Grid computing is based on the „scale-out“ property of distributing functions via a network (grid) of a number of coupled systems. In this case, grid is used to express the distribution of an application over a number of computer systems which follow a coordination process which is transparent for the application. The grid coordination has today often been realized as a function within the application or is based on proprietary application server frameworks.    

One recent extension of the grid concept is cloud computing. The buzzword „cloud computing“ popped up in IT jargon a little less than a year ago. In the meantime, the stir it caused in the media appears to have died down. Cloud structures drive the concept of distributed utilization of infrastructure capacities beyond the walls of the computer center and out into the Internet. A virtually assembled infrastructure with distributed components which may, under certain circumstances, not even be operated in the users‘ own computer center any more, but are instead procured as a service via the Internet, represents the most flexible design of a consolidation platform. It provides the option of obtaining infrastructure services such as ­memory or CPU capacity as a utility – just like electricity or water – and then releasing them again quickly or cancelling them ­completely. Figure 2 provides an overview of application consolidation ­options.

The possible decisions range between the two poles for application consolidation: either a monolithic consolidation strategy or a scale-out strategy with distributed provision of services.    So what should an ideal solution for the scaling problem ­based on grid concepts rather than on monolithic structures look like?   

The vision: the meta-OS as solution for all scaling ­problems

A meta-operating system (meta-OS), alternatively called a „Virtual Data Center Operating System“ by VMWare or described by Gartner as a variant of its „Real Time Infrastructure“ (RTI) model, is a virtualization technology. The system manages distributed operating system, network, memory, and CPU resources. They are in turn stored on discrete server units which may be of different sizes. The concept is optimized when only small, standardized server units such as 2 CPU devices in combination with one or two standard rack units are used. The standard hardware makes it possible to add new capacity very flexibly and in small steps as well as to reutilize it. In ideal cases, this technology can even manage systems with differing hardware ­architecture and various operating systems, but this is still ­another wish for the future.