Hi Many apologies that this is so late, I am guilty of not reading the Sun Managers FAQ, please accept my pologies for this, this list is a valuable source of information for me and a SUMMARY of the data received regarding my Performance Poll should have been posted. Again, apologies, I am new to the list and hope you can forgive me!! Problems with our SAN have been highlighted as being down to the controllers in the disk array, they are not powerful enough for the job and having been quoted #52k to replace them with 4 new controllers (each with 1GB cache), we have decided to look into alternative technology as we can get a new array for this kind of price. A summary of all responses that I received can be found below regarding the Performance Poll, I would like to thank everyone that replied to me, thank you! The SAN issues and Brocade issues were resolved via consultants using there own lab and hardware to duplicate our enviroment (PICK RDBMS mixed in with Windows volumes all on the same SAN all using RAID 5). Following is a copy and paste of all performance poll replies, hope they are of interest to you as much as they were to me! ------------------------------------------------------------------- Martin, Here are my results on V880 (4 x 900Mhz, 8GB) connected to two T3+'s each configured as follows: 1GB cache, 9 x 36GB 10K RPM, RAID 5 (7+1, 1 hotspare) write caching enabled. Note: this is not a partner pair. This box is a database server using Oracle and so some filesystems are mounted with Direct IO as Oracle does it's own caching. Hope these figures help. On a filesystem striped accross both T3's (mounted logging) dagda {1089}# time mkfile 1g test real 0m18.54s user 0m0.09s sys 0m8.70s dagda {1090}# iostat output: 25.2 3928.6 201.6 31387.6 0.0 18.5 0.0 4.7 0 99 c2t1d0 23.8 3911.2 190.4 31198.2 0.0 17.7 0.0 4.5 0 99 c4t1d0 On a filesystem mirrored across both T3's (mounted logging,forcedirectio) dagda {1093}# time mkfile 1g test real 1m14.48s user 0m0.03s sys 0m13.89s dagda {1094}# iostat for the T3s looked like this during it: 7.2 1831.4 57.6 14694.3 0.0 0.8 0.0 0.4 0 77 c2t1d0 7.4 1830.4 59.2 14677.3 0.0 0.8 0.0 0.4 0 78 c4t1d0 On another striped filesystem, this time mounted logging,forcedirectio real 1m7.60s user 0m0.05s sys 0m9.06s iostat: 0.0 1042.4 0.0 8329.0 0.0 0.4 0.0 0.4 0 42 c2t1d0 1.0 1042.8 8.0 8325.1 0.0 0.4 0.0 0.4 0 44 c4t1d0 ------------------------------------------------------------------- Hi Martin, Just thought I'd participate in your poll. I have a Sun Fire V880 - with 4 x 900Mhz CPU's and 4 GB RAM.Internal 6 x 73GB 10000RPM hard disks.Connected to an EMC400 SAN with 9 x 73GB 10000RPM disks.These disks are mirrored in twos using Disksuite, except for 3 which are RAID-5. The 'timex mkfile 1g test' command produced the following results for me. / root Mirror RAID-5Real 36.19 20.40 16.07User 0.11 0.13 0.10Sys 6.74 6.08 6.36 I don't have process accounting turned on. ------------------------------------------------------------------- Here are the numbers for a V880 (2 x 900Mhz, 4gb mem) Disks mirrored with disksuite r/s w/s kr/s kw/s wait actv wsvc_t asvc_t %w %b device 0.0 229.4 0.0 28749.0 0.0 5.3 0.0 23.0 0 73 c1t4d0 0.0 229.4 0.0 28749.0 0.0 5.3 0.0 23.3 0 77 c1t5d0 timex mkfile 1g test real 35.80 user 0.09 sys 6.09 and for a 4800 (4 x 900Mhz, 4gb mem) dual attached directly to an Hitachi 9960 r/s w/s kr/s kw/s wait actv wsvc_t asvc_t %w %b device 0.4 1603.6 3.2 12829.0 0.0 0.9 0.0 0.6 0 92 c4t500060E80000000000007B000000011Fd0 timex mkfile 1g test real 1:21.12 user 0.13 sys 7.67 ------------------------------------------------------------------- ROOT@db02# timex