I'm trying to optimize one of our news servers, and was hoping for some
suggestions. I could especially use some advice on how to figure out the
best chunk size and block/inode size.
Specs:
It is a Pentium 166 MMX w/ 128MB running Linux 2.2.6,
raid0145-19990421-2.2.6 patch, and raidtools-19990421-0.90. We are running
software RAID0 with 6 4.3 UW SCSI disks on an Adaptec 2940 controller with
everything set at 40MB/s. The array is used solely for the news spool. The
root disk is on a separate controller.
/etc/raidtab:
raiddev /dev/md0
raid-level 0
nr-raid-disks 6
persistent-superblock 1
chunk-size 8
device /dev/sdc1
raid-disk 0
device /dev/sdd1
raid-disk 1
device /dev/sde1
raid-disk 2
device /dev/sdf1
raid-disk 3
device /dev/sdg1
raid-disk 4
device /dev/sdh1
raid-disk 5
The current filesystem was created using:
# mke2fs -b 1024 -i 1024 -m 0 /dev/md0
The articles in the spool appear to be laid out in the following sizes:
1-2K 48.4%
2-4K 20.1%
0-1K 16.1%
>4K 15.6%
Questions:
1) The block size/bytes per inode was set at 1024 using the reasoning that
there would be many small files (which there are). However, looking at the
above article sizes, perhaps 2K blocks would make more sense. (Maybe even
4K, if the performance is enough.) Comments?
2) Should I even bother using the bytes per inode (-i) option to mke2fs?
The mke2fs man page says this defaults to 4096k, which for a 1024k block
size would mean 1 inode for every 4 blocks. Since most of my files are
1k-2k, setting the bytes/inode lower is a good idea so as not to run out
of inodes, correct?
3) Any tips on how to find the best chunk size for this configuration?
Since almost all of the files are less than 4K, I'm assuming that I should
make my current chunk-size (8k was a starting point) smaller. But should I
go to a 2K or even 1K chunk size?
Thanks for any help!
Dave
=== (Crappy Sig #13)
Dave | E-Mail | The truth is out
[EMAIL PROTECTED] | [EMAIL PROTECTED] | there. Drink
BOFH | for PGP Public Key | milk. MOO.
18 92 8E 39 D5 CD 26 60 3B B1 A2 59 4C 3A 76 5F
