Virtualization and Analytics: Recurring Power8 Themes

June 2, 2014 Alex Woodie

If you struggled with virtualization with your Power7 box, you might want to improve your skills in that regard with Power8, which brings an even more heavy reliance on virtualization. On the plus side, the massive compute and I/O capacities of Power8 open up new possibilities in the analytics world. For IBM i shops, that means taking seriously the potential to run new analytics workloads right next to your transactional loads.

Analytics and virtualization were the highlights of a recent conversation on Power8 with Tim Klubertanz, a senior IBM Power Systems architect at Evolving Solutions, and formerly of IBM‘s Lab Services group in Rochester, Minnesota. Evolving Solutions, based in Minneapolis, was the first IBM business partner in the world to take delivery on a new Power8 server earlier this month. The 4U scale-out server will provide a test bed for Evolving Solutions customers who are considering adopting the new server.

If you’re thinking about the Power8, you’d better bone up on your virtualization skills, says Klubertanz, who works with AIX, Linux, and IBM i customers. In particular, knowing how to use the Virtual I/O Server (VIOS) will be even more important with the Power8 machines, owing to the fewer number of I/O adapters available.

VIOS started becoming important in Power7 when customers struggled to feed the processors with data from disk, Klubertanz says. “When Power7 came along the number of cores was so high that users can’t physically attach enough I/O resources to machines to do what IBM i customers are accustomed to doing,” he tells IT Jungle, “so virtualization becomes really the only way they can get into a machine economically and really make use of the hardware and what’s it capable of doing.”

Although the I/O adapters in Power8 provide plenty of bandwidth for moving data between the processors and storage, virtualization has become even more critical with Power8. “It’s probably more of a requirement for Power8 than Power7. There are more threads per cores, more cores per socket, and right now there are no I/O expansion enclosures for Power8,” he says. “So if you want to run Power8, unless [your workload] can fit within the small number of slots in the system unit, you have to virtualize.”

VIOS, which has its roots in AIX, struck fear into the hearts of IBM i devotees when it was introduced years ago. But customers should work on replacing those fears with knowledge, and not be afraid to learn something new–especially when that knowledge has become so essential to efficiently operating an IBM i-Power Systems environment.

Evolving Solutions showing off its new Power8 system.

“There’s been a reluctance to use VIOS because it’s based on AIX. That’s completely different for the iSeries administrator,” Klubertanz says. “I would say that most customers overestimate some of that involvement. There is quite a lot of automaton and graphical interfaces to the VIOS and the virtualization, so they can do a lot of the management from the HMC now without having to log in and run AIX commands.”

Just as logical partitioning took some getting used to when it was introduced so long ago, VIOS is finding some resistance in the IBM i installed base. But shops are starting to warm up to VIOS, especially since IBM has started supporting the Storwize storage arrays there. The Storwize products are proving hot sellers for IBM and its channel; in particular, the midrange Storwize V5000 is finding a lot of buyers among the IBM i faithful, Klubertanz says.

While the Power8 rollout is expected to boost the channels’ fortunes in the months ahead, storage in general has been boosting bottom lines for resellers like Evolving Solutions. The unrelenting explosion in all types of data–structured and unstructured alike–and the immense interest in data analytics is driving customers to stockpile data.

“Customers are looking for the competitive advantage that an analytics solution might give them,” Klubertanz says. “They’re not even sure what advantage that is yet. They just know they need to be looking at it to stay ahead of their competition.”

Interest in Hadoop has generated its share of hoopla over the past year or so, and the promoters of real-time streaming analytic and in-memory data grid platforms have had time in the sun, too. There’s a huge buzz in the field of “big data” in general, and IBM and its resellers are certainly tuned into the profit potential here.

While IBM i is left out of the big data discussion, Linux is right in the thick of it. And since the Power Systems platform runs Linux, IBM is positioning the platform to take on commodity X86 servers for the privilege to run these new workloads.

The good news for IBM i shops is that they’ve already invested in a platform that software vendors are targeting to run big data analytic workloads. These workloads won’t run on IBM i, which has its strengths firmly in the transactional camp. But by installing and running Linux on their Power Systems server, they can take advantage of the innovation occurring in big data analytics without having to buy and manage more X86 servers.

While IBM i shops continue to consolidate their ERP, DB2 for i, and other transactional workloads, they can also take advantage of the considerable processing overhead in the Power8 servers to consolidate analytic workloads on fewer machines. “The same box is able to handle both workloads and compete at a price point for the analytics type workloads IBM is shooting for,” Klubertanz says.

Midsize companies, in particular, are more likely to leverage Power8’s virtualization capabilities to cohabitate analytics and transactional workloads on the same machine. Running a data mart next to a transactional system helps to minimize duplication of data, but the bigger firms are more likely to use ETL to offload that data to secondary system where the analytic work is done, he says.

In any event, interest in analytics is surging in companies of all sizes. IBM is smartly positioning Power8 as a good investment for scale-out, distributed processing of analytics workloads. For IBM i shops that just want to do transaction processing, there’s more than enough oomph in Power8 to get the job done. And for those shops that are more adventurous, Linux and big data analytics await on the other side of the hypervisor.