A comparison of receiver-initiated and sender-initiated adaptive load sharing (extended abstract)

Derek L. Eager, Edward D. Lazowska, John Zahorjan

ACM SIGMETRICS Performance Evaluation Review · 1985 · 43 citations · 12 references

Concepts

TL;DR

Current locally distributed systems mainly share data and output devices, with limited computational resource sharing, yet load sharing can enhance performance by redistributing work from heavily loaded to lightly loaded nodes. The study aims to facilitate resource sharing in locally distributed systems. Load sharing policies are either static, using only average system behavior and independent of the current state, or adaptive; static policies themselves can be deterministic or probabilistic.

Abstract

One goal of locally distributed systems is to facilitate resource sharing. Most current locally distributed systems, however, share primarily data, data storage devices, and output devices; there is little sharing of computational resources. Load sharing is the process of sharing computational resources by transparently distributing the system workload. System performance can be improved by transferring work from nodes that are heavily loaded to nodes that are lightly loaded. Load sharing policies may be either static or adaptive . Static policies use only information about the average behavior of the system; transfer decisions are independent of the actual current system state. Static policies may be either deterministic (e.g., “transfer all compilations originating at node A to server B”) or probabilistic (e.g., “transfer half of the compilations originating at node A to server B, and process the other half locally”).

References

12