you > upgraded to something like 2.6.22. I've heard many good things about > the areca driver in that kernel version. These tests are being run with the CentOS 5 kernel, which is 2.6.18. The ioDrive driver is available for that kernel, and I want to keep the software constant to get comparable results. I put the Samsung SSD in my laptop, which is a Core 2 Duo @ 2.2GHz with ICH9 SATA port and kernel 2.6.24, and it scored about 525 on R/W pgbench. > This sounds like an interesting development I'll have to keep track > of. In a year or two I might be replacing 16 disk arrays with SSD > drives... I agree, it's definitely an exciting development. I have yet to determine whether the SSDs have good properties for production operations, but I'm learning. -jwb -- Sent via pgsql-performance mailing list (pgsql-performance@postgresql.org) To make changes to your subscription: http://www.postgresql.org/mailpref/pgsql-performance
Re: [PERFORM] Perl/DBI vs Native
On Tue, Jul 22, 2008 at 9:48 AM, Greg Sabino Mullane <[EMAIL PROTECTED]> wrote: >> In case someone is wondering, the way to force DBI to use unix >> sockets is by not specifying a host and port in the connect call. > > Actually, the host defaults to the local socket. Using the port > may still be needed: if you leave it out, it simply uses the default > value (5432) if left out. Thus, for most purposes, just leaving > the host out is enough to cause a socket connection on the default > port. For the further illumination of the historical record, the best practice here is probably to use the pg_service.conf file, which may or may not live in /etc depending on your operating system. Then you can connect in DBI using dbi:Pg:service=whatever, and change the definition of "whatever" in pg_service.conf. This has the same semantics as PGSERVICE=whatever when using psql. It's a good idea to keep these connection details out of your program code. -jwb -- Sent via pgsql-performance mailing list (pgsql-performance@postgresql.org) To make changes to your subscription: http://www.postgresql.org/mailpref/pgsql-performance
Re: [PERFORM] 3ware vs Areca
On Mon, Jul 14, 2008 at 9:13 PM, Greg Smith <[EMAIL PROTECTED]> wrote: > On Fri, 11 Jul 2008, Jeffrey Baker wrote: > >> Their firmware is, frankly, garbage. In more than one instance we >> have had the card panic when a disk fails, which is obviously counter >> to the entire purpose of a RAID. We finally removed the Areca >> controllers from our database server and replaced them with HP P800s. > > Can you give a bit more detail here? If what you mean is that the driver > for the card generated an OS panic when a drive failed, that's not > necessarily the firmware at all. I know I had problems with the Areca cards > under Linux until their driver went into the mainline kernel in 2.6.19, all > kinds of panics under normal conditions. Haven't seen anything like that > with later Linux kernels or under Solaris 10, but then again I haven't had a > disk failure yet either. Well, it is difficult to tell if the fault is with the hardware or the software. No traditional kernel panic has been observed. But most recently in my memory we had an Areca HBA which, when one of its WD RE-2 disks failed, completely stopped responding to both the command line and the web management interface. Then, i/o to that RAID became increasingly slower, and slower, until it stopped serving i/o at all. At that point it was not relevant that the machine was technically still running. We have another Areca HBA that starts throwing errors up the SCSI stack if it runs for more than 2 months at a time. We have to reboot it on a schedule to keep it running. -jwb -- Sent via pgsql-performance mailing list (pgsql-performance@postgresql.org) To make changes to your subscription: http://www.postgresql.org/mailpref/pgsql-performance
