Skip site navigation (1)Skip section navigation (2)
Date:      Sat, 16 Oct 1999 12:58:40 -0700 (PDT)
From:      Mike Meyer <mwm@phone.net>
To:        freebsd-smp@freebsd.org
Subject:   Re: SMP on 4 Pentium III(450NX) failed 
Message-ID:  <Pine.BSF.4.10.9910161240590.31550-100000@guru.phone.net>
In-Reply-To: <199910161823.LAA06715@dingo.cdrom.com>

next in thread | previous in thread | raw e-mail | index | archive | help
On Sat, 16 Oct 1999, Mike Smith wrote:
;->As a general rule, think "more small, cheap machines" rather than "one 
;->big expensive machine".  This is a strategy that has worked well for 
;->many people (yahoo, hotmail, etc.).

Yes, indeed. The exceptions are rare, and getting rarer. Even things
that are heavily compute bound can benefit from this - see the
seti@home project <URL: http://setiathome.ssl.berkeley.edu/ > for an
extreme example.  The other exception that comes to mind is if you
have a large database that needs to stay synchronized, but even then
the worst case is a large, fast DB machine behind an array of small,
fast web servers. If you can partition the DB, then you can replace
that large, fast DB machine with an array of small, fast machines. If
your DB is in some sense fuzzy (by which I mean sharing some of the
properties of the typical web search engine), then even a DB machine
being down is no great loss. Is anyone really going to complain that
you missed 1% of the URLs on the web when a larger percentage than
that are probably out of date?

The headache that comes with it is you have to manage all those
systems. I've not found a clean approach yet. Is anyone working on one
- or better yet, have one?

	<mike




To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-smp" in the body of the message




Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?Pine.BSF.4.10.9910161240590.31550-100000>