There are lies, damn lies, and benchmarks. Note the sequential read
figure. Indexing probably reads the entire database sequentially.
HTH,
Jeffrey
Quoting genge1
I have run bonnie on three machines. If I read the numbers right the middle machine, new000, should be a much worse performer than sys000. This concerns me because I was planning to move sys000 to new000 to get better performance.
I ran a real world application on both. I primarily use filePro, so I reindexed all of the databases on both systems and it took 32 min on sys000 and 32 on new000, this would indicate I will get roughly the same IO performance, but better CPU power on new000.
Am I reading the bonnie numbers correctly? If bonnie says sys000 had better IO than new000 why roughly the same time on a heavy IO program?
Thanks in advance for any advise.
Greg Engel
Here are the stats: Command run with bonnie Version 1.2: bonnie -s 1000
Seek- -Per Char- --Block--- -Rewrite-- -Per Char- --Block--- --04k (03)- Machine MB K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU /sec %CPU sys000 1*1000 4678 98.5 6119 10.3 2881 12.2 4372 91.1 6750 14.2 104.0 4.7 new000 1*1000 1984 45.8 2585 9.0 1165 4.4 3064 61.6 8591 14.4 427.3 12.6 sys680 1*1000 2872 56.4 3049 9.1 2298 6.2 3878 65.4 11618 14.2 237.6 5.0
Machine SuSEVer filesystype CPUmzh/num Memorytotal swapsize sys000 6.4 ext2 300/ 2 516452 136544 new000 7.1 reiser 334/ 2 523848 11256 sys680 7.1 reiser 398/ 1 523856 15992
Notes: all three systems are configured as RAID 5 arrays using 9 gig drives all three are 10k rpm drives
sys000 uses an adaptec controller with an external differential SCSI array new000 uses a DPT SmartRAID IV controller with internal drives sys680 uses an IBM ServeRAID controller and internal drives
-- I don't do Windows and I don't come to work before nine. -- Johnny Paycheck