Blog: Krish Krishnan

« Historical Data - Costly Maintenance | Main | Where Is My Data? »

Is the SMP architecture suited for the Data Warehouse Ecosystem

Whenever a debate comes up amongst RDBMS vendors on who is the fastest database on which platform, you are directed to TPC results and how each one can surpass the other in controlled tests. Well when we talk transaction processing that is one thing, but if we talk about the data warehouse, this takes a whole different meaning.

I do not mean to belittle the TPC and their benchmarks, being a DBA myself for a number of years. But when we talk of the data warehouse ecosystem, we are not talking about small discrete transactions, we are talking about mixed queries that need processing power from the underlying database, storage and network.

The current stack of database, storage and network platform, with all the power and processing capabilities have yet not solved the issue of query speed and sustained performance. Here is where the end users are left frustated and IT often helpless, since the underlying platform cannot cope with the expanding demands, which arise from the fact that the underlying architecture of these traditional solutions are SMP based.

The net result is more spending all over to ensure sustained speed and performance while paying for database and code redesign and deployment. Does this cycle ever slow down lest stop, is there an alternative?.

We do have an answer when we talk of less spend for more performance. Before you go on your next splurge, take a look at the Data Warehouse Appliance. If you need answers ask the hard questions. This is one compiled stack of database, storage and network all built for performance and sustained scalability and most importantly based on MPP architecture.

There are multiple vendors in the market with Data Warehouse Appliances\ offerings, each of them have a solution that can be applied to solve specific problems. Yes they are all up and coming technologies, but rewind the clock and so were the database, storage and network vendors in the early years.

No early adopter of this technology has claimed failure so far, that itself is a testimonial to the fact that the Appliance technology is here to stay and help solve the problems that cannot be solved by the SMP stack. It is not just being MPP that matters, but building the right solution offering at the right price is where the Appliances are making the inroads.

Whether we choose to look at this technology or choose to ignore it for all reasons, it is becoming clear that SMP architecture cannot fuel the data warehouse ecosystem for a long time and will need the MPP architecture to co-exist and provide the combined platfom.

Share: