I've got a 15GB cache on our web server, and a 10GB one on our ftp mirror server, both work pretty well, except they take quite a while to scan at boot, even with the cache-subdir support. I haven't messed with changing the chunk size, but I'd presume that would cut the time down significantly, albeit at the expense of # of distinct files in the cache. (I really need to get back to one of my to-do-list items about reworking the dumpcache tools to build clean w/ openafs as well as possibly merging them into fs.)
There is definately something nice about holding your entire web site inside the cache. Definately appears to boost performance, especially when you have local web indexing that would repeatedly flush out the entire cache if it were small (300MB). I won't deploy any machine with less than a 250MB cache if I can help it. -- Nathan Derek Atkins wrote: > > I was never running with a cache that big. I was running > a 3GB cache and it worked ok. I never filled it nor did > I actually run any performance tests. > > Recent versions of OpenAFS should support it relatively well. > > -derek > > Nickolai Zeldovich <[EMAIL PROTECTED]> writes: > > > > ... is it useful to have a, say, 18Gb client cache? > > > > You might hit some performance problems in the dcache code with a > > cache that large. There's a number of places where the dcache code > > does linear scans of all dcache entries, and you'll potentially have > > 1.8 million dcache entries in the kernel. (You could probably cut > > that down to ~280k by having fewer dcache entries and assuming you'll > > cache mostly whole 64K chunks, but it's still quite a bit.) > > > > It'd be interesting to see what particular performance problems you > > would infact hit with such a large cache, and fix them, but that's > > more in the long run.. > > > > -- kolya > > _______________________________________________ > > OpenAFS-info mailing list > > [EMAIL PROTECTED] > > https://lists.openafs.org/mailman/listinfo/openafs-info > > -- > Derek Atkins, SB '93 MIT EE, SM '95 MIT Media Laboratory > Member, MIT Student Information Processing Board (SIPB) > URL: http://web.mit.edu/warlord/ PP-ASEL-IA N1NWH > [EMAIL PROTECTED] PGP key available > _______________________________________________ > OpenAFS-info mailing list > [EMAIL PROTECTED] > https://lists.openafs.org/mailman/listinfo/openafs-info -- ------------------------------------------------------------ Nathan Neulinger EMail: [EMAIL PROTECTED] University of Missouri - Rolla Phone: (573) 341-4841 Computing Services Fax: (573) 341-4216 _______________________________________________ OpenAFS-info mailing list [EMAIL PROTECTED] https://lists.openafs.org/mailman/listinfo/openafs-info
