EDP GOVERNORS

1. Introduction

EDP governors implements the policy for current budget allocation amoung
clients. In general, the governor decides budget allocation in the
following situations:

	[*] When a client makes an E-state request. If the request can
	    not be met with the remaining current, other clients may be
	    throttled to recover extra current which can then be granted
	    to the requester. If the request is unfarely high, a reduced
	    E-state has to be decided according to the policy.

	[*] When there is an increase in the manager's remaining cap,
	    the governor will try to distribute the surplus amoung
	    clients whose requests were previously rejected or who were
	    throttled during the above step.

	[] When a client has more than one borrower, the loan has to be
	   distributed.

Following sections provides a short description about available
governors.

2. Priority

As the name indicates, this governor implements a priority based
allocation in which higher priority clients are given preference. When a
budget recovery takes place, lower priority clients are throttled before
the higher priority ones. Similarly, during a promotion cycle or during
a loan update, higher priority clients are served first.

If the request can not be satisfied by throttling lower priority
clients, the requested E-state may be lowered at most to E0. This
ensures that higher priority clients are throttled only to provide
minimum guarantee E-state.

3. Overage

Overage governor uses a proportional allocation based on the difference
between the current E-state level and E0 (named the 'overage'). This
causes all clients to increase or decrease in their E-state some what
simultaneously. Hence this is fare allocation policy and ensures that no
client is throttled too much.

4. Fair

Fair governor policy is similar to overage policy, but the proportion is
based on E0-state level of clients.

5. Best Fit

This policy searches for a best-fit solution where the number of
throttles and remaining current is minimum. If the optimal solution
includes an E-state which is less than what is requested, then that will
be approved (subject to the general EDP rules).

Since the perfect solution would involve several passes across all
clients, a trade-off is made to approximate the optimum so that the
algorithm complexity remains linear.

6. Least Recently Requested (LRR)

An arrival-queue based policy where the least recently requested client
is throttled first.

7. Most Recently Requested (MRR)

Another arrival-queue based policy where the most recently requested
client is throttled first.

8. Round Robin (RR)

In this policy, clients are throttled in a round-robin fashion.