Saturday, August 7, 2010

Key Highlights of Oracle Exadata

Oracle has been doing databases for more than three decades now. Over the course of these three decades, they have been busy in accumulating hundreds of years of experience from the industry. The culmination of their hardwork, innovation and vision seems to be the Exadata database machine.

Some of the key highlights of Exadata are:

Exadata is database servers plus storage servers.

Load is divided among both database and storage servers.

Storage in Exadata is no longer a dumb dump of bits. Processing is also done on the storage through the storage servers, so when the results of queries are returned to to the database servers, they are presented to the user along with caching, which provides added performance as the caches in the Exadata are huge and very fast.

Exadata's real beauty is that it exploits the hardware resources very smartly using world-class software. The Exadata software does column projection, predicate filtering and bloom filtering during joins to make the retrieval of data lightning fast. Here 'lightning fast' is not a marketing term, but I literally mean it.

Another blockbuster feature is the storage indexes. Yes you still need indexes even in Exadata, but they are perhaps the 'true' indexes as they fulfill the index definition completely. They make the data fetching fast and they don't take much space and their maintenance is not a headache. In Exadata storage, the space is divided into 1 megabyte region and storage indexes keep the low and high value of that region, and so index is aware of data distribution at such a minute level, and the parallel (helper) processes returns the data in less than jiffy.

Just like the mention of bride is incomplete without the bridal dress, the discussion of Exadata is incomplete without mentioning the Hybrid Columnar Compression. Compression is one of the most lovable weapon of DBAs who manage static data. Compression not only saves space at disk and buffer cache, it also improves the I/O. HCC defines compression units in the data and it groups rows into those compression units, and it only stores unique values of rows in the columns.

Intelligent, non-volatile, database-aware cache, the Flash Cache of Exadata works hand in hand with the Buffer Cache of database. Due to its huge size (up to 5TB), the flash cache often fulfills the needs of queries.

There are many many other working and proven features of Exadata. Larry Ellison is not excited without reason, he has really got a point there. If enterprises and corporations are individuals, then Exadata is like iPhone and iPad for them. It is cool, useful and expensive and they have to have it.

No comments: