Date: Thu, 30 May 2002 14:13:28 +0100 (BST) From: Jan Grant <Jan.Grant@bristol.ac.uk> To: "Lance M.Westerhoff" <lmw188@psu.edu> Cc: freebsd-questions@FreeBSD.ORG Subject: Re: DB2 on FreeBSD?? Message-ID: <Pine.GSO.4.44.0205301402010.14897-100000@mail.ilrt.bris.ac.uk> In-Reply-To: <40948B54-732F-11D6-B5D0-00039357F10C@psu.edu>
next in thread | previous in thread | raw e-mail | index | archive | help
On Wed, 29 May 2002, Lance M.Westerhoff wrote: > I'm not sure if this is the right list to send this question on, but.... > > In the coming weeks we will be building a 20 processor (10 node) AMD > 2000+ MP cluster, and I would really like us to try to use FreeBSD > instead of Linux. If these machines would only be used to run > computations, I suspect we would have no trouble at all. But, we will > also be connecting a 1.5TB RAID to two of the machines (one will be used > to serve databases and web pages, and the other will be a file and > application server). Your first problem is that currently, the stable (release) track of FreeBSD cannot support filesystems over 1TB in size. There is current work (in -current) on UFS2 which addresses this issue. You should, however, be able to partition a large RAID array and use multiple sub-11TB slices happily. > What about database > software? As the databases grow, I expect each of them to encompass > several TB, so I don't think MySQL will work in the long run (though I > could be wrong). You may well be. MySQL uses single files per table, so you'll have a practical limit of 1TB per table inherited from the filesystem unless you look at UFS2. Oracle gives you quite a bit more control over where and how your tables are distributed; you have _some_ control over this with PostgreSQL but currently a table has to (effectively) sit in one file. PostgreSQL has some "object relational" features but that basically means that one table can "inherit" from another; it can make some modelling simpler but pgsql is _not_ an OODBMS. > Lastly, what sort of performance hit, if any, can > we expect with using a Linux-native DBMS with FreeBSD's Linux > compatibility libraries? Providing the application works (if it doesn't, follow up with the emulation crowd), there isn't a fixed overhead for linux "emulation". The linuxulator works by using a different syscall vector and having a set of lightweight linux shims over the freebsd kernel; so running code goes at the same speed, basically. Performance differences will come from fundamental subsystems like the VM system and the networking stack. Here, opinions basically vary and your best bet would be to run your own application-specific benchmarks if you're concerned about accurate performance measurements. If you decide to do this then it's worth following up with your results to the appropriate (freebsd or linux) crowd because you will almost certainly get good tuning advice thrown in for free with every explanation of why one is superior to the other :-) -- jan grant, ILRT, University of Bristol. http://www.ilrt.bris.ac.uk/ Tel +44(0)117 9287088 Fax +44 (0)117 9287112 RFC822 jan.grant@bris.ac.uk Lambda calculus? I hardly know 'er! To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-questions" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?Pine.GSO.4.44.0205301402010.14897-100000>