On-line algorithms service sequences of requests, one at a time, without knowing future requests. We compare their performance with the performance of algorithms that generate the sequences and service them as well. In many settings, on-line algorithms perform almost as well as optimal off-line algorithms, by using statistics about previous requests in the sequences. Since remembering such information may be expensive, we consider the use of randomization to eliminate memory. In the process, we devise and study performance measures for randomized on-line algorithms. We develop and analyze memoryless randomized on-line algorithms for the cacheing problem and its generalizations.
Note: The Institute of Electrical and Electronics Engineers, Incorporated is distributing this Article with permission of the International Business Machines Corporation (IBM) who is the exclusive owner. The recipient of this Article may not assign, sublicense, lease, rent or otherwise transfer, reproduce, prepare derivative works, publicly display or perform, or distribute the Article.