IBM POWERs Scale Up With Its Newest Servers

By Matt Kimball - August 21, 2018

Note: This blog contains contributions from Moor Insights & Strategy principal analyst, Patrick Moorhead

IBM has had a string of success in recent quarters with its Systems business lines and last quarter, was the growth driver for the entire company. Part of that can be attributed to Z mainframes but was also driven by the continued success of POWER Systems. In the continued rollout of servers built on the POWER9 chip, it launched two scale-up servers – the E950 and the E980. How do these servers perform? Why does this matter to enterprise IT? And how can IBM continue to build on the positive momentum it has found in the market? What was launched As mentioned, IBM launched two servers today– the E950 and E980 aimed at the scale-up server market. These servers are targeting workloads such as in-memory databases, deep learning, and analytics (e.g., SAP HANA) that are in use by some of the largest enterprises (IBM boasts 80% of the Fortune 100 deploying POWER-based servers). Figure 1 IBM Scale Up Specifications. IBM When looking at these specifications overall – cores and memory and the large number of PCIe Gen4 slots – these are servers that seem to be tailor-made for those workloads that require maximum resources. IBM seems to have found a balance this go-around of all the necessary resources to make those high-end analytics deployments and deep learning instances run at peak performance. Key value propositions for the E950 and E980 The E950 and E980 deliver on performance, security, and reliability. IBM also touts these servers as driving “affordable scalability.” Performance, security, and reliability all make sense. In fact, these features are what IBM is known for. Affordable scalability? As an ex-IT Director, I’ve never thought of IBM as being synonymous with affordable. But, there’s logic to the claim. When considering the richness of resources (twice the number of threads per core v. the x86 equivalent), POWER9 can pack so much more computational power into fewer cores. Fewer cores to run some of the most demanding workloads can lead to a decrease in software licensing costs. As anybody who has spent time in a TCO model knows, decreasing software licensing costs can allow organizations to realize real cost savings. The E950 and E980 ship with POWERVM (IBM’s Power Systems Hypervisor) to enable auto-scaling of compute and memory resources. This can be important for multi-tenant environments that are elastic or “bursty” as resource requirements will be constantly changing. A quick refresher on POWER9 POWER9 is IBM’s processor and the foundational building block to its family of servers. This CPU is a four-way threaded chip that is designed to deliver outstanding performance-per-core (up to 2x that of its x86 competitor).  With the POWER architecture, IBM has focused heavily on compute and resource-intensive workloads such as business analytics, HPC, AI/ML and deep learning IBM has also built a strong cloud practice on the POWER architecture (read principal analyst Patrick Moorhead’s coverage of Google’s adoption of POWER here). How does POWER compare to x86 based CPUs from Intel and AMD? It’s a different architecture that focuses on absolute performance and the adoption of advanced features such as PCIe Gen4 and IBM’s own CAPI for I/O acceleration. This architecture is what makes IBM a strong candidate for those higher-end workloads. Scale-out v. scale-up and what this all means Scale-out designs are for distributed workloads such as cloud, database and virtualized infrastructure. Moor Insights & Strategy Principal Analyst Patrick Moorhead covered IBM’s launch of POWER9 and the momentum it is finding in the market. Scale-up server designs are for workloads that benefit from packing more processing power into a single platform. Think of a data analytics engine that needs to process terabytes worth of data in real time. This use case requires lots of computational resources (cores, memory, accelerators) located near the data. Scaling across multiple servers introduces latency that will lead to unacceptable levels of performance degradation. IBM’s design of POWER is ideal for both scale-out and scale-up server designs. And this is why servers based on POWER are finding momentum in verticals such as finance, oil & gas, healthcare. Critical success factors POWER is an architecture that has found success in the market. And based on IBM’s recent earnings, it appears POWER9 continues to build on that success. But IBM faces an interesting challenge in that its server lifecycles tend to extend longer than those of its competitors (yes, I’m saying they tend to build products that are too good and last a long time). Patrick Moorhead tells a story of an IBM POWER7 customer who was unhappy it had to reboot after six years. We are talking about some serious longevity here. There are a few things IBM can do to maximize success with the launch of its scale-up portfolio:
  • Think big. The launch of the E950 and E980 is not necessarily about these servers. The launch signals IBM completing its POWER9 portfolio and addressing the needs of enterprise customers – from mainframes to cloud to scale-up. IBM should promote its unique position and embrace the qualities its known for – reliability, security, and
  • Act small. While keeping the “big picture” in mind, selling scale-up deployments into Fortune100 customers is a high touch sales cycle. The teams deploying such systems will have an appreciation for the details around the performance, security and reliability features built into the POWER9 architecture. This audience will also dig deeper into IBM’s position on “affordable scalability.” IBM should be well prepared for this go-to-market approach.
  • Target the install base. As I previously stated, IBM has a unique challenge in that they deliver servers that are almost too reliable. Just writing this seems wrong, but it is, in fact, a hallmark of IBM and one of the reasons for its success in the datacenter. But there are reasons for customers to upgrade. Performance gains and security improvements that are relevant to every customer, and a TCO story that has legs. IBM should develop a campaign that clearly articulates the value of the E950 and E980 and ensure every customer has been educated.
  • Activate the channel. Engage with POWER channel partners and systems integrators. Focus on those partners that have practices built around target verticals. Arm them with the TCO models that show cost savings over current deployments. Give them the resources to prove performance advantages. And build in the incentives to motivate each channel partner to be proactive in its GTM approach.
  • Promote your success. There’s no selling tool more useful than case studies. Especially when those case studies are from an organization’s peers and competitors. POWER has a big footprint in Fortune100. IBM would be well served to continue promoting this fact with evidence-based customer stories.
Closing thoughts IBM is in a good position.  Its cloud and systems businesses have fueled healthy growth, and POWER-based servers are finding success across the industry. The emergence of workloads such as AI/ML and analytics are the kind of workloads and associated markets where IBM has traditionally found success.  And the launch of the E950 and E980 should help further strengthen IBM’s position. Note: blog contains content from Moor Insights & Strategy principal analyst, Patrick Moorhead
Matthew Kimball
+ posts

Matt Kimball is a Moor Insights & Strategy senior datacenter analyst covering servers and storage. Matt’s 25 plus years of real-world experience in high tech spans from hardware to software as a product manager, product marketer, engineer and enterprise IT practitioner.  This experience has led to a firm conviction that the success of an offering lies, of course, in a profitable, unique and targeted offering, but most importantly in the ability to position and communicate it effectively to the target audience.