As hosting providers and service providers seek out the most efficient and powerful infrastructures for their "cloud" support capabilities, high utilization and lowest total costs become imperative. Amazon, Google, Yahoo, IBM, HP, EMC, Sun and Microsoft are all building huge cloud compute capabilities. Large enterprises and governments that also seek high performance and lowest costs will and should seek to emulate these providers and create "private cloud" datacenters where it makes sense.
By using and leveraging virtualization, MapReduce technologies, dynamic provisioning, automated pools of resources, boundary-less data services, and standardized infrastructure, enterprises can support applications, data and services in a cloud fashion. We may be working toward a de facto set of cloud infrastructure standards, but things today are still largely home-grown.
Meanwhile, the large "private" providers of web services to their employees, partners, and end users need to answer some basic questions. For example, How can I get flexible compute resources and how can I get access to them from wherever I am in the world? I think there is some very interesting models being put together, such as Amazon's S3 model. Businesses should tap into that model, using similar means for scaling up and scaling out their environments, without ever having to touch the hardware infrastructure and operating systems.
I expect that HP, IBM, EMC/VMware, Citrix and Microsoft will be leading providers of some or much of the private cloud infrastructure. I also expect global systems integrators to help their clients define and build out these capabilities.
This was first published in August 2008