DB Rambles

Forrester Report on SQL Server 2005 Scalability

An article on E-Week about a Forrester report on high performance computing sparked some thoughts. The report claims that Microsoft’s planned 2005 release of SQL Server will have few, if any, performance improvements in the high-end performance and scalability arena.

I guess I was a little surprised at the findings. If Microsoft has taken 5 years — let’s face it, an eon in the IT world — to produce the next major release of their flagship database server, I would have expected some pretty substantial performance improvements. And I think most Microsoft customers — especially the ones who’ve been waiting this long for SQL Server 2005 — will be pretty peeved if the report’s findings are true.

But then again, is this more a case of media hype than anything else? When you look closely at the parameters on which “high-performance scalability” is based, you tend to wonder if these types of reports really will impact Microsoft’s bottom line, or whether they serve merely to weaken Microsoft’s reputation amongst the database elite. I mean, after all, SQL Server has a devout following because of it’s ease of use, administrative toolset, and MS product integration ability; its devotees aren’t necessarily loyal because of a one-time TPC-C benchmark showing Microsoft could “compete with the big boys”.

The generally accepted parameters for being judged a high-end database installation is a database of at least a terabyte, with thousands of concurrent users. Let’s face it: not many companies are running databases of this size or breadth. The article estimates that only around 2000 of these high-end databases (of which 80 run SQL Server) exist, so, do these reports really impact Microsoft’s reputation in the database arena? Is the database market, or rather the revenue of database software vendors, reliant on these 2000 companies? Unlikely. A significant chunk of revenue is derived not from license sales, but from support and applications integration, which is why you’ll see disclaimers on Oracle and Microsoft’s pricing pages that “license costs account for a small proportion of the database’s overall total cost of ownership”.

So, what do these types of reports really tell us? And who do they aim to influence or inform? I admit there must be some sort of yardstick by which our industry measures the performance of database software, but are the TPC-C benchmarks and these types of reports really applicable to the average IT manager who is looking to purchase database software?

How can the industry develop more realistic or focused measurements which target small to medium-sized businesses and provide information that is relevant to them? As always, much interested to see what people have to say!