Monday, August 27, 2007

Technology Performance Increases - Why Disk Sucks

In a previous life I looked after the FreeBSD port of Firebird. I still check in occasionally to see what the Firebird community is up to.

I remember Jim Starkey announcing his employment by MySQL to write them a new storage engine (codename Falcon). I thought I'd check out the progress of this engine.

One of the articles I came across was Mike Kruckenberg's blog. One of his entries was on Jim Starkey's presentation at the 2006 MySQL Conference.

The bit that caught my eye was the point list for the technology changes that have occurred over the 30-odd years. The stand-out for me was the reduction in disk access speeds - only a mere threefold reduction. Everything else has undergone changes involving an order (or orders) of magnitude.

This speaks volumes to me. The so-called innovation in disk storage has only really been in capacity. We've only really taken the bit of rust, put it on glass or metal and spun it faster. The last ten years has been about how much rust we can put on the platter and how we align it.

I know his is a cynical comment and undermines the engineering and design feats that have occurred, but let's face it, the most common failure points in a computer system nowadays are the mechanical ones.

I'm looking forward to real innovation in storage - that is, breaking away from the traditional rust-on-platter approach and moving to hopefully a non-mechanical process.

No comments: