Jim Goodnight wrapped up the SAS Executive Global Forum. He began by talking about the changing hardware environment for analytics – the explosion of multi-core chips in particular – and how this has driven a massive re-design in SAS as they take all the SAS products and make them support multi-threaded, multi-core set ups. Two particular products that came from this work were the high performance analytic products for EMC Greenplum and Teradata (discussed earlier in my announcements post).
Customers have traditionally sampled their data as data volumes get large. Of course as more data is available the need to sample to make things work in a sensible time becomes more widespread. And yet customers with this data want to use larger samples when they can. SAS would like to let them run their analytics on all the data because working out the right sampling approach is one of the big causes of delay in getting models built and deployed. This is what the new MPP-based solutions do – make it possible to process literally a billion rows using one of these appliances, spreading the load across all the cores and using massively multi-threaded SAS routines.
This, in turn, means that the kind of massive portfolio analysis that companies used to think of as end-of-month analysis can become part of a transaction – doing portfolio impact analysis as part of each operational decision that affects the portfolio. Or it allows retailers to run mark-down optimization every day.