Re: [PERFORM] 3ware vs Areca
On Tue, Jul 15, 2008 at 8:17 AM, Greg Smith <[EMAIL PROTECTED]> wrote: > On Tue, 15 Jul 2008, Jeffrey Baker wrote: > >> But most recently in my memory we had an Areca HBA which, when one of its >> WD RE-2 disks failed, completely stopped responding to both the command line >> and the web management interface. > > What operating system/kernel version are you using on these systems? Debian "etch", which has a 2.6.18 kernel. I have contacted Areca support (as well as the linux-scsi mailing list) and their responses are usually either 1) upgrade the driver and/or firmware even though I have the latest drivers and firmware, or 2) vague statements about the disk being incompatible with the controller. -jwb -- Sent via pgsql-performance mailing list (pgsql-performance@postgresql.org) To make changes to your subscription: http://www.postgresql.org/mailpref/pgsql-performance
Re: [PERFORM] 3ware vs Areca
On Fri, Jul 11, 2008 at 12:21 PM, Greg Smith <[EMAIL PROTECTED]> wrote: > On Fri, 11 Jul 2008, Jeff wrote: > >> I've got a couple boxes with some 3ware 9550 controllers, and I'm less >> than pleased with performance on them.. Sequential access is nice, but start >> seeking around and you kick it in the gut. (I've found posts on the >> internets about others having similar issues). > > Yeah, there's something weird about those controllers, maybe in how stuff > flows through the cache, that makes them slow in a lot of situations. The > old benchmarks at > http://tweakers.net/reviews/557/21/comparison-of-nine-serial-ata-raid-5-adapters-pagina-21.html > show their cards acting badly in a lot of situations and I haven't seen > anything else since vindicating the 95XX models from them. > >> My last box with a 3ware I simply had it in jbod mode and used sw raid and >> it smoked the hw. > > That is often the case no matter which hardware controller you've got, > particularly in more complicated RAID setups. You might want to consider > that a larger lesson rather than just a single data point. > >> Anyway, anybody have experience in 3ware vs Areca - I've heard plenty of >> good anecdotal things that Areca is much better, just wondering if anybody >> here has firsthand experience.It'll be plugged into about 8 10k rpm sata >> disks. > > Areca had a pretty clear performance lead for a while there against 3ware's > 3500 series, but from what I've been reading I'm not sure that is still true > in the current generation of products. Check out the pages starting at > http://www.tomshardware.com/reviews/SERIAL-RAID-CONTROLLERS-AMCC,1738-12.html > for example, where the newer Areca 1680ML card just gets crushed at all > kinds of workloads by the AMCC 3ware 9690SA. I think the 3ware 9600 series > cards have achieved or exceeded what Areca's 1200 series was capable of, > while Areca's latest generation has slipped a bit from the previous one. >From my experience, the Areca controllers are difficult to operate. Their firmware is, frankly, garbage. In more than one instance we have had the card panic when a disk fails, which is obviously counter to the entire purpose of a RAID. We finally removed the Areca controllers from our database server and replaced them with HP P800s. -jwb -- Sent via pgsql-performance mailing list (pgsql-performance@postgresql.org) To make changes to your subscription: http://www.postgresql.org/mailpref/pgsql-performance
Re: [PERFORM] Practical upper limits of pgbench read/write tps with 8.3
