Definition
Latency Hiding ( Distributed System)
Latency hiding overlaps communication with useful computation, so that a process need not remain idle while waiting for a remote response.
The process issues an asynchronous request, performs work that does not depend on the response, and waits only when the result is needed. This reduces exposed waiting time, not the communication delay itself.
Overlapping Work
Let
The model ignores scheduling overhead and resource contention. Batch workloads and parallel applications often provide independent work to overlap; issuing an asynchronous request does not help if the next operation immediately needs its result.
Reducing the number of remote exchanges or performing more work locally instead reduces communication itself, rather than hiding its latency.