Depending on the task and whether the processing units are identical or not:

Article Improving makespan in dynamic task scheduling for cloud robo...

and

Article Ordered balancing: load balancing for redundant task schedul...

can be used to answer your question. The load-balancing approaches used here are proven theoretically to be optimal.

Load balancing in a distributed environment has been a problem looking to be solved for a number of years. Think of a demuxer (demultiplexer). It takes an input and selects one of the multiple lines out for that signal to go on. In a distributed computing software environment, the job scheduler (which runs on a single node, servicing requests) can decide where a job should run.

The load balancer works to steer the traffic to a pool of available servers through various load balancing algorithms. If more resources are needed, additional servers can be added. Load balancers health check the application on the server to determine its availability.

Kubernetes is just one of many software solutions

In high-performance distributed computing systems, robust load balancing refers to the process of efficiently and reliably distributing workloads across multiple computing resources (like servers, processors, or clusters) to optimize overall system performance, minimize latency, and ensure high resource utilization. Robust load balancing is crucial because it can handle various challenges such as uneven traffic patterns, server failures, and changing system conditions while maintaining performance and reliability.

Challenges in Robust Load Balancing:

Solutions and Approaches: