Bulk data transfers on the Internet or how to book some terabytes on red-eye bandwidth

Many emerging scientific and industrial applications require transferring multiple terabytes of data on a daily basis. Examples include pushing scientific datasets from particle accelerators/colliders to laboratories around the world, synchronizing data centers across continents, and replicating collections of high definition videos from events taking place at different time-zones. To accommodate the needs of such applications we designed, implemented, and validated Hermes. The value of Hermes lies in its ability to bypass Multiple Time-Aligned Bottlenecks (or MTABs). MTABs refers to the situation in which receiver side bottlenecks are succeeded or partially overlapped in time by sender side bottlenecks, thus condemning direct bulk transfers to long completion times. Hermes resolves MTABs by performing store-and-forward scheduling and routing through intermediate storage nodes. Thereby, it guarantees that transfers are constrained by individual, as opposed to more serious combined bottlenecks. Hermes can prove valuable to a variety of applications that involve wide-area, but also local, bulk transfers. We demonstrate this through a trace-driven case study, which is motivated by MTABs caused by ISP-induced throttling at senders and receivers residing at distant time zones.