Technical Specification of School Cluster (10.107.225.254)

Master Node with 128GB RAM and 20 Cores

11 nos. of computes nodes with 128 GB RAM, and 20 cores each node

1 GPU node with 128 GB RAM , 20 cores and K40 GPU card

2 SMP servers with 1TB RAM and 64 cores each. One SMP server is out of queue (Compute 5) and one is in the queue.


Queue structure of Cluster
  1. Serial Queue: (Short queue) having 80 cores, walltime 24 hours and 128 GB RAM
  2. Parallel Queue: (Long queue) having 140 cores, walltime 96 hours and 128 GB Ram
  3. Gputest1 Queue: having 20  CPU cores and K40 GPU card
  4. SMP Queue: having 64 cores and 1 TB RAM

Fare-share policy (auto managed by scheduler) is active in background. Statistics flushing interval for fair share policy is 2 months.

 


    Quota Limits for storage

  1. Ph.D.                                2 TB
  2. PG Diploma                      1 TB
  3. M.Sc (2nd year)            200 GB
  4. M.Sc (1st year)             100 GB

 


   Basic Commands for Cluster

  1.  Submit the job to the scheduler :    qsub <script_name>
  2.  Check job status :   qstat , qstat -n1
  3.  Check full information of the job : qstat -f <job_id>
  4. Delete the job from the queue : qdel <job_id>
  5. Check the queue information : qstat -Q
  6.  Check jobs in queue : showq
  7.  Check queue parameters : qmgr  -c ‘p s’

 


FAQ's regarding cluster

 

10.107.225.254

11

128 GB RAM, and 20 cores

1

128 GB RAM, 20 cores and K40 GPU card

2

1TB RAM and 64 cores

4 nos.

Serial Queue : (Short queue) having 80 cores, walltime 24 hours and 128 GB RAM • Parallel Queue: (Long queue) having 140 cores, walltime 96 hours and 128 GB Ram • Gputest1 Queue: having 20 CPU cores with K40 card • SMP Queue having 64 cores and 1 TB RAM

With the approval of Supervisor/Dean the cluster account is created.

No

Storage is assigned initially as per below details: • Ph.D. 2 TB • PGDE 1 TB • M.Sc (2nd year) 200 GB • M.Sc (1st year) 100 GB

With the recommendation of Supervisor and further approval from the Dean SCIS, the storage limit may be extended with proper justification.

Fare-share policy (auto managed by scheduler) is active in background. Statistics flushing interval for fair share policy is 2 months.

2 Months

qsub <script_name>

qstat , qstat -n1

qstat -f <job_id>

qdel <job_id>

qstat -Q

showq

qmgr -c ‘p s’