mkfile 1g /user28/test real 11.35 user 0.16 sys 10.65 V880 8cpu 32gb ram. Connectrix fiber switches tied to EMC Symmmetrix. ------------------------------------------------------------------- # timex -ops mkfile 1g test3 real 13.66 user 0.09 sys 5.95 That's a Sun Fire 280R, 2x 1.2GHz Sparc III, 8GB memory LSI E4400, RAID 5, 18x 73GB 10K RPM LSI 409190 FC cards Brocade 2250 FC switch ------------------------------------------------------------------- This is a V880/8*900MHz, 32Gb RAM, EMC cx400 array direct attached via Emulex LP9000s, Veritas 3.2, EMC ATF. 3512:/var/pkms/clust/logs-> timex mkfile 1g test real 13.33 user 0.06 sys 13.17 On the internal disks: 3512:/var/tmp-> timex mkfile 1g test real 1:13.57 user 0.20 sys 9.53 And the same thing on a V880, 8*750, EMC FC4700 1502 # timex mkfile 1g test real 23.41 user 0.06 sys 15.20 ---- original message ---- Hi V440, 8 GB RAM, 4x1 GHz, Solaris 8 7/03, 36 GB 10 krpm*4 3510 two controllers with 0.5 GB cache each, 36 GB 15 krpm*12, HW RAID5 (10+2 spare) V440 internal, mirror with disksuite: bash-2.03# timex mkfile 1g test real 38.69 user 0.09 sys 6.06 3510: bash-2.03# timex mkfile 1g test real 18.67 user 0.07 sys 4.28 Almost no load, just oracle and som web stuff idle at the time of test. ------------------------------------------------------------------- Hi, I've done a couple of tests with: 1. e250 with A1000 - real 1:17.82 2. e450 with hp-SAN - real 1.14.27 Make sure that you have a healthy battery in the controller.(or if you know that you have ups-protected the controller/disks, turn on the "cache without battery" option) I'm not sure that is your problem, but it could be. ------------------------------------------------------------------- Here are my results SF3800 (4x900MHz, 6GB, Solaris 8), Brocade 2400, EMC CLARiiON FC4700 Disks are a striped set of 3 mirrored pairs of 73GB/10k disks (hardware raid 1/0). Filesystem is VxFS timex mkfile 1g test real 9.62 user 0.00 sys 8.49 iostat results device r/s w/s kr/s kw/s wait actv svc_t %w %b atf2 0.0 501.2 0.0 31651.6 13.2 1.0 28.3 99 99 -------- Original Message -------- Subject: PERFORMANCE POLL Date: Mon, 09 Feb 2004 11:50:13 +0000 From: Martin Thorpe <Martin.Thorpe@DATAFORCE.co.uk> Reply-To: Martin.Thorpe@DATAFORCE.co.uk Organization: Dataforce Group Ltd To: sunmanagers@sunmanagers.org Hi all I am looking for a ball park figure for throughput on different servers/setup, predominantly SAN enviroments/V880 etc. Could you just run a timex mkfile 1g test on your SAN/V880/etc and tell me the result, that or equivalent and give me some idea of your setup. My setup is: SunFire 3800 (2x900mhz CPU, 4GB MEM) - 2400 Silkworm Brocade Switch - LSI e2400 Metastor Disk Array (10x73GB fibre based Seagate 15k rpm, HW RAID 5). TEST: mkfile 1g test SunFire 3800 ROOT DISKS (mirrored via Veritas Volume Manager) - timex mkfile 1g test r/s w/s kr/s kw/s wait actv wsvc_t asvc_t %w %b device 0.6 108.1 4.8 12611.1 0.0 74.8 0.0 688.0 0 100 c0t0d0 0.4 108.5 3.2 12683.3 0.0 76.0 0.0 697.6 0 100 c0t1d0 real 1:23.19 user 0.20 sys 9.98 SunFire 3800 SAN: timex mkfile 1g test r/s w/s kr/s kw/s wait actv wsvc_t asvc_t %w %b device 0.0 71.0 0.0 4437.5 0.0 56.7 0.1 798.6 0 90 c1t1d4 0.3 69.4 0.6 4344.4 0.0 66.0 0.1 947.0 0 95 c3t0d4 real 4:18.66 user 0.16 sys 8.97 Thanks in advance _______________________________________________ sunmanagers mailing list sunmanagers@sunmanagers.org http://www.sunmanagers.org/mailman/listinfo/sunmanagersReceived on Tue Mar 2 10:42:00 2004
This archive was generated by hypermail 2.1.8 : Thu Mar 03 2016 - 06:43:29 EST