Running BEAST on computer cluster

[Gudni Magnus Eiriksson]

Dear all

The Nordic countries (Denmark, Iceland, Norway and Sweden) have recently started operating a common supercomputer (http://nhpc.hi.is/). I have been inquiring about running BEAST on the system. I understand that the available version of BEAST only operates on one processor. Is there a special version for supercomputers / computer clusters?

Best regards,

Gudni

[Giap Nguyen]

Good morning

 

I think you may want to try the following tetragrid:

http://www.phylo.org/portal2/login!input.action

 

Best regards

Giap Nguyen

2012/2/13 Gudni Magnus Eiriksson <gmeir...@gmail.com>

--
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.

[Miller, Mark]

Wayne Pfeiffer of SDSC implemented this for CIPRES. I am sure he can tell your folks how he set it up on beagle if you prefer to run it on your resources...

Mark

[Gudni Magnus Eiriksson]

Thanks Giap

I thought that the tetragrid was only for US nationals.
I will check it out...

Best regards,

Gudni

[Gudni Magnus Eiriksson]

Thanks Mark

I have been running BEAST on computers accessible to me, but without
BEAGEL. The speed is a bit slow (or the runs a bit large). Do you know
how much BEAGLE speeds things up, twofold speed, or.. ?

Best regards,

Gudni

On 14/02/2012, Miller, Mark <mmi...@sdsc.edu> wrote:
> Wayne Pfeiffer of SDSC implemented this for CIPRES. I am sure he can tell
> your folks how he set it up on beagle if you prefer to run it on your
> resources...
>
> Mark
>
>
>
> On Feb 13, 2012, at 8:44 PM, "Giap Nguyen"

> <nguyen...@gmail.com<mailto:nguyen...@gmail.com>> wrote:
>
> Good morning
>
> I think you may want to try the following tetragrid:

> <http://www.phylo.org/portal2/login!input.action>http://www.phylo.org/portal2/login!input.action

>
> Best regards
> Giap Nguyen
>
> 2012/2/13 Gudni Magnus Eiriksson

> <<mailto:gmeir...@gmail.com>gmeir...@gmail.com<mailto:gmeir...@gmail.com>>

> Dear all
>
> The Nordic countries (Denmark, Iceland, Norway and Sweden) have recently
> started operating a common supercomputer

> (<http://nhpc.hi.is/>http://nhpc.hi.is/). I have been inquiring about

> running BEAST on the system. I understand that the available version of
> BEAST only operates on one processor. Is there a special version for
> supercomputers / computer clusters?
>
> Best regards,
>
> Gudni
>
> --
> You received this message because you are subscribed to the Google Groups
> "beast-users" group.

> To post to this group, send email to <mailto:beast...@googlegroups.com>
> beast...@googlegroups.com<mailto:beast...@googlegroups.com>.

> To unsubscribe from this group, send email to

> <mailto:beast-users%2Bunsu...@googlegroups.com>
> beast-users...@googlegroups.com<mailto:beast-users...@googlegroups.com>.

> For more options, visit this group at
> <http://groups.google.com/group/beast-users?hl=en>
> http://groups.google.com/group/beast-users?hl=en.
>
>
> --
> 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<mailto:beast...@googlegroups.com>.

> To unsubscribe from this group, send email to

> beast-users...@googlegroups.com<mailto:beast-users...@googlegroups.com>.

[Mats Töpel]

Hi Gudni

I did some tests with Beagle before I installed Beast/Beagle on our cluster. The results can be found here http://matstopel.se/files/MT_Beast-Beagle.pdf. As you can see on slide 4 and 5, the result (at least when running on CPU) very much depends on the computer (called "Main server", "node0" and "node1") and dataset used (I used the two files "benchmark[1-2].xml" distributed with beast). On these three computers, the execution time was approximately half the time compared to using the compiled native C (?) library. Therefore I wrote the script beagle_optimiser.py (http://matstopel.se/code) to find the "optimal" settings for speed. Many more beagle options could/will be added to the script in order to find better settings.

