Why all workloads don’t belong in the cloud

November 2, 2009

Why all workloads don’t belong in the cloud

I had an interesting conversation with a CIO the other day about cloud computing. He had a simple question: I have an relatively old application and I want to move it to the cloud. How do I do that? I suspect that we will see a flurry of activity over the coming year where this question will be asked a lot.  And why not — the cloud is the rage and who wouldn’t want to demonstrate that with the cloud all problems are solved.  So, what was my answer to this CIO? Basically, I told him that all workloads do not belong in the cloud. It is not because this technically can’t be done. It can. It is quite possible to encapsulate an existing application and place it into a cloud environment so that new resources can be self-provisioned, etc. But, in reality, you have to look at this issue from an efficiency and an economic perspective.

Cloud computing gains an economic edge over a traditional data center when it supports a relatively small simple workload for a huge number of customers. For example, a singular workload like email or a payment service can be fairly optimized at all levels — the operating system, middleware, and the hardware can all be customized and tuned to support the workload. The economics favor this type of workload that support large numbers of customers. The same cannot be said for the poor aging Cobol application that is used by 10 people within an organization. While there might be incremental management productivity benefits, the cost/benefit analysis simply doesn’t work.

So, the answer is pretty simple. You just can’t throw every workload into the cloud. It is not a panacea for all IT problems.  Organizations that are trying to figure out what to do with these pesky old workloads need to look at three options:

1. Decide if that workload is still supporting business objectives in a cost effective manner. If it does the job, leave it alone.

2. That old workload might be better supported by traditional outsourcing. Let someone else keep the application alive while you move into more mission critical tasks.

3. Think about rebuilding that old workload — either by encapsulating key elements and placing them within a modular flexible environment. You might even discover that there are components that are actually useful across the organization. When you discover that sharing components across divisions/department is a productive and pragmatic approach, you might be ready to move those workloads into the cloud.

Cloud Computing
About Judith Hurwitz

Judith Hurwitz is an author, speaker and business technology consultant with decades of experience.

  1. Great piece and I wholeheartedly agree that, “Not all workloads belong in the cloud.”

    There has been a lot of buzz that Dev/Test workloads are cloud ready. In reality, they face the same challenge that you lay out for other applications- how do you determine which are the best for the cloud?

    This is especially important since most enterprise Dev/Test requires a combination of cloud, physical and virtual compute resources and the integration and orchestration of dozens of tools in the tool chain For those that are interested, I try to address how to determine which Dev/Test workloads are cloud ready in the following post. http://tinyurl.com/ykqjanh

  2. Thanks for your comments. While I agree that the provisioning aspect of cloud is a great productivity tool, it does not mean that all workloads belong in that cloud. You mention the term hosting when you talk about the environments that you company supports. There is a clear difference between applications that should be “outsourced” through hosting and those that should be a good fit for a cloud. I look forward to seeing your books. Our book, Cloud Computing for Dummies will be out in about a week.

  3. Whilst not every workload is correct for the Cloud – it is only half the story”Cloud computing gains an economic edge over a traditional data center when it supports a relatively small simple workload for a huge number of customers.” Whilst this is truue, it not the only reason

    For business users there is a far more compelling reason for Cloud applicatons – opportunity value. Their IT dept could take 3-6 months to procure a server and install som enew software. A Cloud based solution is available now – so the benefits flow 3-6 months earlier. Nimbus has hosted its 10 yr old BPM app, Control, (Win32 backend, web front end) for the last 4 years to a level where 90% of new installations for enterprise customers are hosted. Companies like Nestle, Carphone Warehouse, Toyota and SAP.

    BTW I’ve just finished writing 2 books; “Thinking of… Buying a Cloud Soltion? Ask the Smart Questions” aimed at customers and “Thinking of… Offering a Cloud Soltion? Ask the Smart Questions” aimed at ISVs migrating to the Cloud http://www.smart-questions.com

  4. Judith, makes complete sense. Moving over older applications almost always will involve a case for re-engineering. The dirty secret in the cloud world is that there are a limited number of “targets”: operating systems, middleware etc that are supported by the virtualization platform (such as VMware or Xen). That number is growing, but changing the underlying platform to suit the infra stack (and leaving the application alone) is one of the key areas that CIOs need to keep in mind as they contemplate a move to the cloud.

    Thanks for the good thoughts.

Leave a Reply

Your email address will not be published.