March 3, 2017

Should bioinformatics be constrained by IT?

When planning bioinformatics capacity it is easy, especially for short-term projects, to let the parameters of design be constrained by what it is possible to achieve with current IT hardware. Optimal price points on disk storage capacity vs. IO speed, processor cores vs. clock speeds, memory capacity vs. cost-per-gigabyte, and in particular maximum transfer speeds and bandwidth contention on network connections are all looked at in terms of today's limitations when making plans to build tomorrow's systems.

Why do we do that? Why not plan to incorporate tomorrow's IT into tomorrow's bioinformatics, instead of shackling systems that are deployed 2-3 years hence (I'm talking about big projects here, not the small scripting jobs you run off in an average afternoon) to the constraints of IT long passed that quite likely will no longer apply. 

We plan ahead for sequencing technology - there isn't anyone in the industry that isn't keeping a close eye on Oxford Nanopore to see how they might have to design their systems to cope with this new type of data from the first day that it comes online - so why don't we plan ahead for IT?

Advances in IT hardware are constantly pressing on. Moore's law is well known in relation to processor speeds, but it applies to other areas too. In 3 years time, when the project you're planning today finally makes it into the daylight, Moore's law suggests that the hardware available to you will be at least twice as capable as that of today, either in terms of raw processor or IO speed or how many gigabytes of storage or memory can be bought for the same money. So why not plan ahead and say that although your proposed system is impossible and unworkable on today's hardware, by the time it goes into production the systems in place at that stage will be more than adequate to host and support its requirements?

Seems logical to me - except, of course, that nobody has a perfect crystal ball, and if you are unfortunate enough to have planned ahead for an eventuality that didn't actually happen in the end, you'll be left with an unworkable system.

So, as always with the big things in life, make sure you have a plan B - just in case!

Topics: Bioinformatics