[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: xfs vs jfs performance

I am not talking about huge delays but rather occasional 2-5 second delays.

I am using Courier IMAP with the Ext3 file-system and kernel quotas.

Postfix is delivering the Maildir file to the users' space.

The way Courier IMAP works is each mail becomes a separate '.imap' file.

Depending on the file's state, it goes into a different directory.

For example, when a new mail comes in it goes to, Maildir/.new 

When it has been viewed it moves from Maildir/.new to Maildir/.cur.

If I put files in my personal directory they end up in

Since I have about 200 - 250 people logged in during peak periods on a
dual 700Mhz machine that is mostly idle 95% of the time (except for
the off peak hour backups and quota indexing), it appears that 
the file-system must be the bottleneck.

I calculate that Courier IMAP is moving about 200-500 files every minute
during the delays.

Additionally, mail is coming in at the rate of 100-300 messages per minute.

Since ext3 is built on top of ext2, it adds a lot of overhead.

The kernel quotas add more overhead.

Although it is easy to move from ext2 to ext3, it does not offer
any greater read or write performance. 

In this month's Linux Journal, for example, 
there is an article about the new SGI 64 bit machine. One thing that they used for 
metrics was the file-system. According the article both ext2 and xfs
performed about the same on the 'super server'. Reiser and ext3 both performed
about 1/4 that of ext2.  

Since the system is not being taxed in any other noticeable way
according to sar, I feel that the file-system must be the bottleneck.
More specifically, it has to be ext3 or the quotas with ext3.

> On Wed, Feb 12, 2003 at 08:16:47AM +1100, Jean-Francois Dive wrote:
> Hi,

> not that i ever tested any of those 2 new file-system, but i have some
> troubles to believe that the FS'd be the bottleneck in your scenario;
> maybe i'm wrong, and 'd be interested to read some tests too though.
> JeF

Reply to: