The realities of “uptime” for Internet sites is discussed by Steve Levin in an article that was linked on Slashdot, so you’ve probably already seen it. BB agrees wholeheartedly, while being a little confused by references to NOC teams and the huge cast of support personnel seemingly on hand for all web sites - surely all those jobs are done by one entity, aren’t they? Usually us.

Talking about defining uptime goals reminds us to avoid products that guarantee 99.99…% uptime. In a previous existence on a lower plane of being, BB worked for a company selling HA software - we won’t mention their name to avoid embarrassment (our’s). Initially, the product promised the standard four-nines of availability. Then one day, the person editing the sales collateral - who didn’t possess a great technical understanding of the subtleties (i.e. marketing) - thought, “Weeeell, why split hairs? Let’s just round the figure up!” So this HA product will now give you total uptime all the time, presumably by never ever failing over (which causes outage). That begs the question of why you’d need it in the first place, if your infrastructure is so reliable it never fails over anyway.

Actually, you don’t need it. No one needs that kind of confused trash talk near their production systems.