Tek-Tips is the largest IT community on the Internet today!
Members share and learn making Tek-Tips Forums the best source of peer-reviewed technical information on the Internet!
-
Congratulations gkittelson on being selected by the Tek-Tips community for having the most helpful posts in the forums last week. Way to Go!
large file system vs. small file systems
- Thread starter tsuba
- Start date
Performance is going to be much more dependent on how you configure the disks rather than the size of the file system. For example, putting all the disks together using striping will be much faster than each file system on a seperate disk.
Provided your block sizes, etc. are the same, memory for file cache should not be affected - this would be more dependent on the individual files.
Provided your block sizes, etc. are the same, memory for file cache should not be affected - this would be more dependent on the individual files.
- Thread starter
- #3
I think with the default config AIX uses any free memory for file cache. This is why you should see the fre column (vmstat) a bit over 120 on a loaded system even when it should be "idle". Not always the case, of course.
I don't have experience with lots of little files, only large ones (datafiles). IBM Certified -- AIX 4.3 Obfuscation
I don't have experience with lots of little files, only large ones (datafiles). IBM Certified -- AIX 4.3 Obfuscation
True striping is most beneficial when you're doing large sequential reads or writes.
I'm guessing he's got a 36G disk and wants to figure out how to use it best.
tsuba: if you've got just one 36G disk, it doesn't really matter terribly much how you slice it up - you only have one path for data to move through.
If you have multiple disks, you want to spread your data across them as much as you can to distribute the slowest part of the system - the actual reading and writing. AIX's LVM is sufficient for doing this, just create your LVs to use the maximum number of PVs.
Now, from another point of view, I would still create multiple smaller filesystems instead of a single large filesystem. It's probably a personal preference, but I find smaller filesystems easier to manage than large ones, so I cap mine at 8G.
I'm guessing he's got a 36G disk and wants to figure out how to use it best.
tsuba: if you've got just one 36G disk, it doesn't really matter terribly much how you slice it up - you only have one path for data to move through.
If you have multiple disks, you want to spread your data across them as much as you can to distribute the slowest part of the system - the actual reading and writing. AIX's LVM is sufficient for doing this, just create your LVs to use the maximum number of PVs.
Now, from another point of view, I would still create multiple smaller filesystems instead of a single large filesystem. It's probably a personal preference, but I find smaller filesystems easier to manage than large ones, so I cap mine at 8G.
Smaller fs keep runaway things from filling up all the space, that's for sure.
Keeping a standard size is also good for system copies. I had to copy an Oracle db yesterday where two data fs were larger than our standard so I had to do a bit more manual labor to make it all fit, and of course if there was a tuned disk layout it's blown away on the new box. IBM Certified -- AIX 4.3 Obfuscation
Keeping a standard size is also good for system copies. I had to copy an Oracle db yesterday where two data fs were larger than our standard so I had to do a bit more manual labor to make it all fit, and of course if there was a tuned disk layout it's blown away on the new box. IBM Certified -- AIX 4.3 Obfuscation
- Status