Also see the results at http://beast.bio.ed.ac.uk/Benchmarks

Hope this helps
Cheers
#Mats

Mats Töpel (PhD)
University of Gothenburg
Department of Plant and Environmental Sciences
Phone: +46 (0) 70 406 52 92
Skype: mats.topel
Twitter: @matstopel
Web: www.matstopel.se
Laboratory notebook: www.matstopel.se/notebook

[Gudni Magnus Eiriksson]

Many thanks Mats
This will become helpful...
-Cheers, Gudni

[Miller, Mark]

Hi Gudni,

A list of our BEAST/Beagle benchmarks are posted here:
http://www.phylo.org/tools/beast_how_fast.html

As configured we get up to 8 fold or so speedup depending on the data set.
We aren't using GPUs, because we found that the data sets we see routinely
Do not get much benefit from them, but there are situations where this might get you even more speedup.

I believe those benefits are realized when the user has a large number of partitions.
But others can say more about that.

Our resource is available for users anywhere up to 30,000 cpu hours; so if you are running
Huge numbers of data sets or very large runs, you may be best running on your resources.

Best,
Mark

-----Original Message-----
From: beast...@googlegroups.com [mailto:beast...@googlegroups.com] On Behalf Of Gudni Magnus Eiriksson
Sent: Tuesday, February 14, 2012 1:33 AM
To: beast...@googlegroups.com
Subject: Re: Running BEAST on computer cluster

Thanks Mark

I have been running BEAST on computers accessible to me, but without BEAGEL. The speed is a bit slow (or the runs a bit large). Do you know how much BEAGLE speeds things up, twofold speed, or.. ?

Best regards,

Gudni

On 14/02/2012, Miller, Mark <mmi...@sdsc.edu> wrote:
> Wayne Pfeiffer of SDSC implemented this for CIPRES. I am sure he can
> tell your folks how he set it up on beagle if you prefer to run it on
> your resources...
>
> Mark
>
>
>
> On Feb 13, 2012, at 8:44 PM, "Giap Nguyen"
> <nguyen...@gmail.com<mailto:nguyen...@gmail.com>> wrote:
>
> Good morning
>
> I think you may want to try the following tetragrid:
> <http://www.phylo.org/portal2/login!input.action>http://www.phylo.org/
> portal2/login!input.action
>
> Best regards
> Giap Nguyen
>
> 2012/2/13 Gudni Magnus Eiriksson

> <<mailto:gmeir...@gmail.com>gmeir...@gmail.com<mailto:gmeiriksso
> n...@gmail.com>>

-----
No virus found in this message.
Checked by AVG - www.avg.com
Version: 2012.0.1913 / Virus Database: 2112/4806 - Release Date: 02/12/12

[C.-j. Mei]

Well, my test run on your platform is not successful. I had two PhyloGeography computation XML input which I used in successful computations. I uploaded to your portal and started computation. Every thing is "Successful" but the computation terminated after a short period of time. I don't know what does it mean. There is no message to show any thing.

 

Help!

 

C.-J. Mei @ASU

[Gudni Magnus Eiriksson]

Thanks Mark

Large is a relative thing... the run I have in mind is about 2000cpu
hours I think (it is large to me, but I found out it is in fact
small). The NHPC people are taking long to reply to me on the issue of
running BEAST on their system. Is there any way to get access to your
resources, or are they in constant use? I am looking into other
alternatives as well.

-Best regards,

[Miller, Mark]

You can just visit www.phylo.org
Click through to the CIPRES science gateway,
Create an account (a few fields in a form) and you can start running.

Let me know (through the bug tracker links on the website) if you have any problems...

Best,
Mark

[Gudni Magnus Eiriksson]

Many thanks Mark
I will give it a try...
Best regards,
Gudni