Efficient Data Analytics Using SAS 9.4 and HPE GreenLake for File Storage

By Anshul Nagori, senior WW technical marketing engineer, Hewlett Packard Enterprise Development LP (HPE)

In the data-driven world today, enterprises deploy AI, ML, and data analytics applications to gain a competitive edge and make real-time decisions to drive their business.

These applications require super-fast data processing capabilities and utilize huge amounts of storage. HPE GreenLake for File Storage offers that – a highly performant, reliable, and scalable file storage solution. In this blog, I’m focusing on how it runs efficient data analytics using SAS 9.4 application.

The latest report from The Forrester Wave marks SAS as a strong leader in the AI decisioning platforms. SAS requires an infrastructure stack that complements its in-memory processing capabilities and offers better performance than ever before. This translates to a hardware and software configuration that delivers data analytics quickly and accurately. In addition to faster compute, it demands reliable and fast storage.  

GreenLake for File Storage standard density is an all-NVMe, disaggregated, scale-out file storage solution. It offers fast, sustained, and predictable performance for data intensive applications like SAS. HPE has built this storage offering in collaboration with Vast Data Software. It uses the Alletra Storage MP modular hardware platform offering a flexible, disaggregated, scale-out, shared-everything architecture. This means you can independently scale performance as well as compute based on your needs. From a management perspective, GreenLake for File Storage benefits from the awesomeness of GreenLake cloud. It enables streamlined deployment, super-easy file share creation, and a simple self-service cloud experience.

We’ve been testing SAS 9.4 Mixed Analytics workload with the GreenLake for File Storage and the results are stellar. Let’s take a look at the data reduction achieved during this testing. But first, let me take you through our test environment.

Figure 1: SAS 9.4 on HPE GreenLake for File Storage setup

As shown in Figure 1, we used 6 HPE ProLiant DL 360 Gen 10 servers with a total of 208 physical cores as test clients. Each client had 512GB of memory and RedHat Linux Enterprise Linux 9.3 as the OS. These clients were connected to the GreenLake for File Storage system over multiple 100GbE ports. The file shares on GreenLake for File Storage were presented and mounted to the SAS clients using NFS over TCP with nConnect set to 8.

Analytics workloads are generally measured in terms of concurrent number of users and/or jobs. This test began with a set of 30 users running 102 concurrent analytics jobs. And we steadily increased the number of users and jobs in each run, from 30 all the way up to 180. Correspondingly, the number of jobs ranged from 102 to 612. Figure 2 shows the combined read + write throughput obtained on each run. The maximum throughput for 180 users we observed was 15GB/s for jobs of up to 5s duration. The throughput reduces as the average job duration increases for the same number of users. This performance is quite decent for SAS customers using high performing shared storage in their Grid environments.

Figure 2: Combined read + write throughput on each run
Click to enlarge

As you can observe, this throughput doesn’t increase much from 180 concurrent users to 360 users. This means that the GreenLake for File Storage standard density 2×2 configuration performs best for up to 180 concurrent SAS 9.4 users or 612 concurrent jobs.

GreenLake for File Storage – ideal for AI and analytics
It is a modern, disaggregated, scale-out solution offering intuitive cloud experience, simplified management, efficient space optimization for several real-world use cases – especially AI and analytics.

Resource :
Technical paper describing this solution and the performance characterization in detail: SAS 9.4 on HPE GreenLake for File Storage – Run efficient data analytics with SAS 9.4 on a high-performing, disaggregated, scale-out file storage