Recently, IBM Redbooks has published a new Redpaper, SAS Business Analytics Deployment on IBM POWER8 Processor-Based Systems: With IBM XIV Storage System and IBM FlashSystem. This ITSO publication describes the performance tuning guidelines that can help you successfully deploy SAS business analytics on IBM POWER8® processor-based servers. This paper also describes the deployment of SAS analytics on IBM POWER8 processor-based servers with IBM XIV and IBM FlashSystem™ in a hybrid-storage model, along with configuration and tuning guidelines. Here are 5 Things to Know about SAS Business Analytics on IBM POWER8 Processors:
- Differences between the IBM POWER8 and the POWER7. POWER8 supports up to 12 cores per socket compared to 8 cores per socket on the IBM POWER7® processor. The POWER8 processor-based servers have increased core density and compute power within the same form factor. POWER8 also supports SMT8 mode, which increases the maximum amount of execution work per cycle. SMT8 mode with eight hardware threads can improve the performance of workloads that use parallelism by allowing additional instructions to run at the same time.
- SAS and IBM - For more than 30 years, SAS (Statistical Analysis System) and IBM have collaborated to offer highly optimized solutions to more than 2,400 customers worldwide. This enables our mutual customers to best exploit the advanced technology of IBM InfoSphere Warehouse, powered by DB2. With a tighter integration between IBM Data Warehouse software and systems and the SAS technology, our joint customers will be able to reduce the time required to build and deploy vital business models.
SAS analytics deployment architecture on IBM Power S822 server with XIV Storage System and Spectrum Scale
- SAS analytics deployment on IBM Power servers with XIV storage. IBM Power S822 (8284-22A) is a scale-out server with two sockets, 20 cores, and 256 GB memory. The server is configured with a Virtual I/O Server (VIOS) and five client logical partitions (LPARs). The VIOS helps in sharing the Fibre Channel (FC) adapters among the client LPARs by virtualizing the physical FC adapters using N_Port ID Virtualization (NPIV). The test configuration uses a single VIOS, however, the preferences is to use dual-VIOS in production deployments for high availability.
SAS analytics deployment on IBM Power S822 server in hybrid-storage environment
- SAS analytics deployment on IBM Power servers in a hybrid-storage environment
SAS workloads typically have two sets of file systems. While SASDATA stores persistent data that includes input and output files, the SAS work area (SASWORK and SASUTIL file systems) stores temporary data that is deleted at the end of each SAS job. In typical customer deployments, SASDATA requires much more space for storing input files and output files. Hence, it is beneficial to deploy SASDATA file system on an XIV Storage System (which is disk-based storage) and the SAS work area on a FlashSystem.
One of the LPARs is used for testing SAS analytics. Volumes are created on XIV and FlashSystem and are mapped to the LPARs. The SAS workload Spectrum Scale file systems are created on the mapped volumes for running SAS analytics jobs. The file system layout on a hybrid-storage environment is shown in Figure 6. A SASDATA file system is created with 16 volumes mapped from XIV Gen3 storage. The 16 volumes proved to be optimal for the workload used during the testing. The 32 volumes were mapped for SASWORK and SASUTIL, each from FlashSystem 840.
- IBM POWER8 processor-based Power S222 server, IBM XIV Gen3, and IBM Spectrum Scale (formerly IBM GPFS) together form a platform that is well-suited for running even the most demanding SAS business analytics workloads. The IBM Power S822 server performed extremely well for SAS workloads that support the compute and I/O requirements.
Also, IBM POWER8 processor-based servers performed extremely well for SAS workloads in hybrid-storage environments involving IBM FlashSystem. SAS analytics customers can explore the possibility of using hybrid-storage architecture involving XIV or other disk-based storage systems and IBM FlashSystem to accelerate SAS analytics while keeping the overall TCO low.
James A. Pistilli
Please consult the following useful related resources:
IBM Power Systems Information Center
IBM Power Systems at IBM PartnerWorld
SAS on IBM AIX 5L, AIX 6, and AIX 7 Tuning Guides
IBM Power Systems scale-out servers
Implementing IBM FlashSystem 840, SG24-8189
Accelerate insights with SAS Business Analytics and IBM FlashSystem
SAS 9.3 grid deployment on IBM Power servers with IBM XIV Storage System and IBM GPFS
Understanding Processor Utilization on Power Systems – AIX