Re: SMP and 7.x performance questions?

This is a multi-part message in MIME format. --------------558A29198FB Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit Fernando Fernandez wrote: > I already heard in several places that SMP will only work well up to 8 CPU. > If this theory is correct, we should only use SMP with 4 to 8 CPU's, then > start with MPP loosely-coupled architectures. This is not necessarily the case. I've seen machines scale well up to 24 procs. What is really critical is the type of work being done. If the work is computational in nature, then having more processors is generally a win. If, on the other hand, you are moving a lot of data around, then bus bandwidths become the limiting factor; once you hit those, it doesn't matter how many processors you have - you just can't squeeze any more data through. Loosely-coupled may only be a good solution if the work doesn't require a lot of data to be shipped around, otherwise you'll hit the same type of limitation. If you are not running client/server, then additional processors can also be a win as your user count increases (i.e. more processor power for more front-end programs). > Now back to Informix... This 7.1 performance problem in few CPU machines > raises another question: should we use disk fragmentation in those cases? > I believe this must be considered carefully because disk fragmentation is > strongly dependent on parallel execution. > > For example: I think that round robin fragmentation is not worth with low > number of CPU's, but hashing is a good policy because of the method of > skipping fragments. > > Any ideas on this? Again, not necessarily. Keep in mind that very little actual cpu is being spent performing the i/o. And i/o is generally one of the slower operations on a machine. So even with a small number of processors, it can be worthwhile to fragment. In the worst case, where you cannot do a parallel scan, how are you better off scanning a large table on one disk versus scanning that table in smaller pieces on several disks? The overhead of setting up the scan on each dbspace is generally noise in relation to the time of the scan itself. -- Dave Kosenko, Informix Professional Services **************************************************************************** While it is true that there is more than one way to skin a cat, the cat himself generally fails to appreciate the differences. --------------558A29198FB Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit Content-Disposition: inline; filename="IFXDISCL.TXT" ************************************************************************* Note: please do not send me email asking about features, or asking about Informix problems and how to solve them. I answer what questions I can in this forum (comp.databases.informix) when I have the time to spare. For questions on features, call your local sales rep or check out the Informix web site (http://www.informix.com). For technical problems, call Informix tech support. ************************************************************************* Disclaimer: All opinions expressed in this message are well-reasoned and insightful; needless to say, they are not those of Informix Software, its partners or lackeys. Anyone who says otherwise is itching for a fight. --------------558A29198FB--