Hardware requirements and cluster scalability
Good day to everybody at the devel-list. And greetings from Finland! The library of Joensuu has started a project to compare different open source library systems, mainly Koha and Evergreen, with the goal of moving from our current closed source system to open source. We are planning to create a consortium of minimum of a ~20 libraries, but if open ILS are found to be superrior, probably the consortium can easily grow to ~250 libraries with ~5 million articles and ~150 000 transactions/hour. I have been hard pressed to find answers to the hardware requirements of Koha and Evergreen and have been instructed to ask here on devel-list by your library managers. My main interests are the hardware requirements of the said large consortium mainframe. How well does it cluster? What sort of hard disk and ram requirements are there? How many pc units with for example 2GHz cpu, 2Gb RAM do we need? Could every library host their own databases or should they be centralized? What are the average search speeds using simple SELECTs and then complex JOINs. What are the system bottlenecks? Can the OPAC be integrated with Drupal/Liferay so it can be compatible with our new web-services strategy? Are there any publications I could read about these topics? Happy coding! -- Olli-Antti Kivilahti Open Library 2013 Library of Joensuu
On Wed, 25 May 2011 12:01:23 +0300, Olli-Antti Kivilahti <olli-antti.kivilahti@jns.fi> wrote:
...We are planning to create a consortium of minimum of a ~20 libraries, I have been hard pressed to find answers to the hardware requirements of Koha and Evergreen and have been instructed to ask here on devel-list by your library managers.
I cannot answer your hardware questions yet, maybe in a month. But surely by then you'll have your answer. But I can comment on this:
Could every library host their own databases or should they be centralized?
If you have data centralized then your libraries would have to share some common policies among *ALL* libraries. AFAIK there's no bibliographic/system separation in koha, so you cannot have a separate centralized database for your bibliographic data and individual databases for the rest of the stuff. Here (UNCuyo, Mendoza, Argentina) we're about to implement a centralized koha for ~20 libraries, but all of them share the same rules, they must obey a central organism which dictates the policies. I believe trying to centralize data from libraries which have separate rules might give you major headaches. -- Fernando Canizo (a.k.a. conan) - http://conan.muriandre.com/ GCS d? s:+ a C++ P--- L++++ E--- W+++ w--- M-- PE-- !tv b+++ h---- y+++
Le 25/05/2011 11:01, Olli-Antti Kivilahti a écrit :
Good day to everybody at the devel-list. And greetings from Finland! Hi Finland, here is France ;-)
The library of Joensuu has started a project to compare different open source library systems, mainly Koha and Evergreen, with the goal of moving from our current closed source system to open source. We are planning to create a consortium of minimum of a ~20 libraries, but if open ILS are found to be superrior, probably the consortium can easily grow to ~250 libraries with ~5 million articles and ~150 000 transactions/hour. Our largest customer is a little bit smaller, but not that much : 48
(just FYI : there is a monestary in finland that uses Koha. Ask Monk Viktor (mviktor@valamo.fi) ) libraries, 1.3 millions items (and 760k biblios), something like 1 million check-out per year. The hardware is a bi-quadri-core (8 cores), 12GB RAM, RAID10 hard disks (84GB used for system/koha/zebra, and 26GB used for mySQL) The server is far from being overloaded. There is no specific hard disk, and no NAS or something like that. Just take care of what I wrote on our blog : http://www.biblibre.com/en/blog/entry/mysql-default-config-and-very-large-ko... About the pc units, we haven't changed any computer. So some are very old (5years or more). The only requirement is to be able to run Firefox. So even a winXP should be OK. The server is centralized and host the 48 libraries catalogues, that are considered as one. I don't see how you could have every library hosting their own database if you want to have a true network (ie: with easy transfers/holds/check-in/out) between branches. The OPAC can be integrated with Drupal through SOPAC. We've made the ILS-DI connector for Koha, and the ILS-DI connector for Drupal. However, if you want to investigate deeper, you should drop me a mail, there are some problems we've solved and that are not on official SOPAC (in 2 words : the sopac guys want something working with any ILS and any portal, we think it's better to use all drupal specific features) I'm not sure though that it would work well with a 5 million records database. HTH -- Paul POULAIN http://www.biblibre.com Expert en Logiciels Libres pour l'info-doc Tel : (33) 4 91 81 35 08
participants (3)
-
Fernando L. Canizo -
Olli-Antti Kivilahti -
Paul Poulain