Putting In Memory Analytics In Perspective
By Tobin Gilman on Jan 24, 2011
It seems like everywhere you look, vendors , analysts, and journalists are talking about in-memory analytics as if it is something new that will revolutionize business intelligence. It's become the new buzz word. I'm a little surprised by all the hype. In memory technology is certainly becoming more important in BI these days, and it can bring significant advantages for certain applications and deployment scenarios. But it is nothing new. Oracle has been applying creative and innovative ways to utilize the speed of memory in its BI platform for years. And it sure doesn't rank at the top of my list of emerging technologies that represent "the next big thing in BI". I'm a lot more excited about some of the other modern innovations in BI, like the convergence of analytics and business process management, real-time decisioning, mobile access, location intelligence, and collaboration. Those are game changers.
I don't mean to completely discount in-memory technology. I just think it needs to be put in perspective. It is definitely becoming more relevant in BI today for a couple of reasons. Memory prices keep dropping which makes it economically viable to increase capacity for in memory processing. And we all know that memory is a lot faster than disk. And of course data volumes continue to grow, which increases the need for faster processing. Super fast, more cost effective analytic processing is a good thing!
One thing that often gets lost in the in-memory hype is the fact that query and calculation performance are influenced by many variables, and the use of memory instead of disk is just one of them. Among the factors that ultimately impact system performance are warehouse data model design, query structure, CPU capacity, network bandwidth, complexity of analytic calculations, and user concurrency to name a few. So in evaluating the potential benefits of in memory technology, you need to consider it in context your entire BI topology and your analytic application requirements.
Evolutionary advances in how memory is utilized for faster analytic processing continues to be important and interesting. It's worth remembering though that in-memory technology represents just one element of overall system performance, and is but one of many, many exciting innovations in business intelligence. More on those innovations in a future post.