Content added Content deleted
imported>Phrack No edit summary |
imported>mutante m (minor bracket error) |
||
(One intermediate revision by the same user not shown) | |||
Line 1: | Line 1: | ||
'''Comparison of XFS, JFS, ext3 with different Chunk Sizes and the Impact on Linux Software RAID Performance in RAID0/1/1+0/5''' |
'''Comparison of [[XFS]], [[JFS]], [[ext3]] with different Chunk Sizes and the Impact on [[Linux]] Software [[RAID]] Performance in RAID0/1/1+0/5''' |
||
The attached PDF contains a detailed summary comparing the file systems '''JFS''', '''XFS''' and '''ext3''' on linux software raid arrays with different raid-'''chunk-sizes''' in different '''raid-modes (0,1,1+0,5, single disk)''' |
The [[Media:Raid-FS_Mode_and_ChunkSize_Benchmark.pdf|attached PDF]] contains a detailed summary comparing the file systems '''JFS''', '''XFS''' and '''ext3''' on linux software raid arrays with different raid-'''chunk-sizes''' in different '''raid-modes (0,1,1+0,5, single disk)''' |
||
*I compiled this list on |
*I compiled this list on a testing-host running [[Debian GNU/Linux]] Lenny (testing), Kernel 2.6.23, 2x 1024MB DDR2-667 (Dual Channel, 2 Modules), Intel Core2Quad 2400MHz |
||
**at first i tried benchmarking with 4x2 = 8GB of RAM, the results were the same! so i "downgraded" the system to make benchmarks run in less time (bonnie++ needs twice the amount of physical RAM as test-file-size // the testing host has plenty performance and the software raid is in no way affected by RAM, CPU, etc. It pretty much shows the performance on the disks. |
**at first i tried benchmarking with 4x2 = 8GB of RAM, the results were the same! so i "downgraded" the system to make benchmarks run in less time ([[bonnie++]] needs twice the amount of physical [[RAM]] as test-file-size // the testing host has plenty performance and the software raid is in no way affected by RAM, [[CPU]], etc. It pretty much shows the performance on the disks. |
||
*All tests were performed on additional disks in the system. There was neither interference with the hosts boot disk nor was there any load |
*All tests were performed on additional disks in the system. There was neither interference with the hosts boot disk nor was there any load |
||
*benchmarking software: bonnie++ v1.3b with 4GB testing filesize (twice the size of installed RAM) |
*benchmarking software: bonnie++ v1.3b with 4GB testing filesize (twice the size of installed RAM) |
||
Line 15: | Line 15: | ||
If you would like to run tests on your own system, simply |
If you would like to run tests on your own system, simply |
||
*install '''bonnie++''' |
*install '''bonnie++''' |
||
< |
<tt>apt-get update && apt-get install bonnie++</tt> |
||
*mount the file system you want to test |
*mount the file system you want to test |
||
*start bonnie++ |
*start bonnie++ |
||
< |
<tt>bonnie++ -r 2048 -u 0 -g 0 -s 4g</tt> |
||
(-r is the amount of physical RAM in MB, -u and -g are user and group-IDs which have to be "0", -s is the size of the test file in MB, which has to be at leat twice the amount of RAM in your machine) |
(-r is the amount of physical RAM in MB, -u and -g are user and group-IDs which have to be "0", -s is the size of the test file in MB, which has to be at leat twice the amount of RAM in your machine) |
||
Latest revision as of 20:04, 26 February 2008
Comparison of XFS, JFS, ext3 with different Chunk Sizes and the Impact on Linux Software RAID Performance in RAID0/1/1+0/5
The attached PDF contains a detailed summary comparing the file systems JFS, XFS and ext3 on linux software raid arrays with different raid-chunk-sizes in different raid-modes (0,1,1+0,5, single disk)
- I compiled this list on a testing-host running Debian GNU/Linux Lenny (testing), Kernel 2.6.23, 2x 1024MB DDR2-667 (Dual Channel, 2 Modules), Intel Core2Quad 2400MHz
- at first i tried benchmarking with 4x2 = 8GB of RAM, the results were the same! so i "downgraded" the system to make benchmarks run in less time (bonnie++ needs twice the amount of physical RAM as test-file-size // the testing host has plenty performance and the software raid is in no way affected by RAM, CPU, etc. It pretty much shows the performance on the disks.
- All tests were performed on additional disks in the system. There was neither interference with the hosts boot disk nor was there any load
- benchmarking software: bonnie++ v1.3b with 4GB testing filesize (twice the size of installed RAM)
Media:Raid-FS_Mode_and_ChunkSize_Benchmark.pdf
If you would like to run tests on your own system, simply
- install bonnie++
apt-get update && apt-get install bonnie++
- mount the file system you want to test
- start bonnie++
bonnie++ -r 2048 -u 0 -g 0 -s 4g (-r is the amount of physical RAM in MB, -u and -g are user and group-IDs which have to be "0", -s is the size of the test file in MB, which has to be at leat twice the amount of RAM in your machine)