In order to parallelize, we will have various nodes each working on an accumulator. In order to reduce those results together, we need a way (maybe a method in the Accumulator interface?) to combine multiple accumulators together.
For example:
Node 1 averages across (A, F)
Node 2 averages across (G, L)
Node 3 averages across (M, R)
Node 4 averages across (S, Z)
This will result in four different AverageAccumulator instances, and we need a way to reduce those four into a single AverageAccumulator.
In order to parallelize, we will have various nodes each working on an accumulator. In order to reduce those results together, we need a way (maybe a method in the
Accumulator
interface?) to combine multiple accumulators together.For example:
(A, F)
(G, L)
(M, R)
(S, Z)
This will result in four different
AverageAccumulator
instances, and we need a way to reduce those four into a singleAverageAccumulator
.