FreeBSD 5.3 I/O Performance / Linux 2.6.10 | Continued Discussion

Robert Watson rwatson at FreeBSD.org
Fri Jan 28 16:50:25 PST 2005


On Fri, 28 Jan 2005, Mike Tancsa wrote:

> At 04:52 AM 28/01/2005, Robert Watson wrote:
> 
> >Just to get started, using dd to read and write at various block sizes is
> >probably a decent start.  Take a few samples, make sure there's a decent
> >sample size, etc, and don't count the first couple of runs.
> 
> 
> I started over and installed the OS on a separate drive, so that the
> RAID5 partition was one big partition. Bonnie seems fairly equal on the
> 2 platforms. I am just putting together all the IOZONE tests as well
> which I will post later tonight / over the weekend. 

The basic I/O results are encouraging.  Could you remind me which storage
controller you're using?  I'm guessing some of what you're seeing is Giant
content and a lack of preemptibility to let the disk controller interrupt
the file system code -- this would add a lot of latency to turning around
I/O's, lowering the throughput.  Perhaps combined with issues in soft
updates and/or UFS generally.  I've CC'd Jeff Roberson since he's working
in several of these areas. 

For postmark, could you try running the test against a partition with the
async flag set, but no soft updates turned on?  Also, could you compare
with the sync flag on both Linux and FreeBSD (again, no soft updates)? 

Robert N M Watson


> 
> 
> 
>                -------Sequential Output-------- ---Sequential Input-- 
> --Random--
>                -Per Char- --Block--- -Rewrite-- -Per Char- --Block--- 
> --Seeks---
> Machine    MB K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU  /sec %CPU
> F        2000 33789 25.8 34065  8.3 14038  3.8 47484 44.6  67316  9.5 
> 2873.7  5.7
> L        2000 28073 62.0 22391  4.3 12189  2.0 17577 35.5 102132  5.5 
> 1853.7  2.5
> 
> F=RELENG_5
> L=2.6.10 LINUX
> 
> However, on the postmark tests, there seems to be quite a bit more difference.
> 
> RELENG_5 #2
> [nfs]# time dd if=/dev/zero of=/mnt/test bs=16k count=2000
> 32768000 bytes transferred in 0.724865 secs (45205661 bytes/sec)
> 0.000u 0.066s 0:00.73 8.2%      33+281k 0+250io 0pf+0w
> [nfs]# time dd if=/dev/zero of=/mnt/test bs=32k count=2000
> 65536000 bytes transferred in 1.731888 secs (37840788 bytes/sec)
> 0.000u 0.122s 0:01.74 6.8%      28+283k 0+500io 0pf+0w
> [nfs]# time dd if=/dev/zero of=/mnt/test bs=64k count=2000
> 131072000 bytes transferred in 3.131403 secs (41857273 bytes/sec)
> 0.000u 0.245s 0:03.14 7.6%      25+335k 0+1000io 0pf+0w
> [nfs]# time dd if=/dev/zero of=/mnt/test bs=128k count=2000
> 262144000 bytes transferred in 6.973014 secs (37594073 bytes/sec)
> 0.000u 0.509s 0:07.00 7.1%      26+510k 0+2000io 0pf+0w
> [nfs]# time dd if=/dev/zero of=/mnt/test bs=256k count=2000
> 524288000 bytes transferred in 14.253915 secs (36782035 bytes/sec)
> 0.000u 1.003s 0:14.30 6.9%      25+784k 0+4000io 0pf+0w
> [nfs]# time dd if=/dev/zero of=/mnt/test bs=512k count=2000
> 1048576000 bytes transferred in 28.135716 secs (37268502 bytes/sec)
> 0.008u 2.165s 0:28.21 7.6%      24+1386k 0+8000io 0pf+0w
> [nfs]#
> 
> 
> LINUX 2.6.10
> 
> pm>set size 300 100000
> pm>set location /mnt
> pm>set transactions 400000
> pm>run
> Creating files...Done
> Performing transactions..........Done
> Deleting files...Done
> Time:
>          1504 seconds total
>          1503 seconds of transactions (266 per second)
> 
> Files:
>          200107 created (133 per second)
>                  Creation alone: 500 files (500 per second)
>                  Mixed with transactions: 199607 files (132 per second)
>          199905 read (133 per second)
>          199384 appended (132 per second)
>          200107 deleted (133 per second)
>                  Deletion alone: 889 files (889 per second)
>                  Mixed with transactions: 199218 files (132 per second)
> 
> Data:
>          12715.55 megabytes read (8.45 megabytes per second)
>          12728.92 megabytes written (8.46 megabytes per second)
> pm>
> 
> 
> RELENG_5.
> PostMark v1.5 : 3/27/01
> pm>set location /mnt
> pm>size 300 100000
> Eh?
> pm>set size 300 100000
> pm>set location /mnt
> pm>set transactions 400000
> pm>run
> Creating files...Done
> Performing transactions..........Done
> Deleting files...Done
> Time:
>          2490 seconds total
>          2487 seconds of transactions (160 per second)
> 
> Files:
>          200107 created (80 per second)
>                  Creation alone: 500 files (500 per second)
>                  Mixed with transactions: 199607 files (80 per second)
>          199905 read (80 per second)
>          199384 appended (80 per second)
>          200107 deleted (80 per second)
>                  Deletion alone: 889 files (444 per second)
>                  Mixed with transactions: 199218 files (80 per second)
> 
> Data:
>          12715.55 megabytes read (5.11 megabytes per second)
>          12728.92 megabytes written (5.11 megabytes per second)
> pm>
> 
> 
> 
> RELENG_4 (Note, this was on a different partition on the RAID5 card
> 
> pm>set size 300 100000
> pm>set location /card0-a
> pm>set transactions 400000
> pm>run
> Creating files...Done
> Performing transactions..........Done
> Deleting files...Done
> Time:
>          1219 seconds total
>          1219 seconds of transactions (328 per second)
> 
> Files:
>          200107 created (164 per second)
>                  Creation alone: 500 files (500 per second)
>                  Mixed with transactions: 199607 files (163 per second)
>          199905 read (163 per second)
>          199384 appended (163 per second)
>          200107 deleted (164 per second)
>                  Deletion alone: 889 files (889 per second)
>                  Mixed with transactions: 199218 files (163 per second)
> 
> Data:
>          12715.55 megabytes read (10.43 megabytes per second)
>          12728.92 megabytes written (10.44 megabytes per second)
> pm>
> 
> 
> _______________________________________________
> freebsd-performance at freebsd.org mailing list
> http://lists.freebsd.org/mailman/listinfo/freebsd-performance
> To unsubscribe, send any mail to "freebsd-performance-unsubscribe at freebsd.org"
> 



More information about the freebsd-performance mailing list