Research On Scheduling Algorithm Optimization Based On SCTP And SCTP-CMT

With the rapid development of Internet and network technology, the traditional transport layer protocols can't meet the requirements of current used network transmission because of the inherent defects. SCTP, which is a new transmission protocol in the network transmission protocol of next generation Internet architecture, has been the industry consensus. As the diversity of network accessing and equipment costs reducing, but also for toaggregatting the bandwidth, improving the end-to-end transmission throughput, researchers had proposed multi-channel transmission at the same time (Concurrent Multipath Transfer, CMT) solution on the features of SCTP,that is to say SCTP-CMT, which has become the hotspot of their research. The research of key scheduling technology for stream has important theoretical significance and application value.As a new transport protocol, SCTP also has some defects,1) when we transport based on the SCTP protocol, due to the particularity of the current agreement, SCTP did not defined how to schedule the flow; for more typical fixed choice, and which cannot meet the requirements of flexibility and particularity;2) when we transport based on SCTP-CMT, Due to the cache block and the reordering of information, we need to find a simple and the best way to make more full use of multipath and flow, in addition to the large difference in path performance or more complex scenarios, we need to find the optimal combination.In order to solve the first problem, first of all, we introduces an embedded scheduling algorithm (Pluggable Scheduling) concept, In order to make the SCTP applications to more flexible customization flow scheduling algorithm, combining implementation of several scheduling algorithm in SCTP protocol packet and uses the API to select and configure algorithms we want and provides a central interface to customize algorithms. If your application needs to use some of the common scheduling algorithms which had been provided, you can simply call API to select the algorithm. If an application requires some special requirements, but also need to use a customized scheduling algorithm, the user can through the core interface to implement. As a result, this also maintains the balance between operational ease and flexibility. In order to solve the second problem, we introduce fixed flow with the pathmap (fixed stream mapping) algorithm here, because of needing reorder information, the flow need map to a specific path; But due to the different path properties and buffer size limit, it will choose the best combination of flow path; when we analysis the broadband performance, due to the particularity of the application layer, we puts forward an API extensions to optimization, in order to solve the disadvantages of cannot make full use of broadband; In more complex scenarios, the application layer at the receiving end to rearrange information transmission performance degradation caused by the negative effects of using the NR-SACK expands to improve the performance of mapping algorithm.Finally, we summarize this paper and propose the future research directions.