next up previous
Next: Transport Layer Up: SADNA: ATM networks Previous: SADNA: ATM networks

Dynamic Bandwidth Allocation

The aim of this project is to build a layer which will provide dynamic bandwidth allocation services using a fixed bandwidth allocation from the ATM. Since the input to this layer will be very bursty there will be a need to modify the bandwidth allocation from time to time (but not too often).

One problem is that the ATM does not support a modification of the allocated bandwidth of a given Virtual circuit. In order to circumvent this, we will either allocate a few virtual circuits and use them in parallel, or each time allocate a new single virtual circuit and preempt the old one.

Stage 1: Allocation of a single CBR virtual circuit between two end station. This will include measuring what is the maximum effective bandwidth that the stations can sustain.

Stage 2: Implement different heuristics for dynamic bandwidth allocation. The aim is to divide the time to intervals, and at the end of each interval, to perform a new allocation. Until the new allocation takes effect, we will use the previous virtual circuit.

At this stage we will allocate a new virtual circuit with a new bandwidth and release the old virtual circuit. (You will have to take care of the synchronization between the new and old circuit.)

Stage 3: Same as stage 2, but rather than allocating a new virtual circuit, add an additional virtual circuits to increase the bandwidth or delete existing virtual circuits to reduce the bandwidth.

Bonus: Deal with the case in which you cannot increase the bandwidth.

Summary: This project will include a comparative study of the different alternative for deciding how to allocate bandwidth and the two methods of changing the allocation.

The project will write four routines:



next up previous
Next: Transport Layer Up: SADNA: ATM networks Previous: SADNA: ATM networks



Mansour Yishay
Wed Mar 6 15:24:36 GMT+0200 1996