On Thu, Sep 27, 2001 at 04:54:24PM -0500, Bob Tanner wrote: > Quoting Jon Erickson (jon.erickson at neicoltech.org): > > I'm building a box that will act as a firewall \ transparent proxy (squid) > > for roughly 500 clients. I plan to use Reiserfs on the cache partitions > > and would like to use a 2.4.X kernel with it's stateful firewall. > > Can you explain why you'd run Reiserfs on the cache partition? > > The testing I have done, shows ext2 faster the Reiserfs, and on a cache fs, I > don't think journaling is necessary or needed (IMHO). > > Just curious. Reiserfs is definetly faster when a large number of small files is involved like it would be for a squid cache. What benchmark did you use? Postmark (http://www.netapp.com/tech_library/3022.html) is a good benchmark utility for testing a filesystem with a large number of changing file. It's faster because the B* tree can find the specific file in the directory structure much quicker than on an ext2 filesystem. Once the file is found, the throughput is similar (sometimes less) in performance when compared to ext2. It becomes an issue of disk speed at that point. There is also a benchmark section off www.reiserfs.org which has a number of different benchmark comparisons. Regards - Karl