On Mon, Jul 7, 2008 at 3:22 PM, Greg Smith <[EMAIL PROTECTED]> wrote: > On Mon, 7 Jul 2008, Jeffrey Baker wrote: > >> On the single 2.2GHz Athlon, the maximum tps seems to be 1450...what's the >> bottleneck? Is PG lock-bound? > > It can become lock-bound if you don't make the database scale significantly > larger than the number of clients, but that's probably not your problem. > The pgbench client driver program itself is pretty CPU intensive and can > suffer badly from kernel issues. I am unsurprised you can only hit 1450 > with a single CPU. On systems with multiple CPUs where the single CPU > running the pgbench client is much faster than your 2.2GHz Athlon, you'd > probably be able to get a few thousand TPS, but eventually the context > switching of the client itself can become a bottleneck. On a 2GHz Core 2 Duo the best tps achieved is 2300, with -c 8. pgbench itself gets around 10% of the CPU (user + sys for pgbench is 7s of 35s wall clock time, or 70 CPU-seconds, thus 10%). I suppose you could still blame it on ctxsw between pgbench and pg itself, but the results are not better with pgbench on another machine cross-connected with gigabit ethernet. -jwb -- Sent via pgsql-performance mailing list (pgsql-performance@postgresql.org) To make changes to your subscription: http://www.postgresql.org/mailpref/pgsql-performance
[PERFORM] Practical upper limits of pgbench read/write tps with 8.3
I'm spending a third day testing with the ioDrive, and it occurred to me that I should normalize my tests by mounting the database on a ramdisk. The results were surprisingly low. On the single 2.2GHz Athlon, the maximum tps seems to be 1450. This is achieved with a single connection. I/O rates to and from the ramdisk never exceed 50MB/s on a one-minute average. With the flash device on the same benchmark, the tps rate is 1350, meaning that as far as PostgreSQL is concerned, on this machine, the flash device achieves 90% of the best possible performance. Question being: what's the bottleneck? Is PG lock-bound? -jwb -- Sent via pgsql-performance mailing list (pgsql-performance@postgresql.org) To make changes to your subscription: http://www.postgresql.org/mailpref/pgsql-performance
Re: [PERFORM] Fusion-io ioDrive
On Mon, Jul 7, 2008 at 6:08 AM, Merlin Moncure <[EMAIL PROTECTED]> wrote: > On Sat, Jul 5, 2008 at 2:41 AM, Jeffrey Baker <[EMAIL PROTECTED]> wrote: >>>Service Time Percentile, millis >>>R/W TPS R-O TPS 50th 80th 90th 95th >>> RAID 182 673 18 32 42 64 >>> Fusion971 4792 8 9 10 11 >> >> Someone asked for bonnie++ output: >> >> Block output: 495MB/s, 81% CPU >> Block input: 676MB/s, 93% CPU >> Block rewrite: 262MB/s, 59% CPU >> >> Pretty respectable. In the same ballpark as an HP MSA70 + P800 with >> 25 spindles. > > You left off the 'seeks' portion of the bonnie++ results -- this is > actually the most important portion of the test. Based on your tps > #s, I'm expecting seeks equiv of about 10 10k drives in configured in > a raid 10, or around 1000-1500. They didn't publish any prices so > it's hard to say if this is 'cost competitive'. I left it out because bonnie++ reports it as "+" i.e. greater than or equal to 10 per second. -jwb -- Sent via pgsql-performance mailing list (pgsql-performance@postgresql.org) To make changes to your subscription: http://www.postgresql.org/mailpref/pgsql-performance
Re: [PERFORM] Fusion-io ioDrive
On Tue, Jul 1, 2008 at 5:18 PM, Jeffrey Baker <[EMAIL PROTECTED]> wrote: > I recently got my hands on a device called ioDrive from a company > called Fusion-io. The ioDrive is essentially 80GB of flash on a PCI > card. [...] >Service Time Percentile, millis >R/W TPS R-O TPS 50th 80th 90th 95th > RAID 182 673 18 32 42 64 > Fusion971 4792 8 9 10 11 Essentially the same benchmark, but on a quad Xeon 2GHz with 3GB main memory, and the scale factor of 300. Really all we learn from this exercise is the sheer futility of throwing CPU at PostgreSQL. R/W TPS: 1168 R-O TPS: 6877 Quadrupling the CPU resources and tripling the RAM results in a 20% or 44% performance increase on read/write and read-only loads, respectively. The system loafs along with 2-3 CPUs completely idle, although oddly iowait is 0%. I think the system is constrained by context switch, which is tens of thousands per second. This is a problem with the ioDrive software, not with pg. Someone asked for bonnie++ output: Block output: 495MB/s, 81% CPU Block input: 676MB/s, 93% CPU Block rewrite: 262MB/s, 59% CPU Pretty respectable. In the same ballpark as an HP MSA70 + P800 with 25 spindles. -jwb -- Sent via pgsql-performance mailing list (pgsql-performance@postgresql.org) To make changes to your subscription: http://www.postgresql.org/mailpref/pgsql-performance
