I'm using a PentiumII-MMX 333 with:
256MB, Adaptec AIC7880 Array 1000 scsi controller running
7x 8.7GB SEAGATE ST39140W MEDALIST Harddrives
Kernel 2.0.35 with raid0145-19981106-C-2.0.35 patch and Kernel 2.1.127 raid0145-19981109-C-2.1.127 patch
Raidtools 0.90 and Bonnie.

The first Harddrive of the array is used for standard Linux install the other 6 is configured as RAID5.
hence 8.7*6-8.7(for parity) = 43.5GB
raidtab is set to 32kb chunks and left-symmetric.

RAID5 performence test with Kernel 2.0.35

1 meg run     -------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
            1  4498 74.7 93772 91.6 30878 120.6  5824 91.0 151367 -0.0 6631.3 36.5
End of 1 meg run

10 meg run    -------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
           10  5385 84.7 85906 100.7 12214 41.7  3942 60.8 50468 44.4 9309.6 60.5
End of 10 meg run

100 meg run   -------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
          100  5030 79.6 13383 13.9  3479  8.7  3162 45.3  5211  5.1 209.6  3.1
End of 100 meg run

1000 meg run  -------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
         1000  4938 77.7 12060 12.9  3428  9.5  3004 43.4  4508  4.4  93.5  1.4
End of 1000 meg run

1808 meg run  -------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
         1808  4918 77.4 12022 12.6  3283  8.9  2880 41.4  4418  4.2  81.3  1.4
End of 1808 meg run
 

RAID5 performence test with Kernel 2.1.127

1 meg run     -------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
            1  4432 77.9 84418 82.4 16318 47.8  6341 99.1 19301 18.8 13620.7 81.7
End of 1 meg run

10 meg run    -------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
           10  5371 85.0 67909 79.6 27729 78.5  5282 81.5 167312 98.0 2888.4 17.3
End of 10 meg run

100 meg run   -------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
          100  5035 79.7 56675 61.4  7476 23.3  5164 82.0 117892 81.7 11255.5 81.6
End of 100 meg run

1000 meg run  -------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
         1000  4901 77.9 15468 24.4  6852 20.9  4851 71.3 14208 15.4 139.4  2.2
End of 1000 meg run

1808          -------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
         1808  4903 78.2 13992 22.8  6949 21.2  5058 74.4 14275 15.7 125.2  1.9
End of 1808 meg run
 

SO = Sequential Output
SI = Sequential Input

I don't know what happend to K 2.0.35 SI Block test, its way out of proportion.

based on the results K2.1.127 is slightly faster than K2.0.35 but uses a lot more cpu time... so there u have it...

Reply via email to