« Return to Thread: Performance Tests
RE: Performance Tests
by Walker, Bruce J (HP-Labs)
::
Rate this Message:
Performance numbers would be most useful if done as a comparison (say base vs. single node CFS or CFS server node vs. non-server node; chard vs. non chard, etc.). We could then identify which areas need work (CFS stacked on ext3 or remote CFS; large reads vs. small writes, etc.).
Bruce
-----Original Message-----
From: ssic-linux-users-admin@... [mailto:ssic-linux-users-admin@...] On Behalf Of Roger Tsang
Sent: Monday, December 19, 2005 4:45 PM
To: Stephan Böni
Cc: ssic-linux-users@...
Subject: Re: [SSI-users] Performance Tests
On 12/19/05, Stephan Böni <boeni@...> wrote:
>
> Perhaps nice to know: We've made a lot performance tests (disk i/o
> related) with an OpenSSI 1.9.2 (SUSE) cluster containing 4 nodes IBM
> x336 with each 1 Xen 3 GHz CPU / 4 GB RAM and one IBM DS4300 storage
> system attached over fibre channel with 8 RAID5 FC-Disks. Here are the results:
>
> - One system as single system has a good r/w performance.
> - One system as single node OpenSSI cluster (not usable) has a
> terrible r/w performance.
> - Two systems (nodes) in a OpenSSI cluster have a good read but still
> a terrible write performace.
> - 3 and 4 nodes are not better than 2 nodes on r/w performance.
>
> On heavy load OpenSSI crashs sometimes. The disk write performance is
> so terrible, that my old PC with an ide disk is faster. :-( Perhaps
> that could be faster with GFS.
>
> Stephan
>
This is what I'm getting on FC3 SSI-1.9.2 preview AMD64 3000+ nodes with 2GB RAM each with IDE disks (not raided). Running the tests at the CFS server on DRBD storage (procotol C) with Gigabit ICS (MTU
1500) and the filesystem is ext3 (rw,noatime,chard) with failover enabled. Default /proc/sys/vm knobs.
The following is just to give an idea, since there is a certain level of buffering going on here.
Iozone: Performance Test of File I/O
Version $Revision: 3.226 $
Compiled for 32 bit mode.
Build: linux
Contributors:William Norcott, Don Capps, Isom Crawford, Kirby Collins
Al Slater, Scott Rhine, Mike Wisner, Ken Goss
Steve Landherr, Brad Smith, Mark Kelly, Dr. Alain CYR,
Randy Dunlap, Mark Montague, Dan Million,
Jean-Marc Zucconi, Jeff Blomberg,
Erik Habbinga, Kris Strecker.
Run began: Mon Dec 19 19:28:52 2005
Auto Mode
File size set to 81920 KB
Command line used: /opt/iozone/bin/iozone -a -s 80m
Output is in Kbytes/sec
Time Resolution = 0.000001 seconds.
Processor cache size set to 1024 Kbytes.
Processor cache line size set to 32 bytes.
File stride size set to 17 * record size.
random
random bkwd record stride
KB reclen write rewrite read reread read
write read rewrite read fwrite frewrite fread freread
81920 4 592283 619564 988083 980068 878961
20679 941683 1586735 898274 548708 592566 1013722 1011595
81920 8 667475 696829 1150625 1187178 1076745
29790 960869 1083355 1059753 620187 658304 1138362 1167036
81920 16 698469 719081 1375325 1385189 1290629
35736 1309778 1360550 603942 115238 449226 1201207 537780
81920 32 419373 695628 1425044 554109 1414509
16313 1423036 2156929 1372745 678662 495718 1323854 1420667
81920 64 722640 434416 1421208 1353176 1290444
17924 1323576 1510463 1341103 221657 281032 1290404 998561
81920 128 684029 781753 1293133 1301826 1327970
37276 1300853 1603821 1312924 628143 654124 1319672 1317889
81920 256 681287 714472 1061937 1091101 1113102
36389 981677 1061138 1081709 326928 482651 1125198 1138187
81920 512 363705 365272 720733 738417 735018
37955 721412 632064 713837 417016 396106 709902 562062
81920 1024 365789 358002 739383 728818 732717
43929 732887 581319 726994 424326 444714 735659 412643
81920 2048 571919 590733 425885 712323 719789
44347 697946 670611 718332 294249 447814 674666 738364
81920 4096 559207 591715 730456 730619 739503
109754 736915 673545 727609 428350 325316 669850 694349
81920 8192 559704 589009 742150 729338 723188
102824 742553 663567 714036 445961 474620 744605 730358
81920 16384 348019 595041 735355 404941 739103
43135 418527 664757 653248 425461 471821 649071 361038
iozone test complete.
Roger
-------------------------------------------------------
This SF.net email is sponsored by: Splunk Inc. Do you grep through log files for problems? Stop! Download the new AJAX search engine that makes searching your log files as easy as surfing the web. DOWNLOAD SPLUNK!
http://ads.osdn.com/?ad_idv37&alloc_id865&op=ick
_______________________________________________
Ssic-linux-users mailing list
Ssic-linux-users@...
https://lists.sourceforge.net/lists/listinfo/ssic-linux-users
-------------------------------------------------------
This SF.net email is sponsored by: Splunk Inc. Do you grep through log files
for problems? Stop! Download the new AJAX search engine that makes
searching your log files as easy as surfing the web. DOWNLOAD SPLUNK!
http://ads.osdn.com/?ad_idv37&alloc_id865&op=click
_______________________________________________
Ssic-linux-users mailing list
Ssic-linux-users@...
https://lists.sourceforge.net/lists/listinfo/ssic-linux-users
Bruce
-----Original Message-----
From: ssic-linux-users-admin@... [mailto:ssic-linux-users-admin@...] On Behalf Of Roger Tsang
Sent: Monday, December 19, 2005 4:45 PM
To: Stephan Böni
Cc: ssic-linux-users@...
