CloudMonkey field notes

VPS and server sizing for South African SMEs

Choose capacity around your workload, growth and recovery needs—not just a RAM figure on a price card.

By CloudMonkey editorial · Published · Updated

Editorial buyer guide; individual specialist review has not yet been recorded. Our evidence policy separates practical guidance from verified delivery results.

Start with the work, not the server

List the applications, database size, concurrent users, busiest trading periods and scheduled jobs. A brochure site, accounting database and image-processing queue can have very different constraints even with the same visitor count. Capture current CPU, memory, storage latency and response times across normal and peak business activity before comparing quotes.

For a new workload, agree a small pilot with representative test data. Any initial specification is a testable estimate, not a performance guarantee. Ask what can be resized without downtime and whether a move to another host is required for larger disks.

Compare the whole operating cost

Ask whether the monthly price includes VAT, operating-system licensing, management, backup storage, outgoing traffic and after-hours support. Separate the server allocation from the work needed to patch, monitor and recover it. A low infrastructure price does not establish who owns those responsibilities.

  • Record baseline and peak resource use, plus the date and test workload.
  • Agree a response-time target for a business-critical action.
  • Document database growth and storage headroom; include logs and temporary files.
  • Ask where production and backup data are held and how you retrieve them.
  • Set a review trigger for sustained saturation or unacceptable response times.

Plan for a South African office connection

Test from the actual office and mobile connections your staff use. Include a realistic interruption to office power or connectivity in the acceptance plan. A reachable cloud server cannot compensate for an offline office router. Confirm who manages failover connectivity and what staff can do while access is unavailable.

Bring the workload inventory to a managed-cloud discussion and ask for a documented pilot, sizing assumptions and a costed recovery plan. Avoid commitments based only on synthetic benchmark scores.

Keep planning