Re: Best database for web backend on Linux?

There are 2 main ways that systems use multiple CPUs to speed query performance. The 2 main methods are spearheaded by Sybase and Oracle. Sybase forms data affinity to a processor, while oracle defaults to using a queue, such that all processors can perform parts of a query in parallel, regardless of data affinity. This works well for queries that are "unkown". Affinity mechanisms work well where the data paths are well understood. Thor HW Andrew Gideon wrote in message <70lkki$cq7@dixie.tagsys.com>... > >We've been discussing some rather interesting details regarding >databases on Linux, and I've learned a few things. But I'm >wondering about scale issues. We've a lot of web activity >which accesses our databases, so this is likely to be relevent >if a little different from the previous contents of this >thread. > >I remember running Sybase on a multiprocessor UNIX machine >(Sequent? I'm no longer sure), and it worked quite well. >One "bound" a dataserver process to a CPU to minimize cache >failures, but it was otherwise fairly transparent. > >I don't recall whether this affinity was a feature of Sybase >or the version of UNIX in question. > >Anyway, how well do the various databases take advantage of >multiple Intel CPUs under Linux? I'm curious about this >"in the large" (ie. multiple DB processes acccessing common >data sets) and "in the small" (ie. a UNION being spread over >processors that are otherwise idle). > >Thanks... > > - Andrew > >