Windows Support


Simulation Cluster

Current Status of Sim Cluster Nodes

Last Updated: Sun Feb 12 10:15:34 PST 2012
sim.csde.washington.edu Network Load Balancer (NLB) is online

sim1 sim2 sim3 sim4 sim5 sim6 sim7 sim8
green green green green green green green green

Information about the usage of processing resources is included for each node. This is updated every 5 minutes. The table below lists the meaning of each status code:

  • green -- less than 50% CPU used
  • orange -- between 50% and 95% CPU used
  • red -- more than 95% CPU used
  • purple -- no data available

About the CSDE Sim Cluster

The purpose of the CSDE Simulation Cluster is to support CSDE's mission of education, research and scholarly exchange in population studies at the University of Washington. We focus on interdisciplinary research, with an emphasis on understanding the structural mechanisms that link individual behavior to population level outcomes.

This server cluster is dedicated to simulation research in the social sciences, especially demography. Students and faculty from all CSDE-affiliated departments are eligible for access.

To request access, please click here to fill out our online form.

CSDE would like to thank: the UW Student Technology Fee Committee for their funding of this project, Microsoft for their donation of server licenses, and Dell for their donation of equipment.

Technical Details of the CSDE Sim Cluster

The CSDE Sim Cluster is a group of eight Windows terminal servers featuring simulation-specific software.

Click here to view the list of software provided.

Each cluster node is a IBM Blade server with 16 2.53GHz Intel Xeon CPUs and 24.0GB of RAM. They run Windows Server 2003 (64-bit) Enterprise Edition. SIM is the "head node", where the Sim Cluster load balancing is controlled, and where the Z: scratch drive is stored. The remaining servers are "compute nodes", where cluster users run analyis jobs.

Terms of Use:

  • You will be expected to compose a brief note and endorsement for future funding proposals
  • You will load the cluster in such a way that so that others may also still use the resource
  • If running multiple jobs, you will start them on several nodes, rather than all on a single node
  • If you find problems with the SIM nodes or software installed therein, you will report them to csde_help@u.washington.edu so that we can keep the resource tuned for all users.
  • The SIM cluster will be rebooted once a month, automatically, at 3AM on the night of the last day of the month
  • If you want your disk IO to be fast, copy your data to the D: drive of the node where you are running your analysis.
  • Data on the Desktop or D: drive of the SIM cluster is not backed up. If you want data backed up to tape, put it in the H: or R: drive.
  • Both the Desktop and the D: drive will be wiped whenever the SIM nodes are re-imaged.
  • CSDE is not responsible for jobs or data lost due to unexpected system crashes or maintainence reboots, although we will we do everyhing in our power to minimize these disruptions.
The CSDE SIM cluster takes significant money and resources to run and there are many other folks using it.

As with anything good in this world, the future of this resource depends on your participation and being a good neighbor.

Downloads

Download the latest version of Remote Desktop Connection for Windows: msrdpcli.exe

Read the system administrator's "Usage and Tips" document: Sim-Cluster-Tips.pdf

A visual diagram of the Simulation Cluster Network is available here: Sim-Cluster-Network.pdf

How to Connect to the Sim Cluster

Simply connect to sim.csde.washington.edu using the Remote Desktop Connection client and log in with your CSDE username and password. You'll be automatically redirected to the least busy Sim Cluster node. (The cluster uses a network load balancing system, or NLB.) This establishes your session. Now, with access to the files stored in your CSDE H: drive, you can start your simulation job.

To start a new session for an additional simulation job, simply open another Remote Desktop Connection window to sim.csde.washington.edu and you'll get a second session. Most likely, the new session will be on a different cluster node, but the current state of the Sim Cluster as a whole will determine this.

You can even disconnect while your job is running, and when you log back on to sim.csde.washington.edu you'll be connected to your old session, regardless of which node you were on.

Where Am I?

The My Computer icon on your session's desktop will indicate which cluster node you're using:

ncosgray on SIM3

If you disconnect from your session, you can either reconnect to it by logging back in to sim.csde.washington.edu, or you can connect directly to the node where it's running. For example, I am on SIM3 in the screenshot above; if I were to disconnect, my session would keep running on SIM3. To reconnect and check the status of my jobs, I'd use Remote Desktop Connection to connect to sim.csde.washington.edu or sim3.csde.washington.edu.

The Scratch Drive

The Z: drive, or NetScratch, is a shared network folder stored on SIM which is available to all Sim Cluster nodes via the cluster's high-speed private network.

Feel free to create a folder in Z: and save temporary intermediate files there, but please remember to clean up after yourself when the simulation has completed. If the scratch drives become cluttered it will be necessary to delete their contents; this may occur at any time without warning.

About Backups on the Sim Cluster

No files on the simulation cluster servers will be backed up. If you need to keep copies of programs, data or results in a safe place, use your H: drive.

How to Connect to a Specific Node

We recommend that you always connect to sim.csde.washington.edu so that all cluster users can benefit from the load balancing features of the Sim Cluster with the minimum amount of effort. Doing this also allows you to easily manage your existing sessions. However, you might want to log in to a specific node if you need to reconnect to a disconnected session, or if the network load balancer fails to connect you to an idle server. To achieve this, you can use Remote Desktop Connection to connect to the target node's fully-qualified domain name, e.g. sim3.csde.washington.edu.