Last week, Microsoft announced its latest TPC-H benchmark score, which was quite impressive. Microsoft’s latest score--posted on May 21, 2007--was 60,359 queries per hour (QphH) @3t with a Price/QphH of $32.60 compared with Oracle's best score--posted on May 14, 2007--of 37,813 QphH @3t with a Price/QphH of $38.00. Almost double the performance for a lower cost per query? Sign me up!
A few years ago, many people questioned Microsoft’s ability to handle enterprise-class performance needs. Today, I find that most people, even Microsoft naysayers, are willing to accept that Windows and SQL Server solutions can be scaled to meet their needs. Microsoft naysayers now focus on other reasons for not liking Microsoft. (No, I’m not going to explore that topic this week, but I promise to revisit the subject in the near future.) Microsoft's score should reassure SQL Server customers who might have lingering doubts about SQL Server’s ability to handle their performance needs.
Impressive TPC benchmarks are important and interesting, but realistically most customers don't need the amount of horsepower displayed in these tests, and the customers that do need that much horsepower have such specialized needs that a generic TPC test is probably of little relevance to them. However, relational database management system (RDBMS) vendors love to play the TPC leapfrog game. Now that Microsoft has trumped Oracle's TPC score only a week after Oracle posted its record breaking number, I can only wonder how long it will take Oracle to try to leapfrog Microsoft’s score.
The inner child in me can’t help but get excited about the big and impressive performance database tuning numbers that vendors put in front of me. I’m especially susceptible because I’ve earned my living as an RDBMS performance tuning geek for most of my professional life. I suspect that anyone tempted to daydream about Panasonic’s 103" plasma TV trumping Samsung’s previous "world's biggest TV" that was 102" might also be susceptible. However, the more mature IT professional lurking within me then rears it’s ugly head and wonders when vendors will realize that not all customers want or need server environments that cost the better part of $2 million and instead provide benchmarks that make sense for the rest of us. Perhaps this won’t happen in my lifetime, but I hope and pray for the sake of future generations of DBAs that someday benchmarks will be meaningful and relevant again. Until then, I’ll be at Best Buy lustfully staring at giant TV’s. For more information about the recent scores and TPC-H benchmarks, go to http://www.tpc.org/tpch/results/tpch_last_ten_results.asp or http://www.tpc.org/tpch/results/tpch_result_detail.asp?id=107052101 .