[ale] Compaq Reliant web server
Jeff Hubbs
Jhubbs at niit.com
Thu Jul 12 10:43:35 EDT 2001
John
-
<SPAN
class=538232713-12072001>Â
The
strongly-held opinion I have formed over the years is that no multi-processor
box should be bought without all CPU sockets fully populated. You're
paying extra (especially if you go with "name" vendors like Compaq) just to have
even a 2-CPU system, but every day that it goes with one or more CPU slots empty
represents lost value, i.e., you are not getting any returned "power" for the
money you spent. The "you can always add CPUs later" argument is always
trotted out but in practice, that's often not possible without also replacing
the original CPU(s) simply because duplicates are no longer available.Â
Even at that, I've encountered people whose newly-SMPed machine never worked
quite right again until two CPUs of the same speed, stepping, and
revision were used. The oringinal CPU(s) displaced then go on to represent
unreturned value unless and until more money is spent to put a machine around
them.Â
<SPAN
class=538232713-12072001>Â
What
Joe Knapka said about the 32MB P/133 isn't a joke. For 2000 hits/day =
43.2 seconds/hit, a 486DX would do unless it's too slow for you to do admin-type
things like tar/zip a bunch of files. It surprises me about how
miscalibrated people can be about how much machine it takes to do a given
thing. A few months ago, I was testing out a Web server configuration
using Rational SiteLoad, and for the heck of it I aimed it at a P/200 running
Zope, with a user script that just pulled out a lot of static content (no
DB). The 512MB PIII/667 that I was running SIteLoad on started to get in
to the weeds (and therefore into the range of invalidating the test) at about 40
simultaneous simulated users, but the target Zope box seemed like it was only
just starting to feel it - and that's without recompiling the kernel, the
modules, and python for 586.
<SPAN
class=538232713-12072001>Â
This
is why I value "junk" machines so much; they give the Linux/*BSD person great
power.
<SPAN
class=538232713-12072001>Â
I
would say that if they are going to even remotely consider a "dualie," they
ought to use it as a killer file or DB server. I'm starting to realize
that a fast file server on a good network is one of the coolest things an
organization can have.
<SPAN
class=538232713-12072001>Â
-
Jeff
<SPAN
class=538232713-12072001>Â
<SPAN
class=538232713-12072001>Â
<SPAN
class=538232713-12072001>Â
<FONT face=Arial
color=#0000ff>Â Â -----Original
Message-----From: Armsby John-G16665
[mailto:John.Armsby at motorola.com]Sent: Wednesday, July 11, 2001 5:17
PMTo: 'ale at ale.org'Subject: [ale] Compaq Reliant web
server
The IT group here is junking an HP box with my web
stuff. I told them I did not want to port to NT, that I preferred to
stay with Unix. I suggested Linux as an alternative on an Intel
box. They accept the idea. Hooray!Â
They are suggesting a Compact Reliant series server
which comes with a RED HAT CD. They want to know if I need dual
processors. Frankly the box is only getting about 2,000 hits in a 24
hour period. The current HP 9000 workstation has 256 meg of ram and a
single 200 megahertz processor. TOP is usually less than .1 except when
I try to tar several thousand files over several days and then it jumps to
1.03 or so. There is no discernable on the WEB server.
My question is: Am I correct in assuming that an
out of the box installation of Mandrake 8.0/Red Hat 7.0 or equivalent will not
by default see two processors? Don't I have to do a kernel recompile and
select multi processor?Â
John
___________________NOTICE____________________________
This electronic mail transmission contains confidential information intended only for the person(s) named. Any use, distribution, copying or disclosure by any other person is strictly prohibited. If you received this transmission in error, please notify the sender by reply e-mail and then destroy the message. Opinions, conclusions, and other information in this message that do not relate to the official business of NIIT shall be understood to be neither given nor endorsed by NIIT. When addressed to NIIT clients, any information contained in this e-mail is subject to the terms and conditions in the governing client contract.
More information about the Ale
mailing list