Re: [PERFORM] Fusion-io ioDrive
On Tue, Jul 1, 2008 at 6:17 PM, Andrej Ricnik-Bay <[EMAIL PROTECTED]> wrote: > On 02/07/2008, Jeffrey Baker <[EMAIL PROTECTED]> wrote: > >> Red Hat and its clones. The other problem is the 80GB model is too >> small to hold my entire DB, Although it could be used as a tablespace >> for some critical tables. But hey, it's fast. > And when/if it dies, please give us a rough guestimate of its > life-span in terms of read/write cycles. Sounds exciting, though! Yeah. The manufacturer rates it for 5 years in constant use. I remain skeptical. -- Sent via pgsql-performance mailing list (pgsql-performance@postgresql.org) To make changes to your subscription: http://www.postgresql.org/mailpref/pgsql-performance
[PERFORM] Fusion-io ioDrive
I recently got my hands on a device called ioDrive from a company called Fusion-io. The ioDrive is essentially 80GB of flash on a PCI card. It has its own driver for Linux completely outside of the normal scsi/sata/sas/fc block device stack, but from the user's perspective it behaves like a block device. I put the ioDrive in an ordinary PC with 1GB of memory, a single 2.2GHz AMD CPU, and an existing Areca RAID with 6 SATA disks and a 128MB cache. I tested the device with PostgreSQL 8.3.3 on Centos 5.3 x86_64 (Linux 2.6.18). The pgbench database was initialized with scale factor 100. Test runs were performed with 8 parallel connections (-c 8), both read-only (-S) and read-write. PostgreSQL itself was configured with 256MB of shared buffers and 32 checkpoint segments. Otherwise the configuration was all defaults. In the following table, the "RAID" configuration has the xlogs on a RAID 0 of 2 10krpm disks with ext2, and the heap is on a RAID 0 of 4 7200rpm disks with ext3. The "Fusion" configuration has everything on the ioDrive with xfs. I tried the ioDrive with ext2 and ext3 but it didn't seem to make any difference. Service Time Percentile, millis R/W TPS R-O TPS 50th 80th 90th 95th RAID 182 673 18 32 42 64 Fusion971 4792 8 9 10 11 Basically the ioDrive is smoking the RAID. The only real problem with this benchmark is that the machine became CPU-limited rather quickly. During the runs with the ioDrive, iowait was pretty well zero, with user CPU being about 75% and system getting about 20%. Now, I will say a couple of other things. The Linux driver for this piece of hardware is pretty dodgy. Sub-alpha quality actually. But they seem to be working on it. Also there's no driver for OpenSolaris, Mac OS X, or Windows right now. In fact there's not even anything available for Debian or other respectable Linux distros, only Red Hat and its clones. The other problem is the 80GB model is too small to hold my entire DB, Although it could be used as a tablespace for some critical tables. But hey, it's fast. I'm going to put this board into my 8-way Xeon to see if it goes any faster with more CPU available. I'd be interested in hearing experiences with other flash storage devices, SSDs, and that type of thing. So far, this is the fastest hardware I've seen for the price. -jwb -- Sent via pgsql-performance mailing list (pgsql-performance@postgresql.org) To make changes to your subscription: http://www.postgresql.org/mailpref/pgsql-performance
Re: [PERFORM] Quad Xeon or Quad Opteron?
