What you should do (in my opinion) is install a grid scheduler (SGE, torque, 
etc) on the big server.  If you run Galaxy on a separate server, it can be 
configured to submit jobs to the scheduler.  Galaxy also has the concept of a 
Web App and Handler components.  Essentially, handlers take care of talking 
with the scheduler while the Web App will serve pages to users.  By default, 
the Web App and Handlers are combined in the same process.  you can configure 
galaxy to start up multiple handler processes and multiple web app processes.  
Then, you can use Apache or Nginx to load balance user requests between the 
various Galaxy Web Apps.

My recommendation is that you start small to accommodate about 5 concurrent 
users without noticeable performance issues:
1 Galaxy web app
1 Galaxy handler

Recommended reading:
https://wiki.galaxyproject.org/Admin/Config/Performance/Scaling 
https://wiki.galaxyproject.org/Admin/Config/Performance/Cluster


-----Original Message-----
From: Shane Kelly [mailto:s...@shanek54.co.uk] 
Sent: July-29-15 7:32 AM
To: Kandalaft, Iyad
Cc: galaxy-dev@lists.galaxyproject.org
Subject: Re: [galaxy-dev] I need some advice on what type of Galaxy server to 
implement

Hi Iyad,
        Thanks for taking the time to get back to me.
I am an IT guy, and would not know how to answer these questions except to say 
that the server is to be used to facilitate biological and medical research in 
the areas of genomics, transcriptomics, epigenomics and metagenomics. (straight 
from the mission statement :-) )

        I think that they would say  that overall performance is less important 
than being able to run large datasets (1/2-3TB).

        I like the idea of a separate box for the web-server, but I am not sure 
how the web server would communicate with the pipeline box - is that ability 
built into galaxy, or is it a well worn path with plenty of examples that I 
could plagiarize?

        Sorry to be such a newb, but I don't know much about galaxy at all. 
Luckily I have 2-3 months to put this in place...

Again, thank you for your time.

Regards,
Shane




>When you say NGS, is it genome assembly?  If so, what type of genomes 
>and do you have experience with its memory and cpu requirements.  We 
>noted that servers with large amount of memory and cores have a memory 
>bus bottleneck. The other aspect is high processing on the server will 
>impact the performance of Galaxy unless it is given higher priority.
>Note that if you overcommit the server, it can destabilize and bring 
>down the Galaxy web app and database.
>
>My general approach is Galaxy web app + proxy on a separate machine 
>from the handlers.  The analysis server is either running a grid or the 
>handlers.
> 
>I recommend multiple smaller servers if you can get away with it as 
>long as you have one that can accommodate your LARGE workloads.  If you 
>don't care about overall performance, large servers are the way to go 
>as they are more "versatile".
>
>Regards,
>
>Iyad Kandalaft
>
>Acting Chief Bioinformatician in Biodiversity, STB Agriculture and 
>Agri-Food Canada / Government of Canada iyad.kandal...@agr.gc.ca / Tel: 
>613-759-1228 / TTY: 613-773-2600
>
>Bioinformaticien chef de la  biodiversite interim, Direction générale 
>des Science et de la technologie Agriculture et Agroalimentaire Canada 
>/ Gouvernement du Canada iyad.kandal...@agr.gc.ca / Tel:
>613-759-1228 / TTY: 613-773-2600
>
>
>
>
>-----Original Message-----
>From: galaxy-dev [mailto:galaxy-dev-boun...@lists.galaxyproject.org]
>On Behalf Of Shane Kelly Sent: July-28-15 8:23 AM
>To: galaxy-dev@lists.galaxyproject.org
>Subject: [galaxy-dev] I need some advice on what type of Galaxy server 
>to implement
>
>Hi
>       I have been tasked with getting a Galaxy server up and running
>       for a group at work.
>
>       1. No-one can tell me how many users (concurrent or otherwise)
>       there will be 2. Most of the analyses will be NGS.
>       3. Tools will be developed in-house but we will use public
>       domain tools also. 4. There will be a guy running the
>       server/developing tools pretty much full time.
>
>       I have two favoured solutions at the moment:
>
>1. A pipeline processor ( 64 Core, 512G Ram, with DAS of about 150TB ), 
>and a Web server to act as frontend and database server, and another, 
>smaller box for a total install of galaxy, but doing only the 
>development work.
>
>2. An all-in-one server with 128 Cores, 1TB ram, DAS storage of 150TB, 
>and development work done on a VM.
>
>Any input would be hepfull.
>
>Thanks
>Shane
>___________________________________________________________
>Please keep all replies on the list by using "reply all"
>in your mail client.  To manage your subscriptions to this and other 
>Galaxy lists, please use the interface at:
>  https://lists.galaxyproject.org/
>
>To search Galaxy mailing lists use the unified search at:
>  http://galaxyproject.org/search/mailinglists/
>___________________________________________________________
>Please keep all replies on the list by using "reply all"
>in your mail client.  To manage your subscriptions to this and other 
>Galaxy lists, please use the interface at:
>  https://lists.galaxyproject.org/
>
>To search Galaxy mailing lists use the unified search at:
>  http://galaxyproject.org/search/mailinglists/

___________________________________________________________
Please keep all replies on the list by using "reply all"
in your mail client.  To manage your subscriptions to this
and other Galaxy lists, please use the interface at:
  https://lists.galaxyproject.org/

To search Galaxy mailing lists use the unified search at:
  http://galaxyproject.org/search/mailinglists/

Reply via email to