BookmarkSubscribeRSS Feed
Lalith_db
Calcite | Level 5

we are using SAS 9.4 M6 on Linux RedHat 6 Platform. 

 

Our Environment is very robust, 70 TB Lasr Memory with 70 nodes, each node have 40 cpus

number of nodes: 70

CPU(s) per each node: 40
Thread(s) per core: 2
Core(s) per socket: 10
Socket(s): 2

 

being these huge resources available, we are not seeing better performance in LASR.

 

My Question is : is LASR Server process requests in Parallel or Sequential 

 

Please find below screen shoot from LASR logs, each request is processing, not like multiple at a time.

 

Lalith_db_1-1620141503008.png

 

 

 

 

2 REPLIES 2
JuanS_OCS
Amethyst | Level 16

Nice resources indeed @Lalith_db .

Performance is quite a broad topic. In which specific aspect are you not seeing better performance: loading data, rendering reports, anything else? What is the change that you made to make better performance? Are you sure the performance decrease comes from the LASR service? It could be from network, as an example.

 

Furthermore, the logs you are showing us are from a single node, and a single process. 

I assume you already gave a look and work on a daily basis with the SAS HPA guide https://documentation.sas.com/doc/en/hpaicg/3.9/hpaicgwhatsnew.htm , are you aware and work with the gridmon.sh tool? It is one of the best to check how your system is performing, in terms of LASR and TKGrid.

JuanS_OCS
Amethyst | Level 16

In addition to above, I would say distributed SAS systems work in general (but not necessarily) better, when the distributed system is large, but not every node is too large. 8 or 16 physical core nodes are really optimal, and then it is important to keep a determined CPU/RAM ratio. It would seem better more nodes with less size than less nodes of larger sizes, in order to work with more data optimally.

 

In any case, I would say your best contact now is SAS Technical Support or your SAS representative, in order to take this discussion to the next level. There is no one but SAS itself who can give you the best recommendations when you wnt to achieve best performance.

suga badge.PNGThe SAS Users Group for Administrators (SUGA) is open to all SAS administrators and architects who install, update, manage or maintain a SAS deployment. 

Join SUGA 

Discussion stats
  • 2 replies
  • 1302 views
  • 1 like
  • 2 in conversation