Center for Computational Molecular Science and Technology Georgia Institute of Technology Center for Computational Molecular Science and Technology School of Chemistry and Biochemistry

CCMST Weekly News, February 18 2011

February 18, 2011 3:02 pm EST

1. Announcements
2. Statistics
3. Tip of the Week

ANNOUNCEMENTS

Ggate Open for Production

The new CCMST cluster ggate, aka greengate, is now open for production. The system is not in full configuration, as the GPU nodes have not been delivered yet. For the moment 16 compute nodes are available, for a total of 448 cores and lots of memory and disk space. Any CCMST user who currently has an account on fgate can now login to ggate (ggate.chemistry.gatech.edu) using the fgate credentials. More information about the system is available at the CCMST wiki page.

NOTE: Planned Power Outage Postponed

Please note that the power outage for the MS&E building, originally planned for March 26-27 2001 has been postponed until later in the spring semester. Date to be determined.

STATISTICS

FGATE

Uptime: 35 days
/home directory usage: 75% (1.5 TB available)
/backups directory usage: 83%

LSF usage for Week 6 (2/7-2/13) (times are in minutes)
GroupJobsTotal CPUAvg CPUAvg WaitAvg Trnr.
Bredas 328 160656 8% 490 2327 2966
Hernandez 543 858028 44% 1580 61 1903
Sherrill 692 344721 18% 498 2925 3430
Total 1563 1363405 70% 872 1805 2802

Note: percentages refer to the total CPU time available for the period.

Most productive user of the Week: atucker 769501.

EGATE

Uptime: 115 days
/theoryfs/common directory usage: 41% (395GB available)
/theoryfs/ccmst directory usage: 90% (88 GB available)

LSF usage for Week 6 (2/7-2/13) (times are in minutes)
GroupJobsTotal CPUAvg CPUAvg WaitAvg Trnr.
Bredas 7 1269 0% 181 0 184
Hernandez 1043 581104 38% 557 303 924
Sherrill 1949 218190 14% 112 237 347
Other 119 358 0% 3 0 4
Total 3118 800919 53% 257 250 527

Note: percentages refer to the total CPU time available for the period.

Most productive user of the Week: hagy 266791.

TIP OF THE WEEK

By Massimo

Working on Ggate: a Primer on Scyld Clusterware

Scyld Clusterware is the management software running on ggate.This is a proprietary software developed and distributed by Penguin Computing ( ww.penguincomputing.com).

Here are the main features of Scyld Clusterware:

  • The master node is the central point of control for the entire cluster.
  • Compute nodes appear as attached processor and memory resources.
  • Compute nodes boot from the network and run only a light kernel.
  • Libraries and drivers are provided to compute nodes by the master node on-demand. Scyld ClusterWare provides a unified process space across all nodes in the cluster.
  • All processes in the cluster are visible and can be directly controlled from the master node. Remote execution services on compute nodes are not required.
  • The cluster has the look-and-feel of an SMP system.

Some commands for monitoring and managing processes:

  • The top and ps commands show all processes running in the cluster.
  • kill will kill a process anywhere in the cluster.
  • To see in which node a process is running, pipe the output of ps through bpstat –P: ps –ef | grep malagoli | bpstat -P
  • beostatus –c will show the load status of each compute node.

Using the bpsh command to interact with compute nodes:

  • bpsh executes a command on a particular node.
  • bpsh n5 ls /scratch/malagoli shows the contents of the scratch directory of node 5.
  • bpsh -a ls /scratch/malagoli shows the contents of all the local scratch directories (on all nodes).
  • bpsh -ap ls /scratch/malagoli prepends the node id.
  • bpsh n12 rm -rf /scratch/malagoli/badjob deletes the scratch files of a bad job on node n12.

Do you have usage tips that you want to share with the other CCMST users? Please send them to Massimo (massimo.malagoli@chemistry.gatech.edu) for inclusion in the Tip of the Week section.