Protocols for fast, long-distance networks

Networks operating at 1 Gbit/s, 10 Gbit/s or even 100 Gbit/s and spanning several countries or states are now becoming commonplace. More and more users have to transfer routinely between multi-GB and multi-TB datasets over these gigabit networks. There is a growing range of application domains for such massive transfers including dataintensive Grids (e.g., in Particle Physics, Earth Observation, Bio informatics, and Radio Astronomy), database mirroring for Web sites (e.g., in e-commerce), and push-based Web cache updates. Although this high speed network infrastructure is already emerging, available transport and application protocols perform poorly over such networks. Standard TCP (TCP Reno or NewReno) is a reliable transport protocol that is designed to perform well in traditional networks. However, several experiments and analyses have shown that this protocol incurs substantial penalties when used for bulk data transfer in fast, long-distance networks. Starting at CERN (Geneva) during the International DataTAG project in 2003, the International Workshop on Protocols for Fast Long-Distance Networks (PFLDnet) has brought together researchers from the US, Asia, and Europe working on these problems. This issue of Computer Networks focuses on the challenges of protocols for fast, long-distance networks, building upon the community of PFLDnet, the IEEE Gigabit/High-Speed Networks workshops, and other workshops held over the past decade. For this issue, twenty four original papers were received from the combination of an open call and solicited as extended versions of the best papers of the past two PFLDnet workshops. Seven were accepted and appear here, including three from PFLDnet. They include two studies of the relation-