Beast runs in Supercomputer

[Emrah Coraman]

Hi,

Is there any guidelines for running beast in a supercomputer?

Best wishes,

Emrah

[Emrah Coraman]

Thank you very much for your answer.

But I have additional questions. I visited the National Center for High Performance Computing and they have some options.

They have one machine with specifications:
HP Integrity Superdome SD32B
Processor Intel Itanium2 1.6 GHz (dual-core)
Number of processors 32
Number of compute cores 64
Memory architecture Shared
Memory amount 128 GB
Disk amount 4.6 TB
High performance network InfiniBand 20 Gbps
Operating system RHEL 5.1 IA64

And another one:
HP ProLiant BL2x220c G5 Blade
Processor Intel Xeon 2.66 GHz (E5430 quad-core)
Number of compute nodes 82 + 23 =105
Number of compute cores 656 + 184 = 840
Memory architecture Distrubuted
Compute node memory amount 16 GB
Compute node disk amount 120 GB
High performance network InfiniBand 20 Gbps
Operating system CentOS 5.4 x86_64

They agreed to support our project. We will have a remote access and so we can start the job.

However,  I do not have any idea if we will need to do any adjustments for running BEAST in those machines or will need Beagle for using them. One of them is a SMP and the other is a MPI.

I will appreciate if someone can comment which will be better and how to install and run BEAST in such computers.

Kind regards,

Emrah



On 16 July 2011 09:07, AJ <plant.bio...@gmail.com> wrote:
>
> BEAST can be run easily on a supercomputer, though some of the
> specifics (i.e., how to tell the supercomputer to run BEAST) vary by
> supercomputer. If you're just interested in running BEAST on -any-
> supercomputer, try the Cipres Portal at http://www.phylo.org/sub_sections/portal/.

[Mats Töpel]

Hi Emrah

The installation of BEAST will be similar on both systems. You don't need to do any modifications to BEAST, in order to run it on the two systems you describe.
Either you ask the system administrator to install it where everyone has access to it (usually /usr/local/bin), or do your own installation somewhere in your 'home'- folder. If you choose the latter, you will be able to upgrade BEAST regularly, without having to ask the administrator for help. But if you are unfamiliar with these things, it is probably better to leave it to someone else to do.

Deciding between the SMP and the MPI system: On the latter you will probably have to write some kind of script to submit your analyses to the queue-system (SGE and Condor are two software packages that handles queues on a cluster). On the other hand, you can reuse the script and will only have to write it ones. Your system administrator can probably help you with that, or have instructions for writing the scripts.

The SMP system may be easier to interact with, if it doesn't have some kind of queue-software installed (Also, BEAST only uses a single core for each analysis, so you don't really need MPI). In that case, you just start the analysis as on a regular computer. Still, you may be interested in using the program "screen" so you don't accidentally terminate your analysis! Use 'man screen' to find out more about that.

It is often (but not always) good to use Beagle when you run your analyses on CPU's. It is probably best to have your systems administrator install it for you, as it can be a bit tricky. When I run BEAST with beagle on our local cluster, I use a script that I call "beagle_optimiser" to find the best options for speed. You can download it from http://matstopel.se/code if you are interested. After installing 'beast', 'beagle' and 'beagle_optimiser', you can run your analyses like this (or including it in your queue script):

    beast_optimiser -r -i <NAME_OF_YOUR_XML_FILE>

This will start your analysis on one of the nodes on the cluster and then test different beagle options to find the one that will make your analysis run as fast as possible on that hardware.

Regards
Mats Töpel

--
You received this message because you are subscribed to the Google Groups "beast-users" group.
To post to this group, send email to beast...@googlegroups.com.
To unsubscribe from this group, send email to beast-users...@googlegroups.com.
For more options, visit this group at http://groups.google.com/group/beast-users?hl=en.