Virtual Path Bandwidth Control Method for ATM Networks: Successive Modification Method

Shigeo SHIODA  Hisao UOSE  

IEICE TRANSACTIONS on Communications   Vol.E74-B   No.12   pp.4061-4068
Publication Date: 1991/12/25
Online ISSN: 
Print ISSN: 0916-8516
Type of Manuscript: Special Section PAPER (Special Issue on Network Control)

Full Text: PDF>>
Buy this Article

Asynchronous Transfer Mode (ATM) technology is expected to be used in constructing a B-ISDN. ATM networks must support a variety of services, e.g., voice, data, and image communications with different grade of service requirements. The demand for these services and their traffic characteristics, however, are not yet clear. To implement B-ISDN under this situation, it is necessary to establish a network control scheme that can' absorb the difference between the estimated traffic and the traffic that is actually offered. In ATM networks, virtual path bandwidth control is a key control scheme for absorbing this traffic estimation error, and several control algorithms have already been proposed. When we try to further utilize the VP resource by dynamically reallocating the bandwidth according to the short-term traffic variation, however, we need control schemes that are highly responsive. This is achieved by using control intervals that are shorter than the intervals over which traffic fluctuates. Control algorithms based on central controllers generally need to collect a large amount of information from geographically widespread network facilities and solve a large optimization problem. This can make them difficult to use with short control intervals in large networks. An alternative enabling the shorter control periods is to use multiple distributed controllers that use only local information. This paper proposes two new VP bandwidth control algorithms suitable for this distributed implementation. In these algorithms, decentralized controllers are located at network nodes including ATM switch (ATM-SW) or ATM cross connect (ATM-XC) function, and each controller observes the quality of the VPs relevant to it. The bandwidth is modified successively as these distributed controllers communicate with each other. We therefore call this method "successive modification method" (SMM). Numerical evaluation using a model network shows the effectiveness of these algorithms for preventing the performance degradation caused by large-scale traffic imbalance within a network. Comparison with the batch modification method (BMM), which has no feedback effect, shows that the proposed SMM with approprate control intervals can be more responsive to traffic variation over time, but is slightly inferior when network conditions are static.