[OPEN-ILS-GENERAL] Suggested Server Specs

[Susan Scott]

Hi, Don,

We are considering Evergreen as a replacement for our SIRSIDynix program 
which will die next year.

We were wondering if we can put AquaBrowser to use with Evergreen.  Do 
you know?

Thanks, Susan Scott, Palo Alto Unified

Don McMorris wrote:

> HI Karen:
> I'm going through a backlog of e-mail, and came across this message.
> I apologize for the delay!
>
> I'm going to post an answer on-list "for the record".  If you have any
> more questions, please don't hesitate to contact the lists again in
> the future!
>
> There aren't "hard-and-fast" rules per se as far as server specs.
> Evergreen is _very_ flexible in server configuration.  It also runs on
> commodity hardware (IE: you don't need a huge Alpha server for it!
> ;)).
>
> Although you could _probably_ get by with a single server, at least
> having a couple for redundancy is advised.  The beauty of the OpenSRF
> platform (which Evergreen is built upon) is it's _very_ easy to add
> systems to if you need (IE: if you decide you don't have enough
> servers initially, you can just add a couple without much modification
> of your existing setup!).
>
> I would advise initially 2 identical machines with the following:
> ->Standard x86 server (single or multiple processors... your call
> ->RAM: As much as possible (2-4GB would be a good number)
> ->Hard drives: Multiple SCSI or SATA2 in a RAID5 configuration with
> hardware controller with an end size of greater than 80GB
> ->Dual Gigabit Network Interfaces (one "lan" will be for the cluster
> communication, the other for the public Interface)
> These would be your main "worker" machines (mirrored databases,
> redundant jabber server configurations, most OpenSRF modules).  If you
> desired, you could even utilize the resources of existing systems...
> For example, your existing web server could be retrofitted for the
> OPAC.  If you wanted to, other servers could have the OpenSRF system
> loaded on them with some of the workers to reduce the load on the
> "dedicated" servers.
>
> If you find that there is a bottleneck on the servers, the OpenSRF
> system allows you to dynamically add another server with minimal
> change to the existing infrastructure.  Also, if a system should fail,
> the effect should be minimal.
>
> Just as a reminder, you don't have to support it in-house if you do
> not want to.  There are companies whom will provide contractual
> support for you.  Anything from them hosting the server at their
> location to configuring and installing the server or maybe for just
> some occasional emergency support.  You are NOT alone with Evergreen!
> And, of course, if you DO decide to support it yourself, we have the
> community mailing lists if you happen to run into trouble.
>
> Again, I apologize for the delayed reply! If you have any more
> questions, please don't hesitate to get in touch with the lists again!
>
> Sincerely Yours,
> --Don McMorris
> On 5/7/07, Karen Collier <kcollier at kent.lib.md.us> wrote:
>
>> Hi,
>>
>> I'm wondering if anyone here can offer some advice on server specs.  
>> I work for a small public library, and we're trying to set up a test 
>> server ourselves that could later become the actual production server 
>> for Open-ILS.  We don't want to sink unnecessary money into the 
>> server, but we also don't want it to be ridiculously slow and unable 
>> to handle the load.
>>
>> We have a little over 50,000 unique titles, with duplicates bringing 
>> us to around 60,000 items.  We have between 10 and 11 thousand 
>> patrons, from a population of 19,000.  We have 3 branches (one main 
>> and two satellite).
>>
>> Any suggestions on number of servers, memory, hard drive space, and 
>> whatever other specs matter would be appreciated.  If the answer is 
>> different for minimum or recommended, I'd like to hear both numbers 
>> and compare.
>>
>> Thank you,
>> Karen Collier
>> Public Services Librarian
>> Kent County Public Library
>> Chestertown, MD
>>
>>