Great Circle Associates List-Managers
(May 1996)
 

Indexed By Date: [Previous] [Next] Indexed By Thread: [Previous] [Next]

Subject: Re: Mailing list hardware question
From: Bill Bogstad <bogstad @ blaze . cs . jhu . edu>
Date: Thu, 30 May 1996 12:47:35 -0400
To: "Jonathan M. Bresler" <jmb @ FRB . GOV>
Cc: Mark J Bradakis <mjb @ spitfire . cs . utah . edu>, list-managers @ GreatCircle . COM
In-reply-to: Your message of "Thu, 30 May 1996 09:30:58 EDT."

Jonathan Bresler writes:

[all fine up to here]

>	ftp, web, mail.  there are all net and disk intensive programs.
>a *good* quality 586 will do fine.  i would recommend against linux.
>its networking and *synchronous* disk speed are not as fast as FreeBSD.
>read the usenix96 paper by mary baker and kevin lai of stanford
>http://plastique.stanford.edu/~mgbaker/publications/usenix96.bench.ps
>context switching will be critical as well.
>
>short summary:
>	context switch:  FreeBSD flat graph @ 100usec
>			 Linux exponential graph crosses FreeBSD at 19 procs
>	network tcp:	 FreeBSD: 65.95MB/s
>			 Linux:   25.03MB/s
>	disk synchronus: FreeBSD: 65ms
>			 Linux:   no data available
>			 Linux: (asynchronous) 0ms
>
>			 0ms means that the data is *not* on the disk.
>			 should something bad happen the data is gone.
>			 ouch.

	I really don't wish to start a flame war about Linux vs. FreeBSD on
this mailing list.  I read enough of that on the associated USENET
newsgroups.  However, I believe one of the recent flamewars uncovered some
issues with this particular paper involving versions of software being used
(old vs. new), relevancy of the numbers being tested, etc.  The developers
of both systems are constantly working on improving their software and since
their source code is available they really can't hide what they are doing.
In other cases, the authors of the systems disagree on whether there is any
benefit to a particular feature.  One area is the issue of meta-data on a
filesystem (not the files themselves, but the directory structure etc.)
FreeBSD uses a filesystem which tries to update this information in such a
manner that the structure is never compromised.  The developers of Linux
essentially say that if the data isn't kept in the same state why do we care
about the directory structure being perfect as long as we can fix it easily
and reliably after a crash.  This translates to less fscking after a crash
under FreeBSD and faster filesystem manipulation under Linux.  (Or at least
that's the way I remember the discussion.)

	In any case, like all benchmarks it should be read critically.
(Hell, all sources of information should be read critically; including this
note and the one to which I'm responding.)

	Let's just say that: Yes, PC hardware can probably handle the load.
Since both OS'es are inexpensive and as far as functionality for your
purpose is concerned are pretty equivalent, why not try both?  Alternatively,
go with the one you are familiar with or others around you are familiar with.
At some point your time is worth more then the money you save by getting a
120 instead of a 166 MHz Pentium...

				Bill
				bogstad @
 cs .
 jhu .
 edu

Indexed By Date Previous: Re: Mailing list hardware question
From: Kjetil Torgrim Homme <kjetilho @ ifi . uio . no>
Next: Re: mailers appending complete replied text
From: Al Gilman <asgilman @ access . digex . net>
Indexed By Thread Previous: Re: Mailing list hardware question
From: Kjetil Torgrim Homme <kjetilho @ ifi . uio . no>
Next: Re: Mailing list hardware question
From: "Jonathan M. Bresler" <jmb @ FRB . GOV>

Google
 
Search Internet Search www.greatcircle.com