Difference: ClusterNodes (6 vs. 7)

Revision 72018-06-19 - DavidPane

Line: 1 to 1
 
META TOPICPARENT name="WebHome"

Cluster Nodes

All of the nodes on the cluster have hyper-threading disabled. This means that each core can only work on one thread at a time.

Things to consider about performance:

Using all of the processors on a machine to their maximum capacity will certainly bring down the machine.

You must leave resources for the OS and overhead.

The processes/threads may also be:

  • CPU bound (needs lots of CPU resources)
  • Memory bound (needs lots of RAM resources)
  • I/O bound (Network and/or hard drive resources)
All of these three resources are finite and any one can limit the performance of a system. Users should consider which of these their particular program will be consuming. There will be a point where the performance of your multi-threaded processing could actually slow down with more threads. It could actually get significantly slower or even crash the node.

Rack #1 (Protected with UPS)
Changed:
<
<
Legacy Name from HawkPsych-o Name# of CPUsCores Per CPUTotal Memory (GB)Queue
dsi-compute-0-8 compute-0-39-2 12 6 24.7 dsi
dsi-compute-0-7 compute-0-39-1 12 6 24.7 dsi
cnbc-compute-0-4 compute-0-37-2 12 6 99.1 default
cnbc-compute-0-3 compute-0-37-1 12 6 49.5 default
cnbc-compute-0-2 compute-0-35-2 12 6 99.2 default
cnbc-compute-0-0 compute-0-35-1 12 6 49.5 default
cnbc-compute-0-6 compute-0-33-2 12 6 99.1 default
cnbc-compute-0-5 compute-0-33-1 12 6 99.1 default
>
>
Legacy Name from HawkPsych-o Name# of CPUsCores Per CPUTotal Memory (GB)Queue
dsi-compute-0-8 compute-0-39-2 12 6 128 dsi
dsi-compute-0-7 compute-0-39-1 12 6 128 dsi
cnbc-compute-0-4 compute-0-37-2 12 6 96 default
cnbc-compute-0-3 compute-0-37-1 12 6 48 default
cnbc-compute-0-2 compute-0-35-2 12 6 96 default
cnbc-compute-0-0 compute-0-35-1 12 6 48 default
cnbc-compute-0-6 compute-0-33-2 12 6 96 default
cnbc-compute-0-5 compute-0-33-1 12 6 96 default
  /data3

/replicated3

/data2

/replicated

Changed:
<
<
compute-0-16 16 8 65.9 default
compute-0-15 16 8 65.9 default
compute-0-14 16 8 65.9 default
compute-0-13 16 8 65.9 default
compute-0-12 16 8 65.9 default
compute-0-11 16 8 65.9 default
compute-0-10 16 8 65.9 default
psycho-o.hpc1.cs.cmu.edu 16 8 24.7
UPS For Rack #1
UPS For Rack #1
>
>
compute-0-16 16 8 48 default
compute-0-15 16 8 48 default
compute-0-14 16 8 48 default
compute-0-13 16 8 48 default
compute-0-12 16 8 48 default
compute-0-11 16 8 48 default
compute-0-10 16 8 48 default
psycho-o.hpc1.cs.cmu.edu 16 8 48
UPS For Rack #1
UPS For Rack #1
 
Rack #2 (The equipment in this Rack is NOT Protected with UPS)
Changed:
<
<
Psych-o Name # of CPUs GPU CardsCores Per CPUTotal Memory (GB)Queue
empty
empty
empty
empty
empty
empty
empty
empty
empty
empty
compute-0-21-2 12 6 99.1 default
compute-0-21-1 12 6 24.7 plaut
compute-0-19-2 12 6 24.7 plaut
compute-0-19-1 12 6 49.5 plaut
compute-0-17-2 12 6 49.5 plaut
compute-0-17-1 12 6 49.5 plaut
compute-0-15-2 12 6 24.7 plaut
compute-0-15-1 12 6 24.7 plaut
empty
empty
empty
empty
empty
compute-1-5 24 4 6 128 taising
compute-1-3 24 4 6 128 taising
compute-1-1 24 4 6 128 taising
>
>
Psych-o Name # of CPUs GPU CardsCores Per CPUTotal Memory (GB)Queue
empty
empty
empty
empty
empty
empty
empty
empty
empty
empty
compute-1-21-2 12 0 6 96 default
compute-1-21-1 12 0 6 24 plaut
compute-1-19-2 12 0 6 24 plaut
compute-1-19-1 12 0 6 24 plaut
compute-1-17-2 12 0 6 32 plaut
compute-1-17-1 12 0 6 48 plaut
compute-1-15-2 12 0 6 24 plaut
compute-1-15-1 12 0 6 24 plaut
empty
compute-1-13 16 4 8 128 tarrq
compute-1-11 16 4 8 128 taising
compute-1-9 16 4 8 128 taising
compute-1-7 16 4 8 128 taising
compute-1-5

12

4 6 128 taising
compute-1-3 12 4 6 128 taising
compute-1-1 12 4 6 128 taising
 
This site is powered by the TWiki collaboration platform Powered by PerlCopyright © 2008-2019 by the contributing authors. All material on this collaboration platform is the property of the contributing authors.
Ideas, requests, problems regarding TWiki? Send feedback