The Last Mile Problem

This past week I was vividly and personally reminded about what I think is probably the single biggest problem to cloud computing: the last mile. Cloud providers do provide service level agreements (SLAs) for different levels of promised availability but that really doesn’t help anything when the problem happens in that last mile of wire between your network provider and your local location. In my case, I was just out of service for over two days because a service truck took out the telephone wires right outside my office. Thanks to the fates and some unusually skillful driving (or whatever) my office was the only building affected– lucky me. Both the phone and my Internet connections were gone and it was a bit enlightening about exactly how much I tend to use those two things. It took Qwest two days and two repair calls to get the everything working again.

Related: SQL Server in a Microsoft Private Cloud

All in all, looking at the pig picture it was really more of an annoyance than a disaster. All my main servers are located on-premise and my product testing and other office work continued on without too much trouble. Email and web-browsing were out. However, with so  many vendors focusing  on the cloud these days I couldn’t help but think about what would have happened if the critical services I was using were cloud based. For this kind of local outage the cloud provider could not have held been responsible. While their services are dependent on that last mile of networking they are also independent of your Internet service provider. In this case, the outage would have been the responsibility of the customer. So there would have been no refunds for the cloud provider not meeting their SLA. An SLA in itself is no guarantee of availability.

Certainly there are ways to help mitigate the last mile problem. For instance, you can arrange for high priority services level from your Internet provider and you could arrange for duplex Internet service providers. Both of these measures would help but even that isn’t enough to save you from metro and larger geographical Internet outages which although rarer than local problems do happen. When it comes right down to it I was reminded the best availability is achieved through highly available on-premise infrastructure. This keeps the responsibility and control of your systems in your own hands. You aren’t held hostage by either another vendor’s SLA or by that last mile of wire coming in from the telco provider to your premises.