On the competitiveness of autonomous servers

Autonomous servers are introduced as a mechanism for solving many online problems, such as dispatching elevators in an office building, taxis in a city, or disk heads in a system with multiple copies of a file. The difference between autonomous servers and conventional online algorithms is that no centralized controller coordinates the servers. In essence, distributed online algorithms are studied under the assumption of limited knowledge: each server is independent of the others and has limited knowledge of the positioning of the other servers. The performance of such servers is examined using analytical and experimental methods. Applications and open problems are discussed.<
>