[OAI-general] RE: OAI-general digest, Vol 1 #74 - 1 msg

[H.M. Gladney]

There is a bothersome implicit assumption in these notes--that an
e-print archive needs to manage (part of) its own servers.

Another possibility is that an archive buy storage and Web services from
one of giant server farm companies that many businesses find practical.
Doing so is likely to have several large advantages over any kind of
"roll your own" (aka RYO) solution: avoided administrative headaches,
automatic capacity management (as alluded to in the attached note), very
short delay between proposal and service delivery, and probably (TBD)
lower cost than RYO.

If someone can identify strong arguments against this proposition, I'd
much enjoy hearing them and perhaps having a friendly debate.

Regards, Henry

Henry Gladney, Ph.D       (408)867-5454
HMG Consulting
20044 Glen Brae Drive, Saratoga CA 95070
http://home.pacbell.net/hgladney/

----Original Message-----

> -----Original Message-----
> From: ePrints Support [mailto:support@ecs.soton.ac.uk]
> Sent: Friday, January 18, 2002 3:39 AM
> To: Bob Kemp
> Cc: EPrints Underground List; jisc-dner@JISCMAIL.AC.UK;
> september98-forum@amsci-forum.amsci.org;
> oai-general@oaisrv.nsdl.cornell.edu
> Subject: Re: [OAI-general] Re: [EP-underground] Sustaining ePrints
>
>
> One issue in the design of eprints-2 is that it is geared up to allow
> many (smaller) archives to be run on one physical computer.
> Which means
> that a larger organisation could buy one "chunky" computer and then
> provide eprints archives with seperate customisations, OAI
> exports (and
> colour schemes) for various departments which don't have the
> resources and
> expertise to do this themseleves. This should reduce the cost
> in parts (and
> system admin hours!) to get several archives going.

In addition, should a particular archive become heavily used, moving it
to a
dedicated box is a snap. You really loose nothing with a multi-archive
(versus monolith) design, and gain so much. Thanks to OAI protocols,
developing an institute wide interface to all the small archives isn't
much
of a hassle -- in fact kinda unnecessary due to ARC
(http://arc.cs.odu.edu/). If you can't tell, I'm a fan of this design.

> This would probably work out more cost effective than buying
> and installing
> a machine and having learn the ins-and outs of the system for a single
> archive. EPrints is free, so we don't have to grub for
> licenses by requiring
> one copy per archive :)

Hurrah for free (GPL) software!

> My own server, running eprints-alpha-2, is running
> demoprints.eprints.org,
> eprints.aktors.org + a couple of other test archives. Later,
> it will be
> running our department publications archive and an archive of
> Ted Nelsons
> material.
>
> The bottle neck is processor power, caused by larger datasets
> but mostly
> by "hits". So large, popular archives should probably run on
> their own
> machine.

Some more metrics: on my un-optimized server, eprints/apache/mysql
altogether require about 40MB of memory per archive. Most web hits are
from
robots and some from OAI harvesters (as it should be).  We have 10
archives
now, though some are just skeletons waiting for our authors to "see the
light" so to speak.

=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=
Ed Sponsler - Sr. Applications Development
Caltech Library System
eds@library.caltech.edu
http://library.caltech.edu/digital


--__--__--

_______________________________________________
OAI-general mailing list
OAI-general@oaisrv.nsdl.cornell.edu
http://oaisrv.nsdl.cornell.edu/mailman/listinfo/oai-general


End of OAI-general Digest