EPrints.org

• Software
• Services
• About Open Access

eprints_tech messages

From: "Mariano A. Belladonna" <mariano.belladonna AT cab.cnea.gov.ar>
Date: Thu, 25 Jun 2009 10:46:00 +0200

http://mailman.ecs.soton.ac.uk/mailman/listinfo/eprints-tech
*** EPrints community wiki - http://wiki.eprints.org/
Hello everyone,

I'm in the process of evaluating and deciding which is the most 
appropriate hardware platform for the dedicated server where we're going 
to install the institutional repository on. I know this involves many 
factors including the expected size of the documents to be uploaded by 
authors, the expected traffic and use of the site, the important of 
redundancy and/or backups and so on, but, considering I don't have any 
practical experience on determining this kind of things, and considering 
also that this is the first attempt for the institution to set an IR, I 
thought that basing on other's IR experiences might help me deciding.

To sum up: What HW platform are you running your repository software on? 
Based on your experience, what HW would you recommend to set up a 
dedicated server for an IR?

As much details as possible would be greatly appreciated.
Thanks for your help.

From: Ben Wheeler <b.wheeler AT ulcc.ac.uk>
Date: Thu, 25 Jun 2009 23:39:54 +0100

http://mailman.ecs.soton.ac.uk/mailman/listinfo/eprints-tech
*** EPrints community wiki - http://wiki.eprints.org/
On Thu, Jun 25, 2009 at 10:46:00AM +0200, Mariano A. Belladonna wrote:
> To sum up: What HW platform are you running your repository software on? 

We use IBM X-series servers running Red Hat Linux. 
Linux is the best OS for running EPrints. Red Hat wouldn't be 
my choice of distro (I'm a Debian fan) but it's ok. We've been
using IBM x-servers for some years and found them very reliable;
disks fail from time to time (as with any machine), that's about it.

> Based on your experience, what HW would you recommend to set up a 
> dedicated server for an IR?

Two PC servers: one development, one production. 

The dev machine can be lower spec than the production but the OS and
all the software should be the same.

Some sort of hardware RAID, either using disks within the servers,
or in an external thing (we use internal disks in RAID-5). Basically
you want the machine to stay up if a disk fails, and to be able to
hot-replace it, and you'll also benefit from faster disk accesses.

Spec your diskspace for at least twice as much space as you think
you'll actually need, or make sure you have slots spare to add more
later. Use LVM so you can add new disks to existing partitions.
If you have a lot of journals/subjects/types/people and generate
views for them all, the views can easily take up far more diskspace
than the documents themselves. 

As much other redundancy within the machine as you can afford, eg
dual power supplies. 

Or two simpler PCs on a failover switch for the production service.
(Not sure if EPrints is compatible with load-balancing across multiple
servers)

4G memory at least, possibly much more depending on your expected load.

Some sort of backup solution where you keep multiple backups, possibly
for a very long time (months) as sometimes if something goes missing
no-one notices for a very long time. Regular offsite backups.

Ben

From: Yuri <yurj AT alfa.it>
Date: Fri, 26 Jun 2009 09:44:19 +0200

http://mailman.ecs.soton.ac.uk/mailman/listinfo/eprints-tech
*** EPrints community wiki - http://wiki.eprints.org/

> 4G memory at least, possibly much more depending on your expected load.
>   

 4G? :-/

From: Zeno Tajoli <tajoli AT cilea.it>
Date: Fri, 26 Jun 2009 09:59:17 +0200

http://mailman.ecs.soton.ac.uk/mailman/listinfo/eprints-tech
*** EPrints community wiki - http://wiki.eprints.org/
Hi to all,


about this topic:
I confirg the preference to a Debian Linux as OS
for hadware: go to DELL / HP web site and select a server with no 
more than one year.
Very important: 4 GB of RAM
I subscribe all consideration of Ben Wheeler

Bye

Zeno Tajoli
CILEA - Segrate (MI)
tajoliAT_SPAM_no_prendiATcilea.it
(Indirizzo mascherato anti-spam; sostituisci quanto tra AT con  AT )

From: Ben Wheeler <b.wheeler AT ulcc.ac.uk>
Date: Fri, 26 Jun 2009 10:43:24 +0100

http://mailman.ecs.soton.ac.uk/mailman/listinfo/eprints-tech
*** EPrints community wiki - http://wiki.eprints.org/
On Fri, Jun 26, 2009 at 09:44:19AM +0200, Yuri wrote:
> http://mailman.ecs.soton.ac.uk/mailman/listinfo/eprints-tech
> *** EPrints community wiki - http://wiki.eprints.org/
> 
> > 4G memory at least, possibly much more depending on your expected ↵
load.
> >   
> 
>  4G? :-/

? 

What's the problem with that?

From: BillN <billn AT fairpoint.net>
Date: Fri, 26 Jun 2009 06:37:27 -0700

http://mailman.ecs.soton.ac.uk/mailman/listinfo/eprints-tech
*** EPrints community wiki - http://wiki.eprints.org/
I would add to Ben Wheeler's excellent recommendations:

1. UPS power supply
2. For large sites performance needs: A separate disk/channel for the 
*views* of journals/subjects/etc. They need not be Raid protected as 
they can be regenerated as needed, but Raid1 offers some read speed 
advantages.

BillN
<Systems guru>