The Windows HPC Team Blog : Development

Dryad and DryadLINQ on HPC Server

JVert — Wed, 15 Jul 2009 23:04:00 GMT

Microsoft Research has an interesting project called Dryad which is investigating programming models for distributed data-parallel problems. If you've heard some of the hype around map/reduce programming or the Hadoop framework, you might be familiar with these type of problems. If you have gigabytes, terabytes, or even petabytes of data that you need to churn through, this is a powerful tool. On top of the Dryad framework is a programming model called DryadLinq which brings the power and familiarity of .NET's LINQ syntax to cluster computing. You can write your data analytics algorithm as a LINQ expression in your favorite .NET language and submit cluster jobs directly from Visual Studio.

I am delighted to announce that Microsoft Research has released a distribution of Dryad and DryadLINQ that runs on top of HPC Server 2008. You can download this release from here and contribute feedback and suggestions at the Microsoft Connect site here. If you want to learn more about Dryad and DryadLINQ, you can find more information on the Microsoft Research site. The source for DryadLINQ is also included in this release.

Releasing this has been a great collaboration between the HPC team and Microsoft Research. I am excited to get this work out there and see what people can do with it. So try it out and give us some feedback!

John Vert

Architect

Windows HPC Server

Update - Channel9 has a video interview with Erik Meijer and Roger Barga discussing this: http://channel9.msdn.com/posts/Charles/Expert-to-Expert-Erik-Roger-Barga-Introduction-to-Dryad-and-DryadLINQ/

a powershell problem

Barndawgie — Wed, 07 Jan 2009 04:22:00 GMT

We've gotten reports of an installation glitch that makes it so that node management cmdlets won't work when run from a remote client (they'll work just fine run on the headnode) due to a registry key that doesn't always get set up correctly. Fortunately, there is a simple work around! If you are trying to run PowerShell from a remote client and get an error like this:

PS> get-hpcnode
Get-HpcNode : Unable to find assembly 'Microsoft.Ccp.Sdm.Store, Version=2.0.0.0, Culture=neutral, PublicKeyToken=null'.
At line:1 char:12
+ get-hpcnode <<<<
    + CategoryInfo          : NotSpecified: (Microsoft.ComputeCluster.CCPPSH.GetNode:GetNode) [Get-HpcNode], Serializa
   tionException
    + FullyQualifiedErrorId : Microsoft.ComputeCluster.CCPPSH.GetNode

You should be able to remedy the problem by adding the following registry key: HKEY_LOCAL_MACHINE\SOFTWARE\Microsoft\HPC\BinDir" with a value of "C:\Program Files\Microsoft HPC Pack\Bin\". You can either do so manually, or just run the .REG file attached to this post.

Then, just open a new PowerShell window and you should be good to go! We hope to have a fix for this at some point soon.

Converting an iterative sequential application to data parallel using CMD scripts

ranjan — Thu, 05 Jun 2008 19:09:00 GMT

I have an application that is based on a sequential iterative call from a command script. It is a Monte Carlo simulation application is essentially easy to put in a data parallel format and deployed to run on CCS cluster easily.

Desciption of the sequential iterative version:

A main.exe program generates the n inputs for variation runs. The program has three inputs: Directory for results output, Project name, Number of variations
For example:
main Results2008 test_ccs 100

A DOS script (run_sub_DOS.bat) called by main execute the ‘distributed’ program n times iteratively. it calls the application "simulation" iteratively.

this is how the script looks:

@echo off

FOR /L %%i IN (1,1,%3) DO simulation %1 %2.%%i%

To modify this into a data parallel version I first changed this script to the following (call it run_sub_CCS.bat). This script, as mentioned above in the sequential case, is called by "main.exe".

set workdir=\\RRS-HN\lhuang\RRS-Example\Test1
FOR /F "usebackq tokens=4" %%j in (`job new /numprocessors:4-4 /jobname:RRS_TESTjob`) do set JobID=%%j
FOR /L %%i IN (1,1,%3) DO job add %JobID% /workdir:%workdir% /stdout:frf.out.%%i simulation %1 %2.%%i%

Then I had the following two wrap scripts to call this. That is I calle the scripts - CCS_Begin (which calls main) and CCS_Finalize in that order. These does some extra work of copying data to the CCS nodes. I decided to leave them as it is to point out the real environment.

@Rem CCS_Begin: script to run main

@call clusrun /readynodes rmdir \\localhost\C$\ranjan\RRS /Q /S

@call clusrun /readynodes mkdir \\localhost\C$\ranjan\RRS\Sub

@call clusrun /readynodes xcopy \\RRS-HN\ranjan\RRS\Sub\\localhost\C$\ranjan\RRS /Y /S

main Results2008 test_css 4

@FOR /F "usebackq tokens=1" %%k in (`job list /status:NotSubmitted`) do set JobID=%%k

job submit /ID:%JobID%

@Rem CCS_Finalize

FOR /F "usebackq tokens=1" %%k in (`job list /status:finished`) do set JobID=%%k
if %JobID%==%1 call clusrun /readynodes rmdir \\localhost\C$\ranjan\RRS /Q /S

(adopted from real application scenario).

