| 摘要 |
Systems and methods for dynamic intelligent mirror selection are presented. Dynamic intelligent mirror selection may include evaluating available mirrors at a client to determine an optimum mirror from which to download a portion of data. For each portion of data downloaded, reevaluation of the available mirrors may be provided such that dynamic selection of the mirrors may reflect changes in network conditions. The mirror selection may be based at least partially on observed characteristics of mirrors observed while downloading from a mirror, such that no synthetic work load is needed to benchmark mirror performance. The observed characteristics may include transfer speed, error counts, latency, and mirror load. Furthermore, a random bonus and usage bonus may be provided to help in facilitating the evaluation of available mirrors.
|
