> -----Original Message----- > From: [EMAIL PROTECTED] > [mailto:[EMAIL PROTECTED] On Behalf Of rbl000 > Sent: Friday, January 19, 2007 1:14 PM > To: [email protected] > Subject: [U2] Single Part vs. Multi part UniVerse Files
[snip] > Needless to say while the single threaded access conformed to > our original expectations, it did little to enlighten us > other than to point towards a link between multi-threading > and multi-parting. Just speculating here, but it could be that multi-threading is resulting in the CPU cache being trashed once the file is combined. This has been observed when running SQL server on a Xeon box with hyperthreading enabled. When 2 threads that are accessing different data get assigned to the same CPU, the CPU cache has to be discarded with every clock cycle causing a severe performance degredation. Maybe your batch processes against the multipart version of the file result in the kernel assigning threads that appear to be accessing the same area of storage to the same CPU. This would allow the cache to frequently be relevant to multiple threads. Once the file is combined, however, the hashing algorithm would make the distribution on disk unpredictable for every thread, resulting in frequently discarded caches. -John ------- u2-users mailing list [email protected] To unsubscribe please visit http://listserver.u2ug.org/
