SAS Grid Computing Technical FAQs

1. Does the installation and configuration of SAS Grid Manager require a plan file?

Ans : Yes. There are sample plan files for Data Integration, Enterprise Miner or Foundation SAS grid installations, or a custom plan file can be used.

2. What are the main uses for running SAS on a grid?

Ans : There are three major functions of a SAS grid, all of which make use of intelligent resource-based load balancing of the SAS workload:

 - Workload balancing enables multiple users in a SAS environment to distribute workloads to a shared pool of resources. Better performance is achieved by allowing the grid middleware to handle allocation to jobs to the appropriate machines. This provides greater scalability & manageability for environments with numerous SAS users and SAS jobs.

 - Accelerate processing for a long running SAS job. Allows users to distribute parallelized SAS workloads to a shared pool of resources to automatically find and use the best available resource. The ability to have pieces of work run in parallel on different nodes in the grid allows the over all job to complete much more quickly than if all the pieces were run in sequence.

 - Scheduling jobs to a grid enables users to distribute jobs within workflows to range of hosts & automatically find and use the best available resource for each job.

3. What products are needed on the grid nodes?

Ans : SAS Grid Manager must be licensed for the number of nodes in the grid. SAS Grid Manager includes the Platform Suite for SAS. LSF, which is part of the Platform Suite for SAS, needs to be installed on all the nodes in the grid. Base SAS and SAS/CONNECT are also required, as well any SAS products required by the execution that is being distributed. For example, if you are running SAS/STAT in a grid environment, you would need SAS/STAT on any of the grid nodes where the SAS/STAT programs can run.

4. What SAS products and solutions take advantage of a SAS Grid automatically?

Ans : SAS Data Integration Server: Data Integration Studio includes a Loop Transform which allows portions of a job flow to be submitted to multiple processors either on an SMP box or on a grid.

 SAS Enterprise Miner: Allows parallel execution of independent nodes either on an SMP box or on a grid. Several other SAS products can take advantage of a grid with some minimal customization.

 Ideally any SAS program (with proper modifications) including stored processes can be deployed in a grid environment

5. If SAS is installed in a 4way hard partition within a 16way machine in a grid, will SAS Grid Manager be able to limit the execution of SAS to just those 4 CPUs on that node?

Ans : SAS would only execute on the 4 CPUs in the partition - not because of SAS Grid Manager but because the partitioning makes it appear as a 4-way machine. There is no way for SAS to see the other 12 CPUs.

6. Are Sun Solaris "containers" supported in a grid?

Ans : Yes – in SAS 9.1.3 support Solaris containers requires a hot fix for the Platform software – contact SAS Tech Support to obtain the hotfix. A large SMP box would be much more expensive that several smaller boxes – so using Solaris containers would make sense if the customer already has the box and wants to carve part of it out for the grid and use other containers for other apps.

7. Can grid computing be done with SAS Grid Manager via Foundation SAS without a full Business Intelligence (BI) implementation?

Ans : Yes. SAS Grid Manager does require the SAS Metadata Server and use of a plan file during installation and configuration, however SAS BI Server or SAS Enterprise BI Server are not required. You do have to define a SAS Grid Logical Server in SAS Management Console in order for the SAS application to “see” the grid. You must also add the appropriate grdsvc_enable calls to your SAS program unless the program is generated from a grid-automated solution such as Enterprise Miner or Data Integration Server. Any SAS program that can be segmented into sub-tasks of execution and/or data can be run in a grid, and/or multiple SAS programs can be submitted to the grid for load balancing.

8. How does the SAS Grid Manager environment compare to using the MP CONNECT feature of SAS/CONNECT?

Ans : The capabilities available with SAS Grid Manager are an enhancement to the parallel processing capabilities of MP CONNECT. We have integrated SAS/CONNECT with the grid expertise of Platform Computing in order to bring dynamic resource based load balancing and better monitoring capabilities SAS/CONNECT is required along with SAS Grid Manager.

9. Does SAS Grid Computing work with Risk Dimensions?

Ans : If you are interested in running Risk Dimensions on a grid, the consultants in SAS Professional Services and the RiskAdvisory group have tools to enable this.

10. Can Windows 32 bit servers and Windows 64 bit servers be combined in one grid?

Ans : Yes this can be done. Since 32 bit and 64 bit data sets are stored in different formats CEDA will be used any time these servers access data created by the other and there will be translation done on the fly. There could be a performance implication depending on how often this happens.