Hi folks, Derrick has committed some patches to make AFS happy about having the cache in advfs. This is accomplished by using the advfs built-in lookup-by-inumber mechanism instead of groveling around in internal structures. This might mean that the performance is less than optimal, a simple test seems to indicate that it's somewhat slower than a UFS-cache. First start of afsd with many cache-files is much faster though (no surprise). The alternative is making Compaq release the relevant interface structures for advfs, which I don't want to mess around with. I did try doing it the `right' way by using bs_get_current_tag + bf_get, but the same simple test (`find -exec wc' on a large file set) didn't show any performance improvement. Anyway it would be great if more people could try this out before 1.0.4 or 1.1 or whatever the next release will be called. Especially interesting is performance figures. If it isn't obvious: this is *only* for the client cache, vice partitions still has to be UFS. /Johan _______________________________________________ OpenAFS-devel mailing list [EMAIL PROTECTED] https://lists.openafs.org/mailman/listinfo.cgi/openafs-devel
