hgj1357
asked on
Raid 0 Chunk size
I am using Red Hat Linux Enterprise 2.4
I have two 500Gb disks in a Raid 0 array. The files stored on this array will be large (at least 100Mb each).
What chunk size should I use?
Also, when I format my raid volume (as ext3) can I set block size, and is there a block size that will work better for me?
These files will be read-only.
Many thanks
I have two 500Gb disks in a Raid 0 array. The files stored on this array will be large (at least 100Mb each).
What chunk size should I use?
Also, when I format my raid volume (as ext3) can I set block size, and is there a block size that will work better for me?
These files will be read-only.
Many thanks
ASKER
after creating the raid volume is formating like this:
mkfs -t ext3 /dev/md0
how would I specify a block size?
Thanks
mkfs -t ext3 /dev/md0
how would I specify a block size?
Thanks
Hi,
To format the HDD you would issue:
mkfs.ext3 /dev/md0
To format the HDD you would issue:
mkfs.ext3 /dev/md0
ASKER
Actually, mkfs -t ext3 /dev/md0
but my question was how do I set the block size.
but my question was how do I set the block size.
ASKER
Actually (again) this is what I think I need
mke2fs -b 4096 -t largefile4 ext3
right?
mke2fs -b 4096 -t largefile4 ext3
right?
To be fair:
mkfs.ext3 /dev/md0
is equivalent to:
mkfs -t ext3 /dev/md0
Anyway, block size is set in /etc/raidtab. There is a directive:
chunk-size 16
Which would set the chunksize to 16k. Chunk size *must* be a power of 2. So 2,4,8,16,32,64,128 etc are all valid. Optimum chunk size is dependent of disc speed, disk-cache, bus speed, availability of DMA, and interface type! As a result it is extremely difficult to calculate the optimum size. However, if the system is running with a > 3GHz processor, and a very high speed disc (15000rpm SCSI) for example, then I might be tempted to go for a 64k chunksize for the large file reads that you are specifying. However, for lesser systems, I'd probably stick at something lower. For most practical applications, 8 or 16 is more than ample. Also bear in mind that there is no point in extracting the data from the hard drive(or buying the highest speed hard drive or creating the highest spec RAID array) if the bottlenet to getting the information is a 10/100 ethernet card!
HTH:)
mkfs.ext3 /dev/md0
is equivalent to:
mkfs -t ext3 /dev/md0
Anyway, block size is set in /etc/raidtab. There is a directive:
chunk-size 16
Which would set the chunksize to 16k. Chunk size *must* be a power of 2. So 2,4,8,16,32,64,128 etc are all valid. Optimum chunk size is dependent of disc speed, disk-cache, bus speed, availability of DMA, and interface type! As a result it is extremely difficult to calculate the optimum size. However, if the system is running with a > 3GHz processor, and a very high speed disc (15000rpm SCSI) for example, then I might be tempted to go for a 64k chunksize for the large file reads that you are specifying. However, for lesser systems, I'd probably stick at something lower. For most practical applications, 8 or 16 is more than ample. Also bear in mind that there is no point in extracting the data from the hard drive(or buying the highest speed hard drive or creating the highest spec RAID array) if the bottlenet to getting the information is a 10/100 ethernet card!
HTH:)
ASKER CERTIFIED SOLUTION
membership
This solution is only available to members.
To access this solution, you must be a member of Experts Exchange.
ASKER
Got it!
Have a look at:
http://howtos.linux.com/howtos/Software-RAID-HOWTO-9.shtml
The site gives you a table of what chunk size to use for RAID Level 0