Most, if not all SAP customers find that HANA’s in-memory architecture provides the opportunity to eliminate aggregate tables and OLAP cube structures… HANA is so quick to aggregate data on-demand, that building a fragile and complex hierarchy of derived data and a series of derivation job-streams is no longer required.
Oracle recently announced Exalytics as an add-on to Exadata where derived data can reside in-memory to improve OLAP query performance. Oracle has also just upgraded their Exadata offering with a significant injection of Flash memory in-between the disks and the processors. Flash memory is sort-of the poor-man’s in-memory database alternative. They say that “hot” data will then reside on Flash silicon, queries will avoid slow disk I/O, and be screaming fast.
HANA set the bar for “screaming fast” by defining “fast” as eliminating the requirement for pre-aggregated data. “Fast” is fast enough to aggregate on-demand.
So one question is: can Exadata eliminate the requirement for pre-aggregated data? If so, then Exalytics is useless. If Exadata is not quite good enough to eliminate the aggregates but it is as fast as Oracle claims… then the pre-aggregated cubes could reside in Exadata Flash rather than on an expensive add-on server… and again, Exalytics is useless. The only other possibility is that Exalytics was designed by Oracle to solve a real problem… that Exalytics is not useless… but that Exadata is not as good as the hype. There is a disconnect somewhere over there…
VN:F [1.9.22_1171]Exalytics vs. Exadata,