Date: Wed, 6 Dec 2006 10:06:56 -0500 From: John Baldwin <jhb@freebsd.org> To: Josh Paetzel <josh@tcbug.org> Cc: freebsd-chat@freebsd.org Subject: Re: Venting my frustration with FreeBSD Message-ID: <200612061006.56852.jhb@freebsd.org> In-Reply-To: <200612051707.46705.josh@tcbug.org> References: <200612041443.15154.josh@tcbug.org> <200612051736.47980.jhb@freebsd.org> <200612051707.46705.josh@tcbug.org>
next in thread | previous in thread | raw e-mail | index | archive | help
On Tuesday 05 December 2006 18:07, Josh Paetzel wrote: > On Tuesday 05 December 2006 16:36, John Baldwin wrote: > > On Tuesday 05 December 2006 17:06, Josh Paetzel wrote: > > > On Tuesday 05 December 2006 15:36, pete wright wrote: > > > > On 12/5/06, Josh Paetzel <josh@tcbug.org> wrote: > > > > > On Tuesday 05 December 2006 11:19, Nick Hibma wrote: > > > > > > > 1) SMP scalability. 4-way boxes are relatively common, > > > > > > > and hardware with higher CPU counts is only going to get > > > > > > > more and more common. I'm no industry expert, but 5 years > > > > > > > from now will my clients be considering buying 32 and 64 > > > > > > > way boxes? Possibly. Will FreeBSD be in a positiion to > > > > > > > compete favorably vs. the alternatives on such hardware? > > > > > > > > > > > > People have been working on this for years. It's a > > > > > > difficult thing to get right. Sun has been spending a *LOT* > > > > > > of time doing this for Solaris, and I bet that even Linux > > > > > > isn't there yet. > > > > > > > > > > Linux actually scales very well in this area. My friends in > > > > > the supercomputer business tell me that people are > > > > > successfully using linux on 1024-way SSI boxes. It doesn't > > > > > scale quite as well as IRIX, but a lot of people opt for > > > > > linux anyways. > > > > > > > > > > For instance, NASA Columbia, which is a cluster of 20 512-way > > > > > SSI Altix's is successfully running linux, and comes in #8 on > > > > > top500.org's supercomputer list. > > > > > > > > yea, i'm pretty familiar with those systems and i would have to > > > > say that the Altix is indeed quite impressive. but, i would > > > > not equate the ability for SGI to implement a large SSI cluster > > > > like this to a "normal" user being able to implement a similar > > > > setup with a stock linus kernel or stock distro for that > > > > matter.... > > > > > > > > -pete > > > > > > What sort of 'normal' user has access to that kind of hardware? > > > > > > Of course they aren't running a stock kernel or distro, but > > > neither are a lot of the guys using linux on real-time embedded > > > hardware. Google doesn't run a stock kernel or distro either, and > > > Verio and Yahoo don't run stock FreeBSD distributions or kernels > > > either. > > > > I would wager that Yahoo's FreeBSD kernel is a lot more stock than > > the Altix one for Linux though. I think the poster's point is that > > you aren't going to get an OTS OS to run on a 512-way cluster, and > > that if one had time and hardware one could probably hack FreeBSD > > up a bunch to run on a 512-way system just as SGI hacked up Linux. > > Not to be pedantic, but the 512 way systems I mentioned are SSI boxes, > not clusters. Granted I did mention a cluster of SSI boxes.... > > But yes, your point is taken. My point is no one has bothered (AFAIK) > to do the needed work to FBSD. Yes, but is that due to errors on the part of FreeBSD or is that because some professor likes Linux better and so he tasked his graduate students to go hack up a variant of Linux that runs on a 512-way machine? Scaling on a 512-way machine is quite a different ball of wax from scaling on 4-way, and scaling up to 32 and 64 is going to be another ball of wax as well. -- John Baldwin
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?200612061006.56852.jhb>