As this year comes to an end, and I'm on my final (scheduled) flight, I thought I’d take a moment to reflect on some cloud computing trends we’ve been seeing here lately at Carpathia.
At Carpathia, we see a lot of pretty unique customer requirements, and from our vantage point, we also have a good view of trends both in terms of technology and workloads. One of the nice things about not being a one trick pony is that we can deliver the full spectrum of infrastructure services, ranging from colo to cloud. With that said, this may not be a particularly well-received post from the pure play cloud crew.
From our perspective, we really see somewhat of a maturity model presenting itself. It appears to have three distinct tiers and goes something like this:
1. Entering the Cloud
We often see customers getting started with hosting infrastructure for a new application, short-term project, workload-like development or test, or a variety of other reasons for which they’re not able to commit to a long-term contract (12+ months). This is what we see as our on-boarding funnel for the cloud: low friction adoption. Of course, I'm over simplifying this dramatically since we also see long-term customers bursting into the cloud for many of the same reasons mentioned above. Cloud bursting has been our most successful usecase at Carpathia with a majority of our cloud customers blending cloud and other services offered by us or inside the datacenters we operate in (note this isn’t the cloud bursting moving on-prem to cloud, I’ve seen no demand for this at all). One thing in common with these short-term customers is that they want nothing to do with servers or storage. They want to consume infrastructure to distribute their software/application/content and want the ability to turn on a dime (or turn off on a dime).
2. Exiting the Cloud
At some point, the previous tier of customers reach some kind of inflection where they start looking at the workloads in the cloud and think more about what their future looks like. Is the product viable? Are there long-term requirements for the solution? How is the performance and availability? Then comes the discussion – “what would happen if we moved XYZ out of the cloud and onto dedicated infrastructure?”
We can quickly alleviate one concern, which is, “I do not want to procure/deploy/run IT infrastructure”, be it a cloud or a dedicated solution; we can offer the same production experience. The trade off in this tier is agility vs. cost savings. Where in the cloud a new server is 3 minutes away, in the dedicated world you have logistics to consider. It takes time to get additional servers/storage delivered/deployed. On the flip side, if your VMs are running for extended periods of time, this is your cheaper option. Many customers hedge their bets and keep cloud (whether that be Carpathia’s, Amazon's, etc.) and pair it with their dedicated environment, in many cases the dedicated environment providing more exotic IT capabilities of flash arrays, high memory configurations, GPUs, etc. Something else of interest here is that many of these customers forgo virtualization in the dedicated infrastructure; instead looking at highly automated IT using tools like puppet. I was very pleased to see Gartner pick up on this trade off in the latest MQ. The good news is security isn’t a reason we here folks wanting to work in this manner any more. I think we finally have crossed that hurdle.
3. Doing it All Yourself
At some point there comes another discussion following this trend: “We are now operating at a scale where we want to consider using our own capital to procure the infrastructure (the previous tier was all about OPEX), and we have our own support team and want to run all/some of the infrastructure ourselves.” This becomes something we internally have tagged “molo” AKA managed colocation. -- our facility, your capital and a combination of our services and yours.
I guess there is one final tier; this is where very few in the industry finally reach, and that is doing everything – including building a facility yourself and variants of that (i.e. buying a POD from a Digital/Dupont). I'd group the ultra-growth companies into this mix, such as Facebook, Google and Zynga.
The great thing about working at Carpathia is that we can engage with any of the above customers -- no matter where they are on the continuum -- and deliver relevant services to support their needs. Since we can engage at any level (or all levels), we can also provide very unbiased views on the infrastructure landscape and quickly become a trusted partner for our customers.
Feel free to add comments on your own thoughts on this migration path. I'm sure we are not the only ones seeing this, but we might be one of the few openly talking about it.
(Updated 12/21/11 at 9:00PM -- Per some Twitter DMs, commented that scenario 2 has nothing to do with “security” and also further clarified the cloud bursting scenario.)