On Fri, May 23, 2008 at 3:41 AM, Andrzej Zawadzki <[EMAIL PROTECTED]> wrote: > Hello, > > We're planning new production server for PostgreSQL and I'm wondering > which processor (or even platform) will be better: Quad Xeon or Quad > Opteron (for example SUN now has a new offer Sun Fire X4440 x64). > > When I was buying my last database server, then SUN v40z was a really > very good choice (Intel's base server was slower). This v40z still works > pretty good but I need one more. > > AFAIK Intel made some changes in chipset but... is this better then AMD > HyperTransport and Direct Connect Architecture from database point of > view? How about L3 cache - is this important for performance? > Intel's chipset is still broken when using dual sockets and quad core processors. The problem manifests itself as excessive cache line bouncing. In my opinion the best bang/buck combo on the CPU side is the fastest dual-core Xeon CPUs you can find. You get excellent single-thread performance and you still have four processors, which was a fantasy for most people only 5 years ago. In addition you can put a ton of memory in the new Xeon machines. 64GB is completely practical. I still run several servers on Opterons but in my opinion they don't make sense right now unless you truly need the CPU parallelism. -jwb
Re: [PERFORM] Update performance degrades over time
On Wed, May 14, 2008 at 6:31 PM, Subbiah Stalin-XCGF84 <[EMAIL PROTECTED]> wrote: > Hi All, > > We are doing some load tests with our application running postgres 8.2.4. At > times we see updates on a table taking longer (around > 11-16secs) than expected sub-second response time. The table in question is > getting updated constantly through the load tests. In checking the table > size including indexes, they seem to be bloated got it confirmed after > recreating it (stats below). We have autovacuum enabled with default > parameters. I thought autovaccum would avoid bloating issues but looks like > its not aggressive enough. Wondering if table/index bloating is causing > update slowness in over a period of time. Any ideas how to troubleshoot this > further. Sometimes it is necessary to not only VACUUM, but also REINDEX. If your update changes an indexed column to a new, distinct value, you can easily get index bloat. Also, you should check to see if you have any old, open transactions on the same instance. If you do, it's possible that VACUUM will have no beneficial effect. -jwb -- Sent via pgsql-performance mailing list (pgsql-performance@postgresql.org) To make changes to your subscription: http://www.postgresql.org/mailpref/pgsql-performance
Re: [PERFORM] SELECT 'DBD::Pg ping test'
On Wed, Apr 23, 2008 at 12:19 AM, sathiya psql <[EMAIL PROTECTED]> wrote: > Hi All, > > This query is being executed nearly a million times >SELECT 'DBD::Pg ping test' Something in your Perl application is use $dbh->ping(). See perldoc DBI. It's possible that this is happening under the hood, because your application is using connect_cached() instead of connect(). -jwb -- Sent via pgsql-performance mailing list (pgsql-performance@postgresql.org) To make changes to your subscription: http://www.postgresql.org/mailpref/pgsql-performance
Re: [PERFORM] 3-days-long vacuum of 20GB table
On Fri, Apr 18, 2008 at 10:34 AM, Jeffrey Baker <[EMAIL PROTECTED]> wrote: > > On Fri, Apr 18, 2008 at 10:32 AM, Jeffrey Baker <[EMAIL PROTECTED]> wrote: > > > > On Fri, Apr 18, 2008 at 10:03 AM, Tom Lane <[EMAIL PROTECTED]> wrote: > > > "Jeffrey Baker" <[EMAIL PROTECTED]> writes: > > > > This autovacuum has been hammering my server with purely random i/o > > > > for half a week. The table is only 20GB and the i/o subsystem is > good > > > > for 250MB/s sequential and a solid 5kiops. When should I expect it > to > > > > end (if ever)? > > > > > > What have you got maintenance_work_mem set to? Which PG version > > > exactly? > > > > This is 8.1.9 on Linux x86_64, > > > > # show maintenance_work_mem ; > > maintenance_work_mem > > -- > > 16384 > > That appears to be the default. I will try increasing this. Can I > increase it globally from a single backend, so that all other backends > pick up the change, or do I have to restart the instance? I increased it to 1GB, restarted the vacuum, and system performance seems the same. The root of the problem, that an entire CPU is in the iowait state and the storage device is doing random i/o, is unchanged: r b swpd free buff cache si sobibo in cs us sy id wa 1 1 30328 53632 60 691471600 904 