Scalable Computing and Communications: Theory and Practice

28

–––––––––––––––––––––––

Toward Optimizing Cloud Computing: An Example of Optimization under Uncertainty

Vladik Kreinovich

28.1   CLOUD COMPUTING: WHY WE NEED IT AND HOW WE CAN MAKE IT MOST EFFICIENT

In many application areas (bioinformatics, geosciences, etc.), we need to process large amounts of data, which require fast computers and fast communication. Historically, there have been limits to the amount of the information that can be transmitted at high speed, and these limits have affected information processing.

A few decades ago, computer connections were relatively slow, so electronically transmitting a large portion of a database required a lot of time. It was, however, possible to transmit the results of the computations quite rapidly. As a result, the best strategy to get fast answers to users'requests was to move all the data into a central location, close to the high-performance computers for processing this data.

In the last decades, it became equally fast to move big portions of databases needed to answer a certain query. This enabled the users to switch to a cyberinfra-structure paradigm, when there is no longer a need for time-consuming moving of data to a central location: The data are stored where they were generated, and when needed, the corresponding data are moved to processing computers; see, for example, References 1‒5 and references therein.

Nowadays, moving whole databases has become almost as fast as moving their portions, so there is no longer a need ...