Scheduling Seamless Downloads in Cache Networks

                     Ed Coffman
                 Columbia University

A file F is sited at a given node of a network; all other nodes are
able to cache part or all of F.  Client requests for F also
originate at these nodes.  The seamless download paradigm requires
that requesting nodes start receiving F at the time
the requests are made, and that they receive
it continuously until the downloads are complete.  We study the problem
assigning segments of F to the node caches in such a way that seamless
on-line schedules are realized and the total cache size needed
is minimized.  We also discuss many open problems arising out of
the seamless download scheduling model.