Prefetching over Heterogeneous Channels of Communication

T.L. Highley Jr. and P.F. Reynolds Jr. (USA)


Communication systems, network communications, fileprefetching, multiple channels, load balancing


Communication options are proliferating. On a single device, there could be several network options for remote communications: satellite, traditional Ethernet, a cellular network, etc. We explore concurrent use of multiple channels for performance and cost effectiveness. We introduce a new model for multiple, heterogeneous channels of communication, where each channel is characterized by its own bandwidth, latency, and per-bit monetary cost. File prefetching across heterogeneous channels is explored, using a new approach inspired by previous work on cost-benefit analysis for file prefetching. Our algorithm seeks a balance between good performance and reasonable monetary cost. It succeeds in finding a middle ground between a channel that is too slow and a channel that is too expensive by splitting the prefetch load across the two channels in order to attain performance that is both cheaper than possible with only the faster, more expensive channel, and faster than possible with only the slower, less expensive channel. Additionally, for some channel pairs, our algorithm can utilize two channels to attain performance that is both faster and cheaper than using either channel alone.

Important Links:

Go Back