A RHEL server is dedicated to one customer, but it has unused capacity and runs applications with specific library requirements. Assuming the goal is to provide more functionality for that customer, what options would you consider?
How would you use spare capacity on dedicated RHEL hardware?
Asked by
-almost-at-capacity
Answers
Containers are a reasonable option for new workloads. They can package those workloads with their own libraries without replacing the libraries used by the existing applications. They still share the host kernel, though, so they aren't a complete substitute for a VM if the new software needs a different kernel or stronger isolation.
I wouldn't jump straight to a full orchestration platform on a production host. First check whether the proposed runtime is supported alongside the existing software, then test resource limits and the effect on the legacy service. CPU, memory and disk activity can all become bottlenecks even when average utilization looks low.
Reply to this answer
Another route is to virtualize the existing server, then run it and the new workloads as separate VMs on a hypervisor. That gives each environment its own OS and libraries, and makes it easier to allocate resources between them.
It's a bigger change than adding a container, though. Check vendor support, licensing, performance and the migration plan before assuming a physical-to-virtual move is feasible. You'd also need a tested rollback and a maintenance window.
Reply to this answer
Start with the agreement and the support requirements. Dedicated might mean the customer expects the whole machine to be reserved, or there may be vendor minimums that explain the apparent surplus. Even if the extra workload is for the same customer, get approval and define what performance guarantees must remain in place.
-almost-at-capacity:
That makes sense. I was thinking of adding functionality for the same customer, with their agreement.
Reply to this answer
Measure before choosing a solution. Look at peak usage as well as averages, including memory, storage throughput and network activity. Then compare the expected benefit with the work and risk of changing a dedicated production system. Sometimes the spare capacity is there for a reason and leaving it alone is the cheaper option.
Reply to this answer
Add your answer
Answers are reviewed before publication.
9minutes_to_closing:
Resource limits matter here. Spare capacity is only useful if the existing application can still handle its busy periods.