DARPA ULTRALOG Final Report - Industrial and Manufacturing ...

Manuscript for IEEE TRANSACTIONS ON PARALLEL AND DISTRIBUTED SYSTEMS 6
A 1
A 3
N 2
N 1
100
0
A 2 A 4
N 3
100
0
Fig. 1. An example network
2.2 Control actions
The network can utilize two different kinds of control actions in controlling its behavior:
algorithm selection and resource allocation.
Algorithm selection
A component can use one of alternative algorithms to process a task. Different alternatives
trade off CPU time and value of solution with more CPU time resulting in higher solution value.
As one can find optimal mixed alternatives, a component has a monotonically increasing
piecewise-linear convex function, say value function, with CPU time as a function of value. We
call the value in the function as value mode that a component can select as its decision variable.
A value function is defined with three elements as f v ), v v 〉 as shown in Fig. 1.
〈 i ( i i(min),
i(max)
This function indicates that component i’s expected CPU time 1 to process a task is f i (v i ) with a
value mode v i and v i(min) ≤ v i ≤ v i(max) . We assume that components cannot change the mode for a
task in process.
Resource allocation
When there are multiple components in a node, the network needs to control its behavior
through resource allocation. In the example network, node N 1 has two components and the
1 The distribution of CPU time can be arbitrary though we use only expected CPU time.