Results tagged “cloud considerations” from The Savvis Blog

"Will migrating to the cloud save me money?" This is a question that comes up fairly often in my discussions with customers. The reality is that there is no clear yes-or-no answer. It depends on a number of factors.

If you're currently looking at cloud adoption for your enterprise and are approaching it with the viewpoint of looking to save money, that is a valid business driver. That being said, a complete replication of your existing data center in a public or private service provider cloud is not guaranteed to save money and from Savvis' perspective isn't the right approach. In a future post I'll talk to some of the common business drivers for cloud adoption that we are seeing.

It's understandable that IT executives are looking at cloud in this way. After all, in the traditional model of IT, as outlined in Figure 1, the business need drives the application and the application drives the infrastructure. So the thought is, regardless of what the infrastructure looks like, as long as it meets the needs of the application, then it should be OK. This together with the idea that public cloud is a multi-tenant environment and so costs are shared across multiple customers leads to the perception that public cloud is cheaper. This isn't always the case.

In an IaaS cloud that paradigm is changed around, as depicted in Figure 2. The business need still drives the application, which still drives the infrastructure, but now the infrastructure has the capability and expectation to meet the business need as well. But it doesn't only have to meet the business need today, it has to meet that need at every point in the future. As we all know, the only thing that is clear about the future, is that it's unclear ... cloudy, perhaps.

A better way to approach the adoption of cloud is to first understand the different types of clouds that are available and what type of workloads would be suitable for each. I plan to write more about this in a future post but to be more specific, the types of questions to consider are:

- Where should you use a private cloud?

- Where should you use a public cloud?

- Where shouldn't you use cloud at all?

- And most importantly, how do you tie all of these different pieces together to form a cohesive solution?

By effectively answering the above questions you will be able to optimize your infrastructure to meet the needs of the application and business. Instead of saving money you will enable your company to be more financially efficient. If correctly planned and implemented a byproduct of this will be lower costs.

Cloud isn't a one size fits all proposition. Your cloud provider should know this.

How can I transform my enterprise to become cloud-centric? There is no right answer to that question. But there is an answer to the question "How can cloud serve my business needs?" There IS a way to harness the power of cloud to drive your business agenda rather than thinking the other way around.

Lots of times I hear my clients asking me "What should my enterprise cloud strategy be?" and "How can you help me accelerate into the cloud?" In my opinion, those are not the right questions to worry about. The concern shouldn't be how to become cloud-centric. Cloud is just one way to service your IT needs.

Instead the question should be "How can my infrastructure be more business-centric?"

We should first try to understand what the needs or challenges are of your business - is it time to market, resiliency or having to align IT spend with business outcomes? Then we should see what kind of enterprise IT architecture (that includes infrastructure and operations architecture) you need to adopt in order to meet those needs and challenges. In that quest for target state architecture, I'm sure cloud can play a pivotal role.

Having said that, there are some simple considerations that can simplify your approach/thinking around making cloud work for your business. They are: Workload, Technology, Efficiencies, Security and Business Case.

I plan to tackle each of these considerations one at a time on this blog, starting here with the most important consideration: Workload.

What does your workload look like? If you were to map the workload demand would it look like a human heartbeat - with ups and downs in very short intervals? Or is it much more seasonal - where it lies low most of the time and spikes up periodically? The distance between peaks is a very important factor in deciding whether or not something should be moved into the cloud.

While on one hand cloud is very well-equipped to handle sudden spikes in workload, there is a "cost" or overhead to RAPID provisioning and decommissioning. In a completely variablized cloud commercial model, the unit cost of resource (like compute) is naturally higher than a fixed-term-based model.

Oftentimes, we use the "pay by the drink" analogy when we talk about the commercial model of cloud. Well, it is very true - when you order drinks by the glass versus buying a bottle, what's more expensive? Obviously, by the glass. So, since, the variable unit rate is much higher than a fixed-term unit rate, unless there is a substantial amount of "rest" period in the workload, it doesn't make economical sense to leverage cloud for your infrastructure needs.

Now, that doesn't mean you SHOULDN'T use cloud in all such situations - you might have another compelling reason why you should. All these considerations are exclusive to each other. Even though one of them might stop you from thinking about cloud, the other ones might out-weigh the negatives and still justify the usage. So, I hate to sound like a consultant, but it DEPENDS on what your BUSINESS needs and priorities are ... that's what will drive your decision.

So, what kind of workload IS suitable for the cloud? A workload that is seasonal - retail applications that typically spike during holidays, financial workloads that peak up during period-endings, educational applications that peak up during admission season or non-production environments of usually very stable and static applications in production that might undergo patches a couple of times a year are just some of these prime applications.

In all these situations, the amount of time where the peak is happening is much lesser than the "off-peak times" and the peak loads are somewhat predictable. So, even though you are paying a much higher unit rate when you are using the cloud resource (such as compute), it is much lesser than what you would have paid if you had procured all of the infrastructure that you need at peak load and let them idle for most of the year.

So, hopefully, based on the above discussion you have a better idea now how to assess your workload for suitability in the cloud. In my next blog entry, I'll talk about Efficiencies in the cloud.