IBM Readies Query Appliance for Mainframes; What About i?
August 17, 2009 Timothy Prickett Morgan
As part of its Smart Analytics System announcement from a few weeks ago, Big Blue also talked a little bit–and I mean only a little bit–about another kind of business analytics hardware that it had cooking in its labs and factories and that would debut soon for its System z mainframes. And hopefully, this appliance, called the Smart Analytics Optimizer, will also soon be available for the Power Systems platform running i 6.1 (and AIX and Linux, if you care).
The Smart Analytics System is a stack of preconfigured Power 550 systems running AIX, parallel file systems, and InfoSphere and Cognos data warehouse tools (and soon, predictive analytics from SPSS, which IBM is in the process of acquiring for $1.2 billion). The Smart Analytics Optimizer is an appliance that links to a System z mainframe through a normal Ethernet link and is used to process very complex queries–the kinds that would normally bring production systems to their knees as they are doing their daily batch and transaction processing work–on the appliance but using live, production data. So you do the kinds of work you would normally do in a data warehouse, but you don’t have to do the extract and load operation from production to data warehouse systems and you do a query on the real, live data–just not on the production machine. In a sense, it is a query co-processor.
Exactly what is inside the Smart Analytics Optimizer is not clear, but I have heard that it makes use of SQL processing technologies that have been tweaked to run on vector processors and utilizing in-memory databases that hold a snapshot of the production data housed on mainframes. The word from IBM is that for certain kinds of queries, the Smart Analytics Optimizer runs a complex query 50 times faster than the mainframe and that the price for a complex query on this appliance will be two orders of magnitude cheaper than running it on the mainframe. I confirmed that the DB2 queries are not running on so-called specialty engines inside the mainframe, and I also think it would be very interesting to cluster or daisy chain such devices together in a manner that allows complex queries to be run all the time without having to resort to a data warehouse at all.
Steve Mills, general manager of IBM’s Software Group, has not confirmed that the Smart Analytics Optimizer would be available for IBM’s other server platforms, but it seems highly unlikely that Big Blue won’t make it available for Power Systems running AIX and i and their respective variants of DB2. All of the same issues about performance and complex queries plague AIX and i shops as do mainframe shops–problems that spawned the big data warehousing wave back in the mid-1990s in the first place.
IBM turns back on server history: To give and to hybrid (The Register)