GUEST POST: Mark Boslet, financial and high-tech journalist
Cloud computing is still a work in progress at most organizations.
Internal pilots and beta tests are underway. But projects are usually small in scope and involve IT development organizations or less important applications. If there is a slip-up, damage is minimal.
This will change. Corporations know cloud is the future of computing, and understand the need to build private clouds. The alternative is to watch business units move workloads to public cloud providers. It is a fight for IT survival, and managers are quickly becoming engaged.
But the complexities can be enormous. A true cloud requires resource pooling, on-demand hardware provisioning, virtualized data-center management, multi-tenant servers capable of running more than one application, and “chargeback” features to monitor and bill users for IT services. Data centers will need to become modular so hardware and software can be routinely added like lego blocks to expand capacity.
Many companies already have key building blocks in place. Virtualization is widely deployed in servers and now storage. Server clustering also is being adopted at a rapid pace. Big workloads are spread across multiple small servers or blades – combining the horsepower into powerful systems.
But private clouds are far more than virtualization, clustering or consolidating servers. This is especially true as companies begin to integrate the legacy computing systems they’ve spent millions of dollars shoehorning into data centers over the years. Big private clouds will grow from small seeds and take years to fully implement.
To get started, top vendors, consultants and analysts offer these tips:
- Top Down Analysis. Take a close look at the IT services a company delivers and the requirements for performance, security, system availability, disaster recovery, latency and bandwidth. Follow this analysis all the way down to the required number of server processor cores, cycles per second, storage transactions per second and costs per transaction. The pressure to get it right is high because it will guide the development of a replicable private-cloud architecture. The goal is to prevent the need to re-architect, says Ken Owens, vice president of security and virtualization technology at cloud service provider Savvis.
- Create Templates. Analysts advise IT managers to use a cookie-cutter approach to design modular templates for server, storage, software and other resources. When more computing power is needed, use the template to add capacity.
- Identify Key Building Blocks. Pick a cloud foundation – database, middleware, operating system, hardware – perhaps to match the products a company already uses. Cloud development then becomes evolutionary rather than revolutionary.
- Examine Workloads. Gauge which workloads are most suited to a cloud. One candidate: computer simulations of underwater gas fields. The application’s enormous demand for computing power comes in periodic bursts, so resources can be allocated elsewhere when simulations finish.
- Start Small. Start with modest projects before going with a large deployment, says Michael Cote, an analyst at RedMonk. Separate the tests from a company’s traditional IT environment and don’t try to integrate existing systems too early. Give the pilots a chance to prove themselves. The efforts will keep premature projects from doing more harm than good.
- Watch for Weak Links. Look for weak links when capacity planning, says Owens. If the utilization of a storage array or switch is high, monitor them closely and be prepared to add additional gear if necessary.
- Create a Scorecard. Organizations should create a scorecard for measuring performance and keeping track of successes and failures.
Mark Boslet is a long time financial, high-tech and clean-tech journalist. His work has appeared in The Wall Street Journal, the San Jose Mercury News, The Industry Standard and newspapers nationwide. He has been published online at Dow Jones Newswires, where he was Silicon Valley bureau chief, Greentech Media and TechPulse 360.