Investigate what is required to effectively run caliper in a K8s environment

[davidkel]

K8s is a popular platform for running containerized applications. K8s has now become more capable around short term executions (ie not long running servers that require to be kept running or scaled when increased workload is required) this means that K8s can now be a platform to run caliper but what is the best way to achieve this ? Also does it make sense for the caliper manager to be able to create workers rather than having it managed externally. This investigation therefore should focus on

• how to deploy a caliper manager into a K8s environment
• how to deploy mosquitto into K8s for remote workers
• how to deploy the correct number of workers
• how to ensure that caliper itself doesn't become the bottleneck in this environment (how do you detect this)
• how to ensure that caliper workers are assigned to a node with enough resources
• how to ensure that a node has the optimal number of workers assigned
• investigate a new way for caliper manager to launch workers in K8s which could become part of the code base
• document this pattern in the caliper docs