xfs
[Top] [All Lists]

Re: 2.6.11-rc3: 80-95% systime when serving files via NFS

To: Anders Saaby <as@xxxxxxxxxxxx>
Subject: Re: 2.6.11-rc3: 80-95% systime when serving files via NFS
From: Steve Lord <lord@xxxxxxx>
Date: Tue, 08 Feb 2005 08:25:52 -0600
Cc: Dave Chinner <dgc@xxxxxxx>, linux-xfs@xxxxxxxxxxx
In-reply-to: <200502081310.30303.as@xxxxxxxxxxxx>
References: <222BE5975A4813449559163F8F8CF5030C22D2@xxxxxxxxxxxxxxxxxxxx> <20050208153427.G147390@xxxxxxxxxxxxxxxxx> <200502081152.55315.as@xxxxxxxxxxxx> <200502081310.30303.as@xxxxxxxxxxxx>
Sender: linux-xfs-bounce@xxxxxxxxxxx
User-agent: Mozilla Thunderbird 0.9 (X11/20041127)
Anders Saaby wrote:


Extra note:

As one of my colleagues pointed out, prime numbers are usually better for hashtable sizes, so I just tried ihashsize=64433 (prime number).

This resulted in ~20% less systime and the following kernel profiles looks like prime numbers definetly _is_ a good idea:


To me that usually suggests that a better hash algorithm is called for,
using the prime number for a bucket count means you get into doing
more complex math to work out the bucket.

Steve (who has not looked at this code at all)


<Prev in Thread] Current Thread [Next in Thread>