Forum OpenACS Q&A: Future relational databases performances?

What are some potentially affordable technologies coming up in the
near future (5 to 30 years in the future) promising a big performance
boost in terms of transaction and query (without giving up on
safety). Any research paper reading geeks here *smile*? Will skills
like good schema designing and tuning ever go out of style? Or less
critical?
Collapse
Posted by Janne Blomqvist on
Sometimes I read research papers, yes. Although quantum mechanics adapts quite poorly to transaction processing theory.. :) Anyway, predicting the future is hard, especially in a fast moving field such as computer technology. As a general rule, short term predictions are generally overoptimistic while predictions far into the future will be very underestimating. So how will the relational database world look in 2005? My guess is that Oracle will still be the standard everything else is compared to. PostgreSQL finally has outer joins. The MSSQL engine will be a multi-gigabyte monster with a nice GUI and essentially the same core features as the present versions have. 2030? No idea... Generally, I don't see any dramatic improvements in transaction processing theory. Transaction processing inevitably needs lots of small accesses to mass storage. Of course hardware will get faster and faster, sometime we will inevitably move to some sort of solid state mass storage, which will improve speeds a lot. One perhaps interresting development is the kdb database (see www.kx.com) which differs from most RDMS:s in that it stores data column-wise instead of row-wise. For some types of queries this is of course faster, but I think most 'normal' applications won't improve a lot by that. Some people think relational databases will be replaced by object databases. I don't think so. As an interesting sidenote, the mysql folks aren't the only ones not groking transactions. See http://www.microsoft.com/sql/productinfo/transadvantage.htm for Microsoft:s excuse why SQL Server doesn't have a multiple versioning concurrency control system (MVCC). And in the process they make a fool of themselves. *sigh*