Portal load comes from user operations and background jobs. A fast homepage tells you little about search, writes or bulk exports. Measure the operations that matter before sizing the environment.
Capture the workload profile
Separate reads, searches, writes and background jobs. Record request volume, errors and p95 response time for each. A p95 value means 95% of measured requests finished within that time. Compare equivalent workflows and intervals.
Find the limiting resource
Correlate slow requests with CPU utilization, disk wait, database connections and queue depth. A growing queue while CPU remains available can point to a dependency or worker limit. Inspect database query time and lock waits before purchasing additional cores.
Contain expensive jobs
Set explicit concurrency limits for imports and exports. Consider a separate worker pool when long jobs crowd out interactive requests. Define how users see delays and how workers retry after failure. A retry should preserve a single intended record or payment.
Run a controlled export test
In a restricted environment, first reproduce ordinary reading and writing. Add an export while keeping the request pattern unchanged. If writes slow as lock waits rise, inspect queries and transaction duration. Repeat the same run after a correction. Report the build, data and conditions alongside the result.
Plan how the portal behaves during failure
Exercise a stopped worker, an unavailable database and a slow external dependency. Record which functions can wait, who receives an alert and how recovery is verified. Test restoration from an independent copy and reconcile records added after that copy.
Size from evidence
Match measured demand and headroom to a specific configuration. Include resource limits, operations, networking and change costs. A physical node may fit a confirmed hardware or isolation requirement. Evaluate continuity separately from the size of a single server.
Related planning guides
See Hosting requirements for a corporate portal for portal requirements and How to choose between a VPS and a dedicated server for the infrastructure decision. Discuss the workload through hosting selection.