Sunday, August 07, 2011
Columnar Databases and SQL Server Denali : Marathon towards being world's fastest analytical databaseI'm reading: Columnar Databases and SQL Server Denali : Marathon towards being world's fastest analytical databaseTweet this !
Have you ever heard of what are columnar databases? You might be wondering this is something new - The answer is No and Yes. Columnstore is not a new technology that has evolved suddenly and is making waves in the database community. It has been in the industry for quite some time. Generally database stored data in the form of records which resides in tables. The storage topology is typically known as rowstore as records are physically stored in a row based format. This methodology has its own advantages with OLTP systems and limitations with OLAP systems. The main advantages of columnstore are better compression, reduced IO during data access and effectively huge gain in data access speeds. Scaling data warehouse computing resources by scaling memory resources and using massively parallel processing does not fit with every business due to budgetary and architecture constraints. Columnstore seems to be a breakthrough technology to play the role of a catalyzer in analyzing enormous amount of data of the scale of billions of records, from enterprise data warehouses.
One of the best examples of columnar database success stories is ParAccel - one of the world's fastest analytical database vendors. Gartner in its latest report, has positioned ParAccel in visionaries category in the magic quadrant. You can get a deeper view on how ParAccel harnesses the power of columnar storage from it's datasheet and a success story.
Microsoft seems to have started its marathon in adding the nitro to SQL Server for adding data access speeds to DBs for OLAP engines. SQL Server Denali is introducing a new feature known as columnstore indexes, know as project Apollo and you can read more about this from here. This is just the first spark in the race of being one of the worlds fastest analytical database, a market into which IBM, GreenPlum, Kognitio, ParAccel and others have already plunged quite some time back. In-memory processing engine like Vertipaq combined with columnstore indexes can yield some blazing speeds in data warehousing environments. Time would tell what is the strategy of Microsoft to incorporate this concept in SQL Server and how SQL Server community reacts to it. Whatever be the case, it's a welcome news for end clients as of now.