Subject: Re: [SSI-users] Performance Tests
On 12/19/05, Stephan Böni <boeni@...> wrote:
>
> Perhaps nice to know: We've made a lot performance tests (disk i/o
> related) with an OpenSSI 1.9.2 (SUSE) cluster containing 4 nodes IBM
> x336 with each 1 Xen 3 GHz CPU / 4 GB RAM and one IBM DS4300 storage
> system attached over fibre channel with 8 RAID5 FC-Disks. Here are the results:
>
> - One system as single system has a good r/w performance.
> - One system as single node OpenSSI cluster (not usable) has a
> terrible r/w performance.
> - Two systems (nodes) in a OpenSSI cluster have a good read but still
> a terrible write performace.
> - 3 and 4 nodes are not better than 2 nodes on r/w performance.
>
> On heavy load OpenSSI crashs sometimes. The disk write performance is
> so terrible, that my old PC with an ide disk is faster. :-( Perhaps
> that could be faster with GFS.
>
> Stephan
>
This is what I'm getting on FC3 SSI-1.9.2 preview AMD64 3000+ nodes with 2GB RAM each with IDE disks (not raided). Running the tests at the CFS server on DRBD storage (procotol C) with Gigabit ICS (MTU
1500) and the filesystem is ext3 (rw,noatime,chard) with failover enabled. Default /proc/sys/vm knobs.
The following is just to give an idea, since there is a certain level of buffering going on here.
Iozone: Performance Test of File I/O
Version $Revision: 3.226 $
Compiled for 32 bit mode.
Build: linux
Contributors:William Norcott, Don Capps, Isom Crawford, Kirby Collins
Al Slater, Scott Rhine, Mike Wisner, Ken Goss
Steve Landherr, Brad Smith, Mark Kelly, Dr. Alain CYR,
Randy Dunlap, Mark Montague, Dan Million,
Jean-Marc Zucconi, Jeff Blomberg,
Erik Habbinga, Kris Strecker.
Run began: Mon Dec 19 19:28:52 2005
Auto Mode
File size set to 81920 KB
Command line used: /opt/iozone/bin/iozone -a -s 80m
Output is in Kbytes/sec
Time Resolution = 0.000001 seconds.
Processor cache size set to 1024 Kbytes.
Processor cache line size set to 32 bytes.
File stride size set to 17 * record size.
random
random bkwd record stride
KB reclen write rewrite read reread read
write read rewrite read fwrite frewrite fread freread
81920 4 592283 619564 988083 980068 878961
20679 941683 1586735 898274 548708 592566 1013722 1011595
81920 8 667475 696829 1150625 1187178 1076745
29790 960869 1083355 1059753 620187 658304 1138362 1167036
81920 16 698469 719081 1375325 1385189 1290629
35736 1309778 1360550 603942 115238 449226 1201207 537780
81920 32 419373 695628 1425044 554109 1414509
16313 1423036 2156929 1372745 678662 495718 1323854 1420667
81920 64 722640 434416 1421208 1353176 1290444
17924 1323576 1510463 1341103 221657 281032 1290404 998561
81920 128 684029 781753 1293133 1301826 1327970
37276 1300853 1603821 1312924 628143 654124 1319672 1317889
81920 256 681287 714472 1061937 1091101 1113102
36389 981677 1061138 1081709 326928 482651 1125198 1138187
81920 512 363705 365272 720733 738417 735018
37955 721412 632064 713837 417016 396106 709902 562062
81920 1024 365789 358002 739383 728818 732717
43929 732887 581319 726994 424326 444714 735659 412643
81920 2048 571919 590733 425885 712323 719789
44347 697946 670611 718332 294249 447814 674666 738364
81920 4096 559207 591715 730456 730619 739503
109754 736915 673545 727609 428350 325316 669850 694349
81920 8192 559704 589009 742150 729338 723188
102824 742553 663567 714036 445961 474620 744605 730358
81920 16384 348019 595041 735355 404941 739103
43135 418527 664757 653248 425461 471821 649071 361038
iozone test complete.
Roger
-------------------------------------------------------
This SF.net email is sponsored by: Splunk Inc. Do you grep through log files for problems? Stop! Download the new AJAX search engine that makes searching your log files as easy as surfing the web. DOWNLOAD SPLUNK!
http://ads.osdn.com/?ad_idv37&alloc_id865&op=ick
_______________________________________________
Ssic-linux-users mailing list
Ssic-linux-users@...
https://lists.sourceforge.net/lists/listinfo/ssic-linux-users
-------------------------------------------------------
This SF.net email is sponsored by: Splunk Inc. Do you grep through log files
for problems? Stop! Download the new AJAX search engine that makes
searching your log files as easy as surfing the web. DOWNLOAD SPLUNK!
http://ads.osdn.com/?ad_idv37&alloc_id865&op=click
_______________________________________________
Ssic-linux-users mailing list
Ssic-linux-users@...
https://lists.sourceforge.net/lists/listinfo/ssic-linux-users
« Return to Thread: Performance Tests
| Free embeddable forum powered by Nabble | Forum Help |
