I was doing a JProfiler install of our webapp/lucene last week and of
course a large part of our app is spent in RandomAccessFile.readBytes ...
This is called by InputStream.readByte which internally uses a
BUFFER_SIZE of 1024 (which is the default).
This value seems too small for a default. Most filesystem block sizes
are 4096 and a lot of controllers will refuse to read in block sizes
smaller than 16384.
On top of that you also have the the physical disks that have a minimum
block size and also RAID controller block sizes.
I tried to tune this variable this weekend with different settings and
had strange results. Mostly because (I'm assuming) that my Linux
filesystem cache and RAID controller cache were getting in the way (and
right now I have no way to fix this).
So my questions are:
1. What do people thing about changing this value? Does anyone else do it?
2. Has anyone run any benchmarks here? Does anyone have any solid
lucene benchmarks they can run with before/after runs with this var changed?
3. Does anyone else think that 1024 is too small of a default value for
modern systems?
4. Does anyone know what the default block size for other filesystems? I
know that XFS is 4096. What about ext2? ext3? JFS? ReiserFS? NTFS? UFS?
etc....
Kevin
--
Use Rojo (RSS/Atom aggregator)! - visit http://rojo.com.
See irc.freenode.net #rojo if you want to chat.
Rojo is Hiring! - http://www.rojonetworks.com/JobsAtRojo.html
Kevin A. Burton, Location - San Francisco, CA
AIM/YIM - sfburtonator, Web - http://peerfear.org/
GPG fingerprint: 5FB2 F3E2 760E 70A8 6174 D393 E84D 8D04 99F1 4412
---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]
