H. S. Rahul, H. Balakrishnan, and S. Seshan. An end-system architecture for unified congestion management. In Proceedings of 7th Workshop on Hot Topics in Operating Systems, pages 52--57, Mar. 1999.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....can be optimized specifically for the particular last mile characteristics (such as cellular modem use) rather than using the default TCP IP timeouts and data sizes. Similarly, client side TCP or TCP like connections can be optimized using connection multiplexing [8] e.g. to share information [20, 21]) While the architecture proposed above does provide for seamless mobility of clients, performance may su#er from the need to always communicate via the home proxy, as in Figure 2(b) If, however, the remote provider has also deployed this architecture, then when a new application is started, it ....

H. S. Rahul, H. Balakrishnan, and S. Seshan. An end-system architecture for unified congestion management. In Proceedings of 7th Workshop on Hot Topics in Operating Systems, pages 52--57, Mar. 1999.

....that while any individual host possesses only limited information about the network, the union of this information across hundreds or thousands of hosts can provide a far more complete picture. Other work has explored the advantages of sharing congestion state between flows within a single host [27, 21, 32] but to our knowledge the benefits of inter host sharing are unexplored (although the possibility is mentioned in [21] In the rest of this paper we ll review TCP s default congestion control mechanisms, discuss how they interact with current traffic patterns, and argue that both the host and the ....

....TCP sessions to the same site to request independent objects in a web page. While this kind of locality can be well exploited by our approach, we believe it is a legacy effect. New versions of HTTP [7, 19] new TCP implementations [32, 21] and new host level congestion control interfaces [27] are likely to reduce this source of flow locality. In Figure 2 we artificially extrapolate the impact of such changes by treating multiple flows between a source and a destination as a single flow, which we will refer to as a persistent flow. In particular, we coalesce all pairs of flows from a ....

H. Rahul, H. Balakrishnan, and S. Seshan. An End-System Architecture for Unified Congestion Management. In Proceedings of 7th Workshop on Hot Topics in Operating Systems, Mar. 1989.

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC