Written by: Jonathan Sharp
This is the third post summarizing the major lessons of the new eBook, “How to Take Advantage of the Cloud: Your Guide to Achieving Value and Performance,” by HP and Latisys. Previously, we addressed “Cloud basics” using the phrase “how to harness the cloud as part of your IT strategy and “Cloud essentials” using the phrase “disaster recovery, storage, security, and compliance. In this post we’ll examine challenges to making the cloud work and challenges that should not be taken lightly.
Dumping entire IT infrastructures into the public cloud risks subjecting users to poor quality of service. It’s not just theoretical: many IT departments have tried running an application on, say, Amazon Web Services and they report that it runs maybe half as fast as it does on-premise. In most cases, users won’t put up with half-as-fast. They tend to start complaining at about 96-percent-as-fast. If the cloud is saving money only by lowering quality, then forcing IT into the cloud is like forcing a carpenter to build your house out of cardboard.
But not all clouds are created equal. Some private cloud service providers offer tiered services where you pick the quality of service level you need. You can choose differing performance levels for different applications. Likewise, although some cloud providers sadly use generic hardware, or hardware that’s not optimized for multi-tenancy, you can choose a cloud provider that uses brand-name servers, blades, firewalls, and load balancers that are built for multi-tenancy and provide greater security and isolation.
When I hear about people resisting the cloud, we’re delighted. Why? Cloud computing is a trendy topic right now, and so it needs to be resisted by people who don’t want to live through a horror story or over-hyped non-event (take Y2K as an example). We tell advocates of cloud computing to listen to the critiques, and the rationales behind them, to hear what they imply about how the critics see the company’s traditional strengths and future challenges. The results of that conversation can be, should be around optimization.
By optimization, I mean right-sizing your IT infrastructure to fit your strengths and needs. You should use the cloud to complement your existing infrastructure as you plan for the future. When you seek to optimize, you decide which applications to move to the cloud, how to balance security and cost savings, performance and elasticity, power and flexibility. Such decisions are hardly impossible, they’re actually similar to decisions that IT has been making for decades: how to deploy and consume, how to secure and manage, and when to build vs. buy.
The whole enchilada
Because IT is about the strategy of empowering workers with information and the cloud is merely the location of (some of) the physical resources to accomplish that strategy, I’m often surprised at how many people say, “If I was starting a company now, I wouldn’t buy any IT resources. I’d put it all in the cloud.” It’s true, there are some companies for whom such a strategy makes a great deal of sense. For example, consider the developer of an iPhone application or game: The demand could be huge and instantaneous, or it could build slowly and then burst, or it might not do much at all. Cloud resources help meet that demand with speed, scalability, and minimal up-front costs.
On the other hand, most enterprises today don’t need the whole enchilada. They do have legacy hardware and applications (and competencies and talent and advantages!). They do have some workloads that need to be on-premises. Rather than the whole enchilada, they just need, if you’ll permit us to extend this metaphor almost beyond its breaking point, a side of nachos.
So I’ll come right out and say it: the cloud is probably not the right choice for all of your workloads and applications. You have meaningful competencies and investments in mission-critical, latency-dependent, complex functions that should absolutely stay on-premises or within your current outsourced colo or managed hosting deployments. You should be skeptical of anyone who tells you otherwise. But you can ask of each application: Is the cloud the right choice for this? You may discover the answer is yes for situations such as:
- Truly variable workloads
- Applications that weigh down your cost structure, for example in software licensing
- Applications that are giving you trouble, because they’re temperamental, old, creaky, crash-prone, memory hogs, user-unfriendly, or otherwise apt to keep you up at night
Of course, even in these cases it’s not quite that simple. Cloud services come in multiple service models and deployment models. You may then need to ask if this particular application needs a basic public cloud or a private cloud designed especially for you. Actually, the way we see it, cloud computing is a subset of IT outsourcing, and so the real question to ask is not, “Can I send this function to the cloud?” but “Can I outsource this function, and if so, where?” Phrasing the question this way helps identify the strategic goals you want to accomplish—financial flexibility, redundancy, security, elasticity—and then lets the solution (cloud, managed hosting, managed services, etc.) arise out of the resulting profile of your needs.
For our expanded thoughts on these and other topics, please see the eBook here.
For more information on HP Converged Cloud visit www.hp.com/go/cloud
Follow me in Twitter @HPCloudSystem
Join the Cloud Advisors discussions on LinkedIn