SAP HANA Now Available On IBM Power
May 6, 2015 Alex Woodie
After years of discussion and months of testing, SAP‘s HANA in-memory database is now officially supported on IBM‘s Power servers, giving joint SAP-IBM customers an alternative to running HANA on X86. IBM on Monday officially launched two special editions of its Power Systems servers designed to run HANA Business Warehouse software this week, while an announcement from SAP is expected soon.
“I don’t think it’s a big shock to anybody that we’re entering the HANA space and that we’re doing it with a processor that was designed for big data and analytics,” says Vicente Moranta, IBM’s director and business line executive of SAP on Power. “It’s a natural step for us.”
Any Power8-based server with the right version of SuSE Linux (SLES 11 SP3 with some PTFs) can now load and run a HANA database, but just for analytic workloads; no transactional workloads are supported yet. The way Moranta sees it, this will benefit not only IBM’s contingent of Power Linux customers, but those SAP customers running core ERP systems on IBM i and AIX operating systems.
“IBM i and AIX clients who run SAP or run other analytics stacks are now able to continue to use that Power platform,” he says. “The market wants choice, and clients who haven’t considered Power in the past now have a choice. Especially in the HANA world, the clients who have wanted to deploy that applications have always had to look at Intel. Now there’s a clear differentiation, a clear conversation, and we believe that’s very valuable.”
IBM launched two Solution Editions specially configured to run HANA. The smaller system, called S824, features 24 Power8 cores, and from 512 GB to 2 TB of memory, which enables it to run HANA databases that are up to 512 GB in size after they are compressed. This system does not support virtualization, and is optimized to run HANA workloads exclusively.
The second system, the E870, features from 40 to 80 Power8 cores, and from 1 to 8 TB of memory; it’s designed to run compressed HANA databases that are up to 1 TB in size. This server also supports virtualization, enabling customers to run other non-transactional workloads on it. Neither of the servers comes with much internal disk; they’re designed to store data on SAN arrays that have been certified via SAP’s Tailored Datacenter Integration (TDI) program.
IBM has more Solution Editions on tap, including a four-socket machine scheduled to be unveiled in mid-May. The companies are also working to expand the amount of server memory that HANA can utilize. Currently, HANA on Power is limited to utilizing 32 GB of memory for each Power8 core.
IBM will also work with SAP to support other HANA workloads besides Business Warehouse. “It’s a portfolio that we’re starting to build to handle SAP on Power,” Moranta says. “What you’re seeing is both companies embracing what clients have been asking for. The clients that already have Power are interested in using that same platform for HANA.”
While IBM declined to provide pricing information for the two Solution Editions now available, IT Jungle was assured that pricing would be very much in line with existing Intel servers. “We’re going to be really, really competitive,” Moranta says.
Similarly, we must take IBM’s word that the benchmark tests that IBM used to see how HANA on Power stacks up against the X86 competition are compelling. “We have done benchmarks. We have them. They look really good,” Moranta says. “They’re looking so good that we’re trying to decide what to do with them. We know most customers don’t make purchase decision based on benchmarks. We want to figure out how we’re going to approach that.”
One thing that seems certain is that HANA will never run under IBM i. “It’s not something that is likely to happen,” Moranta says. “It would add a whole lot of complexity to the supportability. It’s not something we can say we’re going to do. But if a client is interested in having a conversion, again we can share the roadmap under NDA and discuss what we have coming and see how we can adapt our roadmap and our current plan to that requirement.”
For that matter, it doesn’t seem that AIX is any better positioned to get HANA workloads in the future, either. SAP’s strategy for HANA is to support it on SLES. While there hasn’t been a huge rush to install and use Linux partitions in the IBM i installed base, it would be a stretch to call Linux a foreign body at this point. Linux is established in enough IBM i accounts to prevent it from being a major roadblock toward HANA-on-Power adoption.
Compared to the additional cost (we assume) and complexity (we know) of running HANA on one or more Lintel boxes, running HANA in a Power Linux partition on the same machine running SAP’s IBM i-based ERP system would seem to be a win-win for everybody involved. Kudos for IBM and SAP for actually making this a reality.