Transmission protocol for high bandwidth-delay product networks
The inefficient performance of TCP, being designed about 30 years ago, over high bandwidth delay product connections has attracted much attention in the past few years. Some research work attempt to solve this problem by using innovative window estimations. The bursty manner of packet transmission h...
Saved in:
Main Author: | |
---|---|
Other Authors: | |
Format: | Theses and Dissertations |
Language: | English |
Published: |
2009
|
Subjects: | |
Online Access: | http://hdl.handle.net/10356/18832 |
Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
Institution: | Nanyang Technological University |
Language: | English |
Summary: | The inefficient performance of TCP, being designed about 30 years ago, over high bandwidth delay product connections has attracted much attention in the past few years. Some research work attempt to solve this problem by using innovative window estimations. The bursty manner of packet transmission has compounded the challenges in the problem.
This dissertation first gives a brief introduction about the existing transmission protocols based on rate control mechanism, eXplicit Control Protocol (XCP) and Rate Control Protocol (RCP), which are interesting proposals to solve the problem if inefficient TCP. Then, we also propose a new transmission protocol which based on rate control mechanism. The new protocol adopts a crocss-layer framework combine flow-state-dependent dynamic priority scheduling (FDPS) and active queue mechanism to support QoS. The new protocol achieves fair bandwidth allocation, high utilization, near-zero queue size and almost no packet loss, fast responsiveness and great scalability, regardless of link-capacity resources, round-trip delays, and number of sources. Complex scenario simulations have shown that is has more responsiveness and requires shorter convergence time than XCP. The protocol employs a novel way to estimate the number of the active flows without maintaining flow states. The protocol does not need to maintain per-flow states in routers, which makes it suitable for implementation. Furthermore, the protocol has a low computation-complex that makes it suitable for backbone networks with high bandwidth-delay product. |
---|