Category Archives: Exadata

Revisiting SAAS BI in the context of Oracle’s Exadata Storage Server

A while back I wrote about SAAS BI and why I thought it may not be the best strategy to avail BI as a SAAS. A few commenters didn’t agree with me.

Guess what? Today, at Open World, Larry Ellison, the guy who would have better understanding of where the databases are going within the next decade, more than most of us, has unveiled one of the coolest products of the decade in the BI space. No, it’s not yet another cool reporting tool or a sexy graph renderer or a user friendly pivoting or some meta data layer that allows analysts create ad-hoc queries without learning about outer joins and SQL aggregations. The innovation is at the very core layer, the hardware layer.

When you hear Larry talk about TPC-H, the challenges that large companies are facing today with tera-bytes of data, mind-you, inspite of a big database grid, you wonder how the SAAS BI guys can solve the problem of extracting those tera-bytes of data into their own hosted environments.

Exadata Storage Server is essentially a specialized storage hardware that has database level knowledge (think of it as a hard disk with part of the database logic put together) and so, when a query has to be executed, instead of having to return blocks of data from the storage server, only the results set is returned back. This is a big big big thing. If you have a large table and your query requires a full-table scan on it, in the traditional approach, those thousands of data-blocks have to be returned (even with optimizations such as multi-block fetch for FTS). But in the new approach of specialized storage solution with built-in database logic, it just needs to return the summary data. That’s a lot of bandwidth optimization between the storage grid and the database grid. Apparently, currently the pipes between the Exadata storage server and the database grid have so much high-bandwidth that they are no longer the bottleneck. Seems, now instead the hard-drives are the bottleneck! Well, if only the SSDs (Solid-State Disks) mature and become more price effective, then perhaps that equation would shift. But, atleast for now, Oracle’s Exadata server blows the competition away by a huge margin. Forget about SAAS, we are talking about raw processing power here!

The benchmarks shown in the presentation indicate that this solution performed much much better than the solutions like Teradata and Netezza.

Leave a comment

Filed under Exadata, Oracle, SAAS BI