More on Maglev
After a little more searching around, I found some good pieces which bring back to reality the performance claims of Maglev and potential problems with object oriented databases. Let’s not forget that Maglev isn’t even out yet, and we have no idea if it’ll be what it claims to be. In the mean time, there are other well known Ruby intrepreters and VMs, such as Ruby MRI, JRuby and YARV.
Check out these in depth looks at Maglev:
This is by Charles Nutter, contributor to a popular Ruby intrepreter written in Java called JRuby. In it, he calls Maglev vaporware, thus Gemstone’s Maglev performance claims don’t mean much at this point. The key to his point is that we have no idea if Maglev is fully Ruby compliant, and to prove his point, he shows us actual numbers for performance gains that can be achieved in JRuby by removing some compliance features.
Sho Fukamachi questions the Maglev performance claims and the usefulness of OODBs.
My Thoughts
My point in the last post was that mapping from OO languages to relational DBs can get ugly. Naturally then, I’m curious about other ways of looking at the problem, but other solutions could have problems, too.
Design patterns such as Ruby’s ActiveRecord have popped up to abstract relational systems to an OO style. However, abstracting relational DBs to an OO model can often result in more DB queries, and it’s easy for inexperienced programmers to write bad, query heavy code. They just see everything as objects, and they don’t understand the consequences of their manipulations.
For this reason, it’s no mystery to me why people often complain about Ruby’s scalability. It isn’t necessarily Ruby that’s the problem. It might be the programmers. Perhaps an OODB is the solution since it would remove the need to map and reduce the aforementioned consequences, but I really don’t know.