Matthew Miller wrote: > On Wed, Sep 08, 2004 at 12:15:43PM +0100, Douglas Furlong wrote: >> If you create the RAID 0 in striped mode then the read/write speed for >> large files is substantially increased. I believe, in the region of 70% >> -> 80%. However I can't remember exactly off the top of my head. > > Even then, that's a best-case test -- not reflective of something most > users actually do often. ok, just for fun, i grabbed two old 40 GB hard drives. i thought they were exactly the same, but one is obviously faster than the other (see below). i threw them into a RAID-0 configuration and installed FC2 on them. here are the partitions. the geometries are the same and the partitions are exactly the same between drives. [root@mutex root]# fdisk -l Disk /dev/hda: 40.0 GB, 40020664320 bytes 16 heads, 63 sectors/track, 77545 cylinders Units = cylinders of 1008 * 512 = 516096 bytes Device Boot Start End Blocks Id System /dev/hda1 * 1 260 131008+ 83 Linux /dev/hda2 261 1300 524160 82 Linux swap /dev/hda3 1301 77545 38427480 fd Linux raid autodetect Disk /dev/hdc: 40.0 GB, 40020664320 bytes 16 heads, 63 sectors/track, 77545 cylinders Units = cylinders of 1008 * 512 = 516096 bytes Device Boot Start End Blocks Id System /dev/hdc1 * 1 260 131008+ 83 Linux /dev/hdc2 261 1300 524160 82 Linux swap /dev/hdc3 1301 77545 38427480 fd Linux raid autodetect here are the hdparm test results [root@mutex root]# hdparm -Tt /dev/hda /dev/hdc /dev/md0 /dev/hda: Timing buffer-cache reads: 732 MB in 2.00 seconds = 366.06 MB/sec Timing buffered disk reads: 72 MB in 3.01 seconds = 23.93 MB/sec /dev/hdc: Timing buffer-cache reads: 732 MB in 2.00 seconds = 365.33 MB/sec Timing buffered disk reads: 108 MB in 3.05 seconds = 35.42 MB/sec /dev/md0: Timing buffer-cache reads: 724 MB in 2.00 seconds = 361.87 MB/sec Timing buffered disk reads: 132 MB in 3.00 seconds = 43.99 MB/sec i also downloaded bonnie++ and tiobench, but then realized i made a mistake. the non RAID partitions i made were too small to be used for accurate tests. now i realize these test kind of suck because one hard drive is obviously faster than the other, but either way, the RAID device is 84% faster than /dev/hda and 24% faster than /dev/hdb. what about writes? well the results of the "unreliable" bonnie++ test showed mediocre read/write speed gains, but file creation/deletion was over twice as fast on the RAID. about the need for RAID-0... i never thought too much about it until two things happened: 1) doom3 was released 2) i started re-authoring dvds doom3 took literally minutes to load a single level...yuk. when re-authoring a dvd, i found that writing the new dvd to hard drive took over an hour (that actually might be the program's fault since reading it off the hard drive doesn't take nearly as long). either way, i needed a faster and bigger (dvd's quickly fill up hard drives these days!) hard drive...RAID-0 fullfills both those requirements...=) about redundancy...i really have no need for it. besides, modern day hard drives have a mean time to failure of what? 5 years? so two of them in a RAID-0 would have a MTTF of 2.5...i'll probably have a whole new computer by then (with new hard drives). also, until i have a hard drive fail on me, i'll foolheartedly toss redundancy into the wind...;)
