It’s no secret that I’m proud of the network Voxel has built over the past 10 years. We fly the Voxel flag at dozens of key interconnection points across three different continents. And with thousands of active peering sessions, most of our customer traffic is delivered directly to the end network once it leaves our global backbone. Handling 50Gb/sec spikes in traffic when a customer’s video goes viral or when Mozilla serves up a new release is just ‘another day at the office’ for us. There aren’t too many hosting or cloud companies out there that can make similar claims.
Our network is what connects businesses to our VoxCLOUD service, and it generally happens over the public Internet. More and more, however, we find ourselves wanting to extend our network directly to specific customer locations. Most of the time we can do this via either VPN or partnerships like the one we have with Equinix and TelX. In some cases, however, a dedicated connection directly between Voxel and the customer is the only way.
These ‘last-mile’ connections are manual and time intensive but are a necessary evil when the public Internet isn’t good enough. They involve researching providers that can deliver the required service to a particular geography, bidding, contract negotiations and, finally, coordinating the provisioning process.
Today, we announced a partnershipwith IPNetZone (www.ipnetzone.com), the New York City based network operator of the MPLS Exchange Platform (MEP). The MEP will allow us to more easily extend our datacenters and network points by facilitating last mile connectivity across multiple carriers and geographies. Voxel is proud to partner with IPNetZone and we hope to leverage MEP to make the last-mile problem less painful.
These last-mile connections are going to continue to be important to our customers, particularly as we start to promote a hybrid ‘public-private’ strategy. In a lot of situations, often for similar reasons why the public Internet isn’t ‘good enough’, our customers don’t feel comfortable with a 100% public cloud strategy. They want to mix and match. The basic premise is to create a unified fabric of capacity, mixing “private cloud” infrastructure located at the customer location (or a datacenter of their choosing) with “public cloud” resources at a Voxel datacenter.
There’s a lot of hype and misunderstanding about such hybrid clouds. Some (notably Amazon and Salesforce.com) have branded them “false clouds”. Of course, it’s to be expected that they’d be bearish on the idea. They argue that hybrid clouds aren’t elastic and don’t offer a pure OPEX model, which is a valid argument. Meanwhile, those that like to sell software licenses or ship gear (including the likes of VMWare, Cisco, Citrix, Microsoft, and large swaths of the VAR channel) see private cloud as the next big thing.
Who’s right? They both are.
Building your own in-house ‘private’ cloud requires spending all the money upfront, which is pure CAPEX (point one for the naysayers). There will also never be excess capacity if you haven’t already bought and installed it, meaning no elasticity (point two for the naysayers). Lastly, you will need in-house expertise to keep the thing up and running. It’s as if a company travel department decided to ban commercial flights for all employee travel and buy and operate a fleet of jets instead. It’d be expensive, distracting, and if you needed a few extra seats for a few weeks you’d have to buy another jet. Why not just buy airline tickets when you need them?
It seems like a sound argument, but what the naysayers fail to realize is that it’s not that simple. IT infrastructure is still a competitive weapon to most companies, and is far from commoditized. Over time, the infrastructure scattered across the thousands of enterprise datacenters and server closets will move away from the ‘edge’, and towards the ‘cloud’ (using those definitions very loosely). While the public cloud will eventually ‘win’, most enterprises are going to dig their heels in and resist this macro trend, at least at some level. Conservatively speaking, more than 90% of servers are still at the edge, and I think people overestimate how quickly they’ll transition away.
What’s at the root of all of this resistance? They mainly center around control and security, and the real or perceived lack of both. Back to my airplane analogy, while it may seem ridiculous to fly all employees via corporate jet, there are some advantages (besides being a hell of a way to fly!). You’re never going to get bumped from a flight. You have control over who can fly with you, the quality of the crew and the type of aircraft. And you have full access to all safety and maintenance records. These are all great advantages, but at what cost? The dirty secret in the infrastructure industry is that for companies with consistent workloads, assuming even moderate scale, the cloud isn’t actually cheaper than running your own infrastructure at the edge. The cost premium will probably start to erode as the cloud market matures, but it’s something that the current private cloud offerings are quick to point out (point one for the private cloud proponents).
If private clouds don’t have elasticity and a ‘pay per use’ (opex) model, what’s the point? Understandably, enterprise IT likes having complete control of their infrastructure in key situations, especially if it isn’t significantly more expensive, but they already have that. It’s not about virtualization – they’re already pretty well virtualized. VMWare is well entrenched, and about 1 in 4 physical servers will be virtualized in the enterprise this year. Why the interest in private cloud, then? It’s all about operational agility. With private cloud, the enterprise gets the service delivery benefits (instant provisioning, standardiziation, cost accounting) of the cloud, but delivered on their terms.
There’s a real opportunity out there for companies that can help enterprises make their IT more agile. They’re both enamored and threatened by the public cloud, at the same time. We feel that hybrid and private cloud are as much about process and best practices as it is about a technology platform. As a Managed Hosting company, it’s in our DNA to take a consultative, hands-on approach with our customers. We want to be more than an infrastructure provider — we want to be trusted advisors. We’ve also built our cloud stack entirely from our own code and open source components. We’re among a handful of companies that can put together and support the entire stack — top to bottom.
We’re currently working on operationalizing and productizing our private cloud offering with key partners and customers. Our goal is to eventually provide an onramp to the public VoxCLOUD service while providing customers the balance of control, security, and manageability that they’re looking for. Stay tuned for more on that!