As I talk to many companies looking to implement a private cloud solution and I ask them why they want to implement a private cloud, I often hear one or more of the following valid objectives:
- “We have to deliver apps and services to the business faster.”
- “We need to reduce growing capital costs and get more out of what we’ve got.”
- “We need to reduce the cost of supporting current services and scale existing staff/budget to support new services.”
- “We need to be able to show actual usage costs for IT services back to the business teams using them.”
- “We have to automate processes to reduce errors, delays and failures for a more stable and predictable environment.”
- “We have to make our infrastructure more flexible and responsive to changing demand and conditions to ensure we always meet SLAs.”
Unfortunately, and more often than not, we get responses like “everyone else is doing it,” “executives want it,” “our CIO committed to executive staff,” etc.
We also ask them what their vision is, or even more simplistically, what components they feel they need to achieve their objectives. We rarely find companies that have a clear long-term vision and instead we often hear a lot of talk related to virtualization, portal, service catalog, provisioning and a myriad of tools they have heard of or are considering as part of their solution.
While these are all critical components of a private cloud they are only a portion of the solution and alone will not achieve the objectives and ROI sold to the business. This limited view of a cloud most often leads to increased costs, overall greater inefficiencies, and in many cases it costs someone their job. But there is a solution and it’s not hard to achieve.
Best Practices to Implementing a Successful Cloud
There are many best practices to implementing a successful cloud. I won’t attempt to list them all, but a few very important ones include:
- Long-term vision – A long-term vision of your private cloud implementation is critical. It’s a simple case of “begin with the end in mind” and then work backwards into a plan that achieves the ROI.
- Phased vertical approach – Vertical vs. horizontal means implement the full value cycle across a single or a small number of groups rather than providing limited functionality across all groups and then adding additional functionality incrementally. You’ll rarely achieve the ROI if you take a horizontal approach.
- Understand the full cloud value-cycle – Look beyond static cloud 1.0 concepts of portal, provisioning, and virtualization and understand what functionality is needed to enable a dynamic cloud (Cloud 2.0) that uses policies, based on business drivers, to expand, contract, migrate, clone, reserve, burst, etc. to meet the needs of the business via service/application SLA’s.
10 Objectives that Can Be Achieved by Understanding the Full Cloud Value Cycle
When talking to companies about the objectives they want to achieve by implementing a private cloud, they may list one or two of the six common objectives listed above, but why only settle for one or two when all can be achieved through vision and proper planning supported by an understanding of the full cloud value cycle? The cloud value cycle as I define it involves the following items. There are more but this is a pretty complete list of items we see from the most successful cloud implementations.
1. Self Service Catalog – Everyone has one but it’s needed. A self-service catalog that can be configured and supported by web services is key.
2. Optimized Service Placement – Most solutions settle for general round robin placement, which is simply dumb. Mature solutions should allow for intelligent placement based on the needs of the service or the health/state of the environment.
3. Showback and Chargeback – I live in Utah and we have a local brewery called “Polygamy Ale.” They have some funny taglines such as “Bring some home to the wives.” Yes Utahns can laugh at themselves. Another of their taglines that illustrates the need for Showback/Chargeback is “Why have one when you can have two?” The same thing applies to free cloud resources. “Why have one gig of storage when you can have 100?” Companies shy away from chargeback as there are plenty of politics that would deter one from trying. So instead of chargeback, try showback. It’s less political, but it is helpful for users to understand the impact of their choices and makes them think twice before requesting 200 procs and a terabyte of memory. Try it! You won’t regret it and you will be a hero.
4. Reporting and Management Dashboard – I’m sure the one person who has read this far is saying “No duh” but you’d be surprised how many companies fly blind.
5. Auto VM Migration – Most solutions stop at the provisioning of a service. But 2/3 of the value is in the post provisioning space. Policy based VM migration maximizes utilization and helps maintain SLA’s. And no VMWare DRS will not get you there!
6. Auto Response to Events and Incidents – Every service is different especially when you think about homegrown applications. Being able to respond to the individual needs of a service is sweet.
7. Auto Maintenance and Future Reservations – This is especially true for large environments. Some think this is no big deal but to have workload migrated from hypervisors needing maintenance is a very, very nice to have and will quickly become a must have.
8. Power Management – Saves costs and makes you and your company feel good about themselves. We all like to feel good about ourselves. Oh yea, when you do this, publicize it and tell the world! Maybe others will follow. Al Gore will like you, but maybe that’s not a good thing.
9. Auto self-healing – It’s hard to keep up with hundreds/thousands of servers and hundreds/thousands services. When something goes wrong your solution needs to know what to do besides simply notifying the admin. It’s the 21st century for heaven’s sake; you can do this…easily!
10. Auto-decommissioning – It’s easy to provision services but most users never end their services even though they’ve stopped using them. Why would they unless you are charging them (shameless plug for item 3 above)? Not having the proper cleanup or shelving strategy can lead to VM sprawl and excessive costs. At a minimum shelve the VM and re-establish it if the user requests. Saves a bunch on licensing costs.
If you are just starting out with your cloud consider the above. If you already have a cloud it is never too late to implement these concepts and begin to realize the full value of your private cloud.