Why Cloud Infrastructure Needs Dispersed Datacenters

Plum Voice has been providing cloud IVR for over a decade. Over that span of time, this cloud concept—taking care of infrastructure and carriers so our customers don’t have to—has been called many different things. Regardless of whether it you call it cloud IVR, hosted IVR, or outsourced IVR, the simple reality is that somewhere, somehow, there is a real IVR that is answering your real phone calls.

Those real IVRs live in datacenters. And when you entrust your crucial IVR functions to a cloud IVR provider, you should expect them to have more than one datacenter. And, in fact, you should expect those datacenters to be separated by at least a few hundred miles for improved fault-tolerance and disaster recovery.

A couple years ago, hurricane Sandy struck the Northeast. At the time, it was uncertain where it was going to make landfall but we knew if it struck close enough to Boston so as to disrupt our infrastructure there, we could move our customers’ traffic to our other datacenters—including our site in Philadelphia. Conversely, if it struck far enough south to cause problems at our Philly datacenter, we knew our Boston datacenter would be just fine. As it turned out, it landed at nearly the midpoint between Philadelphia and Boston and both datacenters easily weathered the hurricane. Unfortunately it caused horrible damage to New York City and did, in fact, knock out many of the datacenters clustered around the New York metropolitan area.

This brings up capacity, however. We have designed our infrastructure such that no single failure will reduce our capacity to the point where our customers would experience a degradation in service—even if our customers were currently experiencing a burst in traffic. This design requirement is described as N+1 redundancy (i.e. if you have one datacenter worth of traffic, then N+1 redundancy would require that you build two datacenters). This means the true redundant capacity that you need to offer is still the amount of traffic that one datacenter can handle even though you’re paying for two. Let’s say, though, that you have three datacenters. Now you have two datacenters worth of redundant capacity and you still only need to pay for one extra. This is precisely the reason why we have three datacenters in the US: Boston, Philadelphia, and Dallas.

We’re not done yet because Plum actually has 4 datacenters. We have servers in London too. The primary purpose for this point-of-presence in England is to provide the same highest-quality TDM circuits for European customers as we do for our North American customers. There’s also an additional benefit with putting in a datacenter overseas: it’s not in the United States. Every day, we compress, encrypt, and deliver a backup of all of our databases—including our precious source code repository—to one of our servers in England. Just in case.

Here at Plum we love our weather-proof, call-spike-proof, disaster-proof cloud IVR infrastructure. It was great fun to build, exciting to grow, and a pleasure to manage. And we think that makes us the best cloud IVR provider in the business: after all, why would you entrust your communications applications to a cloud company that isn’t obsessed with infrastructure?