By Scott Ryan
How many times have I heard one of these in meetings in the past few months:
When it is stated, it is usually in conjunction with a hand wave and a tone of teenage ‘duh, obviously!’ Pardon me if what pops into my mind is the scene from Old Schoolin which Will Ferrell states, “Hey, we’re streaking…everybody’s doing it!” (warning: there’s a lot of Will Ferrell in that link.)
I mean no offense to anyone who has made any of these statements to me, because many of them are just being told by their management something like, “Hey, I was reading HBR this weekend, and they talk about this whole cloud thing revolutionizing business. Do what’s right…but…really, move everything to the cloud. Before the next quarterly earnings call if you can. I want to cover this on that call.”
So when are people going to realize that this Emperor has no clothes? The ‘move it all to the cloud strategy’ sounds a bit like the “we are outsourcing all IT” strategy go the 80s and 90s. It sounded great at the time, but how many companies found themselves trying to unwind their IBM, EDS/HP, or Perot/Dell, outsourcing contracts 10 years later? Answer: fewer once IBM started suing their outsourcing customers.
“Move it all to the cloud” just doesn’t take some hard realities into account about how applications and IT operates. Let’s take a quick look at both.
Latency, Performance Killer, Qu’est-ce que c’est?
The cold, hard fact is that not all applications are well suited to go to the cloud. Simply drawn the picture looks like this:
Chatty Applications: Some applications produce or request a lot of traffic in small bursts, which the application needs to present quickly to the end-user to provide a reasonable experience. This is not a cloud capable workload.
Chatty Storage: Some applications produce a lot of non-sequential read or writes for data from storage which is needed for the application to progress in it’s processing. Chatty Storage from the cloud is not cloud capable,
Batch Processing: Some applications are a scheduled jobs or background processes that has no user waiting for a response. These types of applications can speak to storage in a cloud in a workable way.
Light Traffic App: Light traffic between the user and an application provides mean infrequent data updates and low effect by being in the cloud. These types of applications are very cloud capable.
There are a number of other reasonable variations of these factor, but I think you get the idea.
The problem with these is network latency. Latency on the Internet is a combination of the time it takes to get from one hop to the next (propagation delay) plus the time spent waiting at each hop to go to the next one (queuing delay) – think: trains and train stops. The higher the latency, the slower the application or storage will appear to be. The fact may be that the app or storage is actually very fast, but it is spending a lot of time waiting for the data or compute. The result is slow performance (shameless plug: Concurrent’s NightStar Tools do a great job of finding application latency bottlenecks.)
Thus, moving all workloads to the cloud is not realistic for every business.
And then there’s support…
So moving workloads to the cloud de-risks some IT decisions and makes application infrastructure more flexible. It also turns capex into opex, which depending on your business measures is maybe good and maybe bad (good CIO summary here).
What the cloud doesn’t change is IT operations and support. People are still needed to monitor the application servers, storage and network, and resolve any outages or performance issues. And then there are those pesky end-users. They tend to have questions about the organization’s applications and their devices. People are needed to provide support to end-users for their applications and devices.
These support needs are precisely why the Managed Services market is exploding, as organizations seek solutions to the support and management of their application workloads. In fact, Managed Service Provider (MSP) services are needed and useful whether a company is hosting their all own application workloads in a private cloud, transitioning some of them to a public cloud, or managing a hybrid of private and public cloud workloads.
Before you get caught up in the exuberance of “going clouding” make certain you have done your homework on your application workload performance needs and that you understand the impact it will have on your support needs. You don’t want to find yourself exposed and alone in front of your customers or management trying to explain why the cloud didn’t solve all your IT cost and support issues.