2960 1216 4720 1 1 74 23 0 1 30328 52492 60 691603600 1152 1380 948 3637 0 0 75 24 0 1 30328 49600 60 691768000 1160 1420 1055 4191 1 1 75 24 0 1 30328 49404 60 691900000 1048 1308 1133 5054 2 2 73 23 0 1 30328 47844 60 692109600 1552 1788 1002 3701 1 1 75 23 At that rate it will take a month. Compare the load generated by create table foo as select * from bar: r b swpd free buff cache si sobibo in cs us sy id wa 2 2 30328 46580 60 691102400 145156 408 2006 10729 52 8 17 23 3 1 30328 46240 60 690097600 133312 224 1834 10005 23 12 42 23 1 3 30328 60700 60 690205600 121480 172 1538 10629 22 14 32 32 1 2 30328 49520 60 691420400 122344 256 1408 14374 13 17 41 28 1 2 30328 47844 60 691596000 127752 248 1313 9452 16 15 42 27 That's rather more like it. I guess I always imagined that VACUUM was a sort of linear process, not random, and that it should proceed at sequential scan speeds. -jwb -- Sent via pgsql-performance mailing list (pgsql-performance@postgresql.org) To make changes to your subscription: http://www.postgresql.org/mailpref/pgsql-performance
Re: [PERFORM] 3-days-long vacuum of 20GB table
On Fri, Apr 18, 2008 at 10:32 AM, Jeffrey Baker <[EMAIL PROTECTED]> wrote: > > On Fri, Apr 18, 2008 at 10:03 AM, Tom Lane <[EMAIL PROTECTED]> wrote: > > "Jeffrey Baker" <[EMAIL PROTECTED]> writes: > > > This autovacuum has been hammering my server with purely random i/o > > > for half a week. The table is only 20GB and the i/o subsystem is good > > > for 250MB/s sequential and a solid 5kiops. When should I expect it to > > > end (if ever)? > > > > What have you got maintenance_work_mem set to? Which PG version > > exactly? > > This is 8.1.9 on Linux x86_64, > > # show maintenance_work_mem ; > maintenance_work_mem > -- > 16384 That appears to be the default. I will try increasing this. Can I increase it globally from a single backend, so that all other backends pick up the change, or do I have to restart the instance? -jwb -- Sent via pgsql-performance mailing list (pgsql-performance@postgresql.org) To make changes to your subscription: http://www.postgresql.org/mailpref/pgsql-performance
Re: [PERFORM] 3-days-long vacuum of 20GB table
On Fri, Apr 18, 2008 at 10:03 AM, Tom Lane <[EMAIL PROTECTED]> wrote: > "Jeffrey Baker" <[EMAIL PROTECTED]> writes: > > This autovacuum has been hammering my server with purely random i/o > > for half a week. The table is only 20GB and the i/o subsystem is good > > for 250MB/s sequential and a solid 5kiops. When should I expect it to > > end (if ever)? > > What have you got maintenance_work_mem set to? Which PG version > exactly? This is 8.1.9 on Linux x86_64, # show maintenance_work_mem ; maintenance_work_mem -- 16384 -- Sent via pgsql-performance mailing list (pgsql-performance@postgresql.org) To make changes to your subscription: http://www.postgresql.org/mailpref/pgsql-performance
[PERFORM] 3-days-long vacuum of 20GB table
This autovacuum has been hammering my server with purely random i/o for half a week. The table is only 20GB and the i/o subsystem is good for 250MB/s sequential and a solid 5kiops. When should I expect it to end (if ever)? current_query: VACUUM reuters.value query_start: 2008-04-15 20:12:48.806885-04 think=# select * from pg_class where relname = 'value'; -[ RECORD 1 ]--+- relname| value relfilenode| 191425 relpages | 1643518 reltuples | 1.37203e+08 # find -name 191425\* ./16579/191425 ./16579/191425.1 ./16579/191425.10 ./16579/191425.11 ./16579/191425.12 ./16579/191425.13 ./16579/191425.14 ./16579/191425.15 ./16579/191425.16 ./16579/191425.17 ./16579/191425.18 ./16579/191425.19 ./16579/191425.2 ./16579/191425.3 ./16579/191425.4 ./16579/191425.5 ./16579/191425.6 ./16579/191425.7 ./16579/191425.8 ./16579/191425.9 # vmstat 1 procs ---memory-- ---swap-- -io -system-- cpu r b swpd free buff cache si sobibo in cs us sy id wa 0 1 30336 46264 60 788235600 250 29911 6 2 87 5 0 1 30336 47412 60 788130800 289648 944 4861 3 2 71 24 0 2 30336 46696 60 788218800 816 4 840 5019 1 0 75 24 0 1 30336 49228 60 787986800 1888 164 971 5687 1 1 74 24 0 1 30336 49688 60 787891600 264048 1047 5751 1 0 75 23 autovacuum | on autovacuum_vacuum_cost_delay| -1 autovacuum_vacuum_cost_limit| -1 vacuum_cost_delay | 0 vacuum_cost_limit | 200 vacuum_cost_page_dirty | 20 vacuum_cost_page_hit| 1 vacuum_cost_page_miss | 10 -- Sent via pgsql-performance mailing list (pgsql-performance@postgresql.org) To make changes to your subscription: http://www.postgresql.org/mailpref/pgsql-performance
Re: [PERFORM] seq scan issue...
