Job Limits

Each LC platform is a shared resource. Users are expected to adhere to the following usage policies to ensure that the resources can be effectively and productively used by everyone. You can view the policies on a system itself by running:

news job.lim.MACHINENAME

Web Version of RZAlastor

There are two login nodes, and 25 pdebug (interactive) nodes (no batch nodes).  Each node has 2 sockets, each of which has 10 2.8GHz Intel Xeon cores for a total of 20 cores and 64 GB memory per node. Turbo mode (dynamic overclocking) has been enabled so users could achieve up to 3GHz.

Banks are not used on this partition.  Jobs are scheduled per core. There is 1 scheduling pool:

  •  pdebug—500 cores (25 nodes)

Hardware

Nodes are Intel Xeon Ivy Bridge with 2 sockets, each with 10 (2.8 GHz) cores for a total of 20 cores & 64 GB memory per node.

Scheduling

RZalastor jobs are scheduled using SLURM. Scheduling is not technically enforced so users are expected to monitor their own behavior and keep themselves within the current limits while following the policies written below:

  •  Users will not compile on the login nodes during daytime hours
  •  A user can have a maximum of 60 processors with a runtime of up to
  •  4 hours in the queue during the day with the following exception:
    •   An occasional job for debugging that takes 61-80 processors for a maximum of one hour can be run as long as it is the user's only job in the queue.
  • The queue (sorted by user) can be found by typing "squeue -S u" at the prompt after setting the environment variable:
    SQUEUE_FORMAT='%.7i %.9P %.8j %.8u %.2t %.10M %.6D %.4C %R'
  • Daytime is  0800-2000 Mondays-Fridays not including holidays
  • No production / processing runs allowed, only development and debugging
  • Users will avoid computationally intensive work on the login node
  • We are all family and expect developers to play nice.  However if someone's job(s) have taken over the machine
    • Call them or send them an email
    • Call Ines Heinz at 3-7900 and she will call them and/or kill the job
    • Call Ines's backup (Ellen Tarwater) and she will get the job killed
  • This approach will be revisited later and additional limits will be set if necessary. If someone  monopolizes the machine, developers can always shift to other RZ resources

Scratch Disk Space: Consult RZ File Systems Web Page:
             https://rzlc.llnl.gov/fsstatus/fsstatus.cgi
 

Zone
RZ
Vendor
Appro
User-Available Nodes
Login Nodes*
2
Debug Nodes
32
Total Nodes
36
CPUs
CPU Architecture
Intel Xeon E5-2680 v2
Cores/Node
20
Total Cores
720
Memory Total (GB)
2,304
CPU Memory/Node (GB)
64
Peak Performance
Peak TFLOPS (CPUs)
16.1
Clock Speed (GHz)
2.8
Peak single CPU memory bandwidth (GB/s)
60
OS
TOSS 4
Interconnect
IB QDR
Parallel job type
multiple jobs per node OR multiple nodes per job
Recommended location for parallel file space
Program
ASC
Compilers
Documentation