Windows HPC Server 2008 April CTP Released

PhilPenn — Tue, 22 Apr 2008 16:33:00 GMT

Please see the latest Community Technology Preview (CTP) release of Windows HPC Server 2008 at the beta program community site (http://Connect.Microsoft.Com). If you aren't yet a registered beta program participant, then you'll first need to join the program using the same web site. This CTP release includes the HPC Pack which may be installed on any x64 edition of Windows Server (an evaluation version is also available for download) as well as the HPC Pack SDK (for developing HPC applications) and the HPC Pack OPK (for OEMs preparing HPC offerings). Be sure to provide feedback on the CTP features using the Feedback Tool on the Connect website. Also, please leverage the HPC MSDN Forums to obtain product support for ITPro's and Developers (http://forums.microsoft.com/WindowsHPC).

clusrun.vbs fix to handle name="value" parameters

ranjan — Fri, 07 Mar 2008 21:31:00 GMT

I want to install an application on a CCS cluster nodes using MSIEXEC. I use clusrun while have the package available on a share. i want the installation target to be at C:\Program Files\product name location on all nodes.

clusrun hangs when I give TARGETDIRECTORY="Program Files\product name" as a parameter for MSIEXEC. A simple workaround I used:

change the clusrun.vbs as follows -

(look at <CCP ROOT>\bin\ for clusrun.vbs).

you need to look at the part that processes supplied arguments in the commnd line (line 101 through 144). It should be as follows.

for i=0 to objArgs.Count - 1
if Left(objArgs(i),1) <> "/" then

          if schedname="" then
           set wshShell = Wscript.CreateObject ("WScript.Shell")
             set WshEnv = WshShell.Environment("Process")
             schedname = WshEnv("CCP_SCHEDULER")
             if schedname="" then
                 schedname = "localhost"
             end if
         end if

         '
         ' build the command line out of everything left
         '
         for j = i to objArgs.Count - 1
            'if the argument contains a space then add quotes to maintain the structure

          if InStr(objArgs(j), " ") = 0 then
             cmdline = cmdline + objArgs(j) + " "
           else
                cmdline = cmdline + Replace(objArgs(j),"=","=""") + """ "
           end if

         next
         task.CommandLine = cmdline
         exit for

end if

I could successfully use MSIEXEC with name value parameter and quotes as well as spaces.

Note: One way to overcome situations of having space in paramter is to place the command within a batch file and execute the batch file using clusrun or job submit.

Windows Communication Foundation (WCF) + HPC

PhilPenn — Thu, 28 Feb 2008 22:47:00 GMT

You'll want to review a recent MSDN forum entry regarding this feature now available with Windows HPC Server 2008 (currently in beta). Read more here.

Getting JOBID in a perl script

ranjan — Wed, 06 Feb 2008 17:54:00 GMT

The following article was the closest one I found when looking for scripting for CCS. I was told `new job ..` will return JOBID but not sure in what format!!- it is actually a string of 4 terms. Was not quite apparent from what was there on the example script which failed to run at the %CCS_JOBID% part (see URL below).

I came up with a hack to fix that - I think we can use scaler instead of the implicit way of coercing the string to number. So first I substr the number as string - then coerce it to number and use it! The hack I used (based on the script that uses %CCS_JOBID% at Look at the script posted on http://technet2.microsoft.com/windowsserver/en/library/30f7fbc9-af05-4bd4-97d5-aa9f97a3f6961033.mspx?mfr=true

# create new job
$id = `job new /numprocessors:4`;
# print $id;
$id = substr($id, 16,4) + 0;
# print $id;
# add parametric tasks
for ($i = 0; $i < 5 ; $i++) {
`job add $id /stdin:\\fileshare\infile$i.txt /stdout:outfile$i.txt myapp.exe`;
}

`job submit /ID:$id`;

Channel9 Video Presents New Job Scheduler OGF Web Service Interface

PhilPenn — Thu, 03 Jan 2008 01:30:00 GMT

View the screen-cast here: http://channel9.msdn.com/ShowPost.aspx?PostID=369487

Channel9 Video Presents the New Job Scheduler Features

PhilPenn — Tue, 18 Dec 2007 02:07:00 GMT

Watch the screen-cast at http://channel9.msdn.com/ShowPost.aspx?PostID=366694.

Channel9 Video about the New Network Direct RDMA Interface

PhilPenn — Mon, 10 Dec 2007 23:39:00 GMT

Check it out here:

http://channel9.msdn.com/ShowPost.aspx?PostID=366683

HPC Application Deployment Scenarios and Redistributable Dependencies

PhilPenn — Thu, 06 Dec 2007 10:43:00 GMT

Windows HPC Server 2008 Beta1 introduces 3 new programmatic interfaces to the Compute Cluster Scheduler. This effectively changes client application deployment scenarios along with client installation dependencies.

Please refer to my blog at http://blogs.msdn.com/philpenn/ for a more detailed commentary.