Analysis and Enhancement of Multipath Mechanisms Focusing on the OSI Transport Layer

点击翻译按钮获取中文摘要

The interconnection of end systems and network components by multiple network paths (multihoming) is a classical approach to increase the availability of the connectivity. Due to the availability of multiple network interfaces in modern terminal devices and the standardization of transport protocols with multihoming support such as Multipath-TCP and SCTP, this scheme also becomes attractive for the end user. Beyond the pure redundancy in case of errors, another question is whether a simultaneous usage of multiple paths is possible in order to increase the data throughput. This approach termed Concurrent Multipath Transfer (CMT) on the one hand requires adaptations of important protocol functions, and on the other hand affects the utilization of the resources in the network in a manner which does not appear in a single-path transmission. The simultaneous use of multiple paths opens up new degrees of freedom but also increases complexity. This facts as well as the correlations among various mechanisms involved, which are not sufficiently understood yet, create a fundamental need for research.The focus of legacy approaches aims at the basic functionality and not on the optimum usage. Well-known mechanisms are used which were, however, designed for an optimum transport via a single path. Both the selection of suitable mechanisms for the optimum transport via multiple paths as well as suitable extensions or, if applicable, new definitions of such mechanisms are not sufficiently investigated up to now. Among other issues, algorithms for the path selection and the allocation of the data streams to the paths are necessary for an optimum usage of the available resources fulfilling the requirements of the application. Furthermore, basic definitions, e.g., related to the fair share of transmission resources have to be adapted or newly defined and the, often implicit, coupling of different functions, e.g., congestion control and fairness, which is typical for current protocols have to be reevaluated. These generic, protocol-independent results significantly affect the design of the CMT-enabled transport protocols.A major goal of the project is the systematic investigation of the new degrees of freedom and the related fundamental challenges of a general application of CMT. The starting point for this are the currently discussed proposals and the insights which have already been achieved by the applicants in this context. A mathematical modelling of the optimization problem for the resource utilization is expected to yield important findings. The outcome of the project should be a largely generic design approach for CMT-enabled protocols which incorporates the new requirements and possibilities. The practical benefit will be demonstrated by a applying this approach in the on-going definition of the current protocol proposals.