Optimizing Capacity Utilization in Point-to-Multipoint Networks

E.C. Foudriat and K. Maly (USA)

Keywords

Network access protocol,pointtomultipointnetworks,optimum capacity utilization.

Abstract

This paper describes a scheme for controlling the media ac cess protocol in point-to-multipoint networks to attain maxi mum capacity utilization. Greater utilization of assigned ca pacity improves the performance. It results in considerable re duction in message delivery time for those messages which can have a back log ready to send typically asynchronous (e mail, files, etc.) traffic. It further results in supporting greater loading before overload occurs. The results are applicable to a number of systems; satellite uplinks, aircraft-to-ground relay stations, and outlying cell phones to a central tower are typical examples. The underlying protocol uses two way control for block and slot assignment. It is designed to support a range of low level protocols using sub framing. It can effectively support multiple message types. ATM [6] and H3M [1, 2] are typical of such systems. Further, it has sub frames boundaries which are dynamic, i.e., there is a sub frame where registered-node requests are assigned fixed blocks or slots and an random ac cess sub-frame that unregistered nodes use on a contention basis. Active nodes are allowed flexible use of their assigned capacity. The access protocol system is developed as a ground to satellite uplinks. In this situation, the uplink system is di videdintonumerousbandswithnodesassignedtoshareaband. Banding is used mainly to reduce the number of guard slots. Nodes request capacity using uplink signaling. As a result, nodes with call and real-time traffic can tailor their capacity requests dynamically. When a call is in silent mode or a real time controller is not signaling, then minimal placeholder size slots can be requested. This allows the satellite to recover this unneeded capacity relative to a fixed slot-size assigned to the message. The recovered capacity can be assigned asynchro nous message types where data at the node is generally pend ing. The capacity assignment algorithm first assigns capacity to the accepted messages from active nodes; then to random access slots used by inactive nodes for setup and small data blocks;andfinallyitsremainingcapacitytoactivenodesasyn chronousmessages.Thislastassignmentenablesconsiderable improvement in the time for these messages to be delivered The paper discusses the protocol in greater detail and refer encesarecentlypublishedpaperwhereinitialresultswerepre sented [3]. This paper extends these result to indicate potential performance improvements under idealized conditions. For example, capacity utilization can decrease delivery times by a factor of 3 or better. A simulation system is being developed to morerealisticallydocumentperformanceimprovements.How ever, due to the wide diversity of network load and operational conditions, eventual use of this type of capacity assignment system would have to rely on an adaptive and/or knowledge based controllers.

Important Links:



Go Back