The potential of SAP's HANA in-memory computing engine permeated the agenda and had attendees' tongues wagging this week at the Tech Ed conference in Las Vegas, but the technology still has some maturing to do.
SAP has relentlessly promoted HANA since its launch in May 2010, touting its ability to quickly analyze large amounts of business data and reporting major early interest from customers. Over time, SAP plans to weave HANA throughout its portfolio as part of a sweeping "renewal," but in reality, the product only went into general availability in June, and detailed stories of customers successfully using it in production have so far been scarce.
One person who has worked on a number of HANA proof-of-concept (POC) exercises said Tuesday that while HANA's technology works well, customers should expect some bumps in the road and a fair amount of effort before they get the results.
"The most important thing is, don't do [a POC] just to test the technology," said Harald Reiter of Deloitte Consulting, during a presentation on the Tech Ed show floor that drew a sizable throng of onlookers. Customers should start a HANA project with a solid use case in mind, "something you really want to go live with," that justifies the financial investment, he said.
Right now, he said, customers have limited choices: Either use one of the few packaged analytic applications SAP has so far released for HANA, or build something from scratch. "Those are the hard ones."
Customers should take pains to line up a full-fledged, dedicated HANA team, according to Reiter. The roster should include a project manager, solution architect, HANA administrator, data services architect, data modeler, security expert and developers, he said.
Part of the challenges Reiter faced on his projects was the fact that SAP has been issuing updates to HANA in rapid-fire fashion.
While the patches often were to fix bugs and are "technically painless to implement," they still require testing to be redone, he said. In addition, at times the updated version of HANA behaved differently from the last.
Data modeling is the key to squeezing the best performance out of HANA, so the data modeler member of the team is most important, Reiter said. HANA's performance can be "totally different from one data model to the next," he added.
SAP has partnered with a number of hardware vendors for HANA, which is available as an appliance in a number of sizes with system RAM varying accordingly.
Customers should realize that a considerable portion of system RAM is going to be consumed by processing, leaving less room for data storage, he said. Companies will want to use HANA to run historical analyses on all of their data but that will be difficult with a smaller appliance, even when HANA's data compression capability is factored in, he said.
Sign up for Computerworld eNewsletters.