When a server component fails, its work is transferred elsewhere and the server is shut down. Regular maintenance may allow it to be repaired with a new fan, power supply or other component. But sometimes failures include the motherboard or another component that is not worth tearing down the server to replace. These servers stay failed, and when the zombie population reaches 20% of the container, it’s shipped back to the manufacturer.
Hindsbo said a key part of cloud economics is spreading the peaks and valleys of many computing users across all the servers in a cloud. It’s hard to achieve high utilization rates of servers in the enterprise because their loads are set and have regular spikes. In the cloud, those spikes are spread out, and utilization of servers can approach what he considers the ideal mark: 90%.
Microsoft on Monday has announced Hyper-V Fast Track and a set of services to speed customer adoption of Hyper-V and Hyper-V management as “an on-ramp to the (private) cloud,” said Barzdukas. It will supply reference architectures and best practices to speed customer adoption. Hardware makers IBM, HP, Dell, Fujitsu, Hitachi and NEC will be supporting the fast-track approach. They will provide a blueprint for what hardware to buy and what Microsoft software to deploy on it to achieve an internal cloud operation. For example, an HP BladeSystem Matrix can be combined with Microsoft Systems Center and Windows Server 2008 with Hyper-V to form an internal cloud combination. Look to Microsoft to take the lessons of cloud infrastructure and make it easier to equip with tools and applications as its bid for the cloud market.