I ran across an interesting quote on www.perl.com that challenged my appreciation of relational database technology—and most particularly Oracle. The article states:
Relational databases started to get to be a big deal in the 1970’s, and they’re still a big deal today, which is a little peculiar, because they’re a 1960’s technology.
Forget the “started to get to be” bit for a moment. Uh, yes, RDBMS technology became important in the 70s (ISI, PRTV and Honeywell MRDS). However, since E.F. Codd didn’t write the defining paper until 1970 it is a stretch to call it “1960’s technology.” Oh well, that perl.com article was written in 1999 after all.
What a Great Idea: I Want to Force Some New, “Cool” Technology into My IT Shop
The bit in the quote that got me thinking was how astonished the author was that 1960’s—well, 1970’s actually—technology was “still a big deal” way back in 1999. You see, I think one thing that actually hurts IT shops is the nearly absurd rate of new technology injection. It seems to me that the datacenters with the highest level of success are those that take new technology as slowly as possible. Am I talking about newer versions of, say, the Oracle database server? No, absolutely not. Adopting a newer revision of Oracle is not radical. Those of us who revere the code rest soundly at night knowing that deep down in kernel are bits and pieces that haven’t really changed much in the last 20+ years—perhaps even 30 years given the KCB role in multi-block read consistency (if in fact MBRC was there in the first version of Oracle).
Why is there a mindset in IT that all old technology must go? Folks, we still use little round, brown spinning things (hard drives). Now there’s a bit of information technology that has been around longer than relational databases and ripe for a fond farewell. DBAs are asking for “disk space” from their storage administrators and that is exactly what they are getting. Forget for a moment that roughly 30 drives worth of large, sequential read throughput can saturate the bandwidth of most high-end FC SAN array controllers. Yes, there are exceptions, but I’m trying to make a point. The point is, here we are 27 years after the introduction of the ST506 5.25” and we are not getting full utilization of our drives—at least not when they are provisioned the way most space is provisioned these days. That is, you ask for 2TB of “space” for your Oracle database and you get it—allocated from something like 6 disks in the same enclosure as dozens of other disks. You are getting space, not bandwidth.