On Thu, Apr 17, 2008 at 11:24 AM, kevin kempter <[EMAIL PROTECTED]> wrote: > Hi List; > > I have a large tble (playback_device) with 6million rows in it. The > aff_id_tmp1 table has 600,000 rows. > - why am I still getting a seq scan ? > You're selecting almost all the rows in the product of aff_id_tmp1 * playback_fragment. A sequential scan will be far faster than an index scan. You can prove this to yourself using 'set enable_seqscan to false' and running the query again. It should be much slower. -- Sent via pgsql-performance mailing list (pgsql-performance@postgresql.org) To make changes to your subscription: http://www.postgresql.org/mailpref/pgsql-performance
Re: [PERFORM] Strange behavior: pgbench and new Linux kernels
On Thu, Apr 17, 2008 at 12:58 AM, Greg Smith <[EMAIL PROTECTED]> wrote: > So in the case of this simple benchmark, I see an enormous performance > regression from the newest Linux kernel compared to a much older one. This has been discussed recently on linux-kernel. It's definitely a regression. Instead of getting a nice, flat overload behavior when the # of busy threads exceeds the number of CPUs, you get the declining performance you noted. Poor PostgreSQL scaling on Linux 2.6.25-rc5 (vs 2.6.22) http://marc.info/?l=linux-kernel&m=120521826111587&w=2 -jwb -- Sent via pgsql-performance mailing list (pgsql-performance@postgresql.org) To make changes to your subscription: http://www.postgresql.org/mailpref/pgsql-performance
Re: [PERFORM] Anybody using the Dell Powervault MD3000 array?
On Wed, Apr 16, 2008 at 1:20 PM, Joshua D. Drake <[EMAIL PROTECTED]> wrote: > On Wed, 16 Apr 2008 16:17:10 -0400 > > > > On Wed, Apr 16, 2008 at 4:15 PM, Jeffrey Baker <[EMAIL PROTECTED]> > > wrote: > > > > > Thinking about buying the Powervault MD3000 SAS array with 15 15k > > > 300GB disks for use as a postgres tablespace. Is anyone using these > > > (or other LSI/Engenio rebadge jobs?). I'm interested in hearing > > > about performance of the array, and problems (if any) with Dell's > > > SAS HBA that comes bundled. Also interested in performance of the > > > maximum config of an MD3000 with two MD1000 shelves. > > > > > > -Jeff > > > "Gavin M. Roy" <[EMAIL PROTECTED]> wrote: > > > Might want to check out the HP MSA70 arrays. I've had better luck > > with them and you can get 25 drives in a smaller rack unit size. I > > had a bad experience with the MD3000 and now only buy MD1000's with > > Perc 6/e when I buy Dell. > > Good luck! > > > > I can second this. The MSA 70 is a great unit for the money. Thank you both. The MSA 70 looks like an ordinary disk shelf. What controllers do you use? Or, do you just go with a software RAID? -- Sent via pgsql-performance mailing list (pgsql-performance@postgresql.org) To make changes to your subscription: http://www.postgresql.org/mailpref/pgsql-performance
[PERFORM] Anybody using the Dell Powervault MD3000 array?
Thinking about buying the Powervault MD3000 SAS array with 15 15k 300GB disks for use as a postgres tablespace. Is anyone using these (or other LSI/Engenio rebadge jobs?). I'm interested in hearing about performance of the array, and problems (if any) with Dell's SAS HBA that comes bundled. Also interested in performance of the maximum config of an MD3000 with two MD1000 shelves. -Jeff -- Sent via pgsql-performance mailing list (pgsql-performance@postgresql.org) To make changes to your subscription: http://www.postgresql.org/mailpref/pgsql-performance
