The Joy of Benchmarking

Aplologies for the LBF (low blogging frequency) – I am currently preparing my “End of Year” report and interview, which is why I am fairly busy writing TONS of other stuff and haven’t had much time for this blog.

I worked on a paper about a novel benchmarking approach for OWL reasoners, which is a highly interesting topic. Basically, we have OWL ontologies, we have reasoners, and we would like to find out how the reasoners work with the ontologies in terms of performance and correctness. Naturally, there are some issues that need to be addressed in order to obtain coherent results: what kind of test sets do we use? How do we make sure that what we measure is what we really want to measure? And how can we avoid interference with other processes running on the computer?

A lot of these questions are answered in this great article on Java benchmarking, which I looked at quite a lot when implementing my benchmarking framework. It addresses the problems of measuring very small times with Java, including issues like a warm-up phase for the JVM, how to deal with garbage collection, and some basic info on statistics as well. So if you’re planning on doing some performance benchmarking, make sure you’re on the safe side in terms of stabilising the measurements and interpreting the results correctly. You should always (ALWAYS) know exactly what you measure. And that still doesn’t mean that your results will be reliable / make sense / are what you expected – trust me, I spent a lot of time staring at spreadsheets with a lot of numbers in them 🙂

Ah well. It’s good to know that I’m not the only one who’s baffled by the randomness that is inherent in computers, as this article on lego robots shows:

From technology to food: Another lunch type place I can recommend is Umami on Oxford Road. While the first time I went to this Japanese/Chinese place was rather disappointing (particularly nasty looking chicken in curry sauce for my dining companion), I decided to try it again in my lunch break one day – and keep returning since. The £5 lunch offer for a starter and a main is good value if you’re starving after a morning of groundbreaking research, the food is tasty and they even do slight modifications of the lunch dishes – great for vegetarianising (ahem) those fried noodles you’ve been craving for days.