Friday, 10 April 2009
Had a strange experience yesterday - went to a project group meeting on network resilience - and really enjoyed it! It was interesting, fun and I understood most of it - should I be worried?
We've done a lot of work to make the core network resilient - two data centres, doubling up of routers, virtual routers at the core, duplicate network connections etc.** Now we're moving out towards the edge and looking at how we provide services to buildings. Currently these are through a small number of aggregation centres located across the campus which provide a great single point of failure. Especially the one that has been for years situated in a basement which frequently floods. The UPSs on them are not particularly reliable and certainly not environmentally friendly. So. we've just begun a major project to get rid of them, and provide duplicate gigabit connections to all buildings, directly linked to the data centres without the need to go through these aggregation centres. Sounds simple doesn't it? But - and there are lots of big buts - we're talking miles of cables, ducts, digging up roads (or using other peoples ducts such as sewers), disruption, expense (the price of kit has risen drastically recently with move of the dollar against the pound). It will be a long job, with some interesting prioritisation along the way.
The first task is to survey what we've got, and where our cables actually go - it will make the London Underground Map look like child's play.
** Our Data Network Manager wrote a really good article for our internal newsletter on how this was done - will see if he'll let me reproduce it here.