Scheduled Downtime
It was 3am on Sunday night and, for whatever reason, I wanted to see if my landlord had cashed a recent rent check. I pulled up my online banking with Citizen's bank and was greeted by this far-too-common message:
Online banking is down for scheduled maintenance.
Ryan S and Eric V, co-posters on this blog, happened to be present for this and each of them suggested we check their respective online banks: TD Bank North and United bank. They were down for "scheduled maintenance" too. Getting a bit worried about the nation's banking infrastructure we checked a couple more banks: Bank of America and Sovereign Bank. Both, from what we could tell without accounts, thankfully were up. Now this was by no means a scientific survey of uptime for online banking at 3 am on a Sunday (though I'd love to see one - it would likely be quite illuminating) but 3 out of 5 banks, all of them fairly major, having online banking downtime at the same time is unacceptable.
Online banking should never be down for scheduled maintenance. Really, no serious public-facing service should have scheduled downtime. There's just no reason for it. Load balancing, redundancy, rolling upgrades - there's no need to take down a whole service. At worst, there should be a performance hit during maintenance. There's no big trick to avoiding scheduled downtime: you just need a decent infrastructure with decent people managing it. It's really not that hard.
A company's public facing web services are easily as important as a flagship branch in meat-space. In fact, I would say that a company's web-based location should be considered the international flagship branch. Every single customer has the potential visit the web-based location - that cannot be said about any physical flagship location for a global, national, or even large regional organization, no matter how optimally it is located geographically. If companies aren't spending equal or more money and effort on their web-based locations as they are on their most premiere physical locations they are making a critical error. Scheduled downtime is a clear sign that they are making this error - it is simply shoddy workmanship of the sort that would never be tolerated at a physical flagship.