leastqueuedepth

Least-queue-depth is a load-balancing heuristic used in distributed systems to assign incoming tasks to the processing element with the smallest current queue length, or depth. The goal is to minimize waiting time and balance workload across servers by directing work toward the least congested queue.

In practice, least-queue-depth often relies on probing only a subset of queues rather than all available ones, to reduce communication overhead. If several queues share the minimal depth, a tie is typically broken randomly or by a round-robin or similar simple rule. This approach is commonly implemented in web servers, computing clusters, message brokers, and data-processing pipelines, where rapid decision-making is essential and state information can be distributed or delayed.

Variants of the approach may include sampling strategies such as the power of two choices, where only two queues are inspected and the one with the smaller depth is chosen, or adjustments for heterogeneous servers where queue depth is weighted by service rate. Some systems combine least-queue-depth with other policies or time-varying measurements to adapt to changing workloads.

Advantages of least-queue-depth include reduced average waiting time and improved throughput in homogeneous environments, particularly under moderate to high load. Limitations arise from reliance on timely and accurate queue length information; stale or inconsistent state can negate benefits, and performance can degrade if servers are heterogeneous or if probing overhead becomes excessive. Related concepts include join-the-shortest-queue, load balancing, and variants of the power-of-two-choices strategy.