Many of the major players have outlined utility computing strategies that are very platform-specific, how do I go about gaining the benefits in a heterogeneous data centre environment?
After a consult, the doctors sit you down in front of their impressively large desk and say:
Dave Roberts, Inkra Networks:
I believe that the deployment of utility computing within the enterprise will arrive as a series of smaller changes to the existing infrastructure. Because of that, I don't expect that people will really adopt "full-blown utility computing," per se. Rather, they'll gradually move to a utility model as the technology matures. Utility computing is not an all-or-nothing proposition; it can be done in stages that bring individual cost and complexity reduction benefits. In each of the three fundamental data centre technology sectors - networking, computing, and storage - products are available to start the journey today.
Despite the hype, no one vendor today delivers all the pieces of a utility-computing solution. It requires multiple vendors and multiple systems, including a wide array of servers, networking equipment, storage systems, and management software. With that said, your deployment will be heterogeneous by nature. However, to ensure that you're selecting the right solution for your organisation, here are some key issues to consider.
First, to start your migration to the utility model on a successful footing, you'll need to have a clear picture of what you've already deployed. Data centres are complex environments and many organisations simply don't have a good understanding of the equipment already in place.
Next, you'll need to balance your existing infrastructure with the strategic investments you'll need to make to move to the utility model. Also, pay attention to management systems. Today, management systems exist that allow you to tie together networking, computing, and storage resource pools. Partnerships and alliances are key to building working systems, and you'll need to ensure your data centre equipment vendors are exchanging information.
Here are some questions to ask of your utility data centre vendors:
• What are the quantifiable operational and capital expenditure benefits I gain by moving to this model?
• How much control do I have over the services and performance levels in my environment?
• What is the adoption strategy? Is it an all-or-nothing approach or something I can do incrementally?
• How well does the system support a hybrid solution where some applications and pieces are using the utility concepts and others are not?
• How does the system manage the virtualised resources to ensure that there is clear performance, security, and fault isolation between them?
• Can the system support advanced billing and chargeback models? (Though enterprises often don't use those today, they eventually want to. The infrastructure put in place today should support requirements for accounting moving forward.)
In a heterogeneous data centre environment, standards will be a critical component of a successful utility computing model. There is a need for standards to bring greater interoperability to data centre equipment and provide a roadmap for utility implementations. Look for data centre equipment vendors that are participating in new standards efforts, such as the Data Centre Markup Language (DCML) organisation (DCML definition). These standards will provide a foundation to enable utility computing and will ensure interoperability between a wide array of data centre elements, including network components, storage components, Unix, Linux, Windows and other servers, software infrastructure and applications.
Data centres are complex environments and you won't be able to rely on one vendor to fulfil your utility computing needs. Customers today must demand that vendors provide an easier way for their various products to communicate, so they can interoperate and leverage the knowledge captured within each system. So, ask your vendors the tough questions and make sure they will deliver on their promises as you start making the transition to the utility model.
Jim Metzler, Ashton, Metzler and Associates:
I completely agree. In implementing utility computing, the journey is just as important as the destination. Think of it this way: most IT organisations are not comfortable (or able) attempting to implement a fully virtualised utility computing architecture all at once. Therefore, it is only logical that they implement these technologies incrementally, over time. As an IT manager, it is imperative for you to show the benefits of each of these steps you are taking. You cannot say, "Just wait three years - then you'll see the benefits from all of this data centre spending." Instead, it is crucial to show the benefits that each stage of deployment brings.
In addition, though some vendors would like you to believe that the promise of utility computing is completely achievable today, that is not the reality. Realistically, we are all hesitant to accept risk, particularly when it comes to something as critical as the data centre. We have relatively flat budgets, and we are dealing with a very complex entity (the data centre). When you consider these issues, it is easy to understand why most IT managers want to take an incremental approach rather than diving in head first. IT managers have learned that the risks associated with integrating new technologies can be significantly decreased by adopting a step-by-step implementation.
Now, having said that, in order to be successful with utility computing you must have a clear vision of your own IT goals. You also need a deep understanding of the different architectures of the major utility computing players in order to determine what model is right for your organisation. As part of your analysis of the different architectures, you must understand where there is commonality and agreement (in the form of standards, etc.), as well as where there is disagreement.
This requires doing some homework. You must be sure to ask granular questions of the various vendors involved to determine which architecture overall offers the best added value and interoperability with your existing infrastructure. The assessment of these architectures and an understanding of how they will work within the framework of your organisation is absolutely critical. The questions that Dave introduced provide a good starting point for the granular level of detail you will need from vendors before deciding if their offerings are right for your organisation.
A good understanding of your own goals as well as the architectural visions of the major utility computing players will also help you determine the migration path for your evolving data centre. You must consider your current storage, server, and networking resources and then determine what steps your organisation must take to achieve a functional utility computing model. By developing a plan to incrementally deploy new data centre technologies every six, nine, or 12 months, you can minimise the risks by controlling each step of the deployment. This also allows you to realise - every six, nine, or 12 months - a tangible benefit in at least one area of your data centre.
In order to fully realise the benefits of utility computing in the heterogeneous data centre, the bottom line is the planning. Take a thorough inventory of your current data centre resources, research the major utility computing players and technology vendors to determine what technologies will best interoperate with your current environment, then plan an incremental approach that will allow you to keep close tabs on the progress of your deployment - each and every step of the way. Diving in to new technologies head first is both risky and ill-advised.
Though I urge you to be critical of the vendors you choose to work with, you don't need to be cynical. There are a lot of exciting technologies available today that are revolutionising the data centre. A planned, incremental road map will help you realise the benefits of these technologies while mitigating the risks of new technology deployments.