I’m sure there will be zillions of other articles covering this in various trade publications, but right now, while I’m thinking about it, I wanted to dash off a couple of thoughts on how the cloud is affected by real clouds… like Hurricane Sandy.
I wanted to send an invoice to the Windows IT Pro folks for an article I had edited… but I couldn’t because the invoicing service I use was down. Ooops.
Over the last two days, I’ve gotten outage notifications from several of the services I depend on, including Trello and Harvest. As I write this on Tuesday, Harvest is back up, but Trello isn’t (nor is FogBugz, which is one of the candidates I’m considering for a hosted bug-tracking service). These outages are not unexpected; the NHC gave us all plenty of advance warning of Sandy’s likely impact. However, I was a little surprised to see how many services actually run on data centers in metro NYC. I have an intuition, but no proof, that the majority of these services are offered by small- to medium-scale companies that cannot yet afford their own dedicated data centers, which is to be expected. I predict that some of the services affected by this outage will move to relocate their services to another area, but some won’t; after all, no location is completely disaster-proof, and there are certainly benefits to having the services you offer hosted “near” your physical location.
The interesting issue is not that these services had failures; that’s to be expected. It is that as cloud service consumers, we now have to be aware of physical location in a way that “the cloud” is supposed to eliminate. File this under “cloud-related promises that turned out not to be completely true.”
Microsoft’s services, and Google’s, and Facebook’s, and Apple’s, and so on are all essentially location-independent. A metro-level failure caused by something like a hurricane or a major earthquake is a problem, but not necessarily one that end customers have to concern themselves with; there are always other data centers that can accommodate the load of the downed sites. In fact, this ability to provide continuity of service is one of the key drivers behind the architecture of Exchange 2010, and now Exchange 2013 extends continuity by simplifying the way load balancing works, thus making it easier to build larger stretched (“stretchier”?) sites.
Not that that helps Harvest or Trello users, of course. So along with the tired-but-still-important advice to ensure that your location isn’t cut off from the cloud by single points of connectivity failure, let me add a recommendation that you periodically survey the service providers you depend on so that you know where their services are hosted and can make arrangements accordingly in case of disaster.