Planning a server refresh? Answer six questions on workloads, consolidation, platform, security, energy and migration before you buy new hardware.
Server hardware typically runs on a multi-year lifecycle, and a refresh is a chance to do more than swap old boxes for new ones. Answering a few questions first can reduce costs, cut power consumption and improve security for years.
What will actually run on these servers?
Start from the workloads, not the hardware catalogue. List each application with its CPU, memory, storage and network profile, growth expectations and licensing model. Some workloads may be better moved to the cloud or to SaaS; others, such as latency-sensitive or data-heavy systems, may belong on-premises.
Can you consolidate?
Newer processors with higher core counts often allow the same workloads to run on fewer hosts. Fewer servers can mean lower power, cooling and rack-space costs. Check how your software is licensed — per core, per socket or per host — because consolidation can change licence costs significantly in either direction.
Which platform model fits?
- Traditional rack servers with separate storage offer flexibility and familiarity.
- Hyperconverged infrastructure combines compute and storage in clustered nodes and simplifies scaling.
- Consumption-based on-premises offerings let you pay for capacity as you use it.
How will you secure the hardware?
Look for hardware root of trust, secure boot, signed firmware and a clear process for firmware updates. Out-of-band management interfaces should sit on an isolated network with strong authentication.
What about energy and sustainability?
Compare performance per watt, not just peak performance. Ask vendors about power-management features and how they handle take-back and recycling of retired equipment.
How will you migrate and retire the old estate?
Plan the migration window, test rollbacks and securely wipe or destroy storage in decommissioned servers, keeping a record for compliance.
Tip: run a short proof of concept with your heaviest real workload before committing to a platform.
Building the business case
A strong server refresh business case goes beyond hardware age. Finance teams respond best to a clear comparison of the cost of staying put against the cost of change.
- Support and maintenance: extended warranties on ageing hardware often rise sharply after the standard term ends.
- Energy and cooling: newer processors deliver significantly more performance per watt, which can reduce power and cooling costs.
- Licensing: consolidating onto fewer, more powerful hosts can reduce per-socket or per-core software licences, but check vendor rules carefully.
- Risk: older platforms may lack firmware updates and modern security features such as hardware root of trust and secure boot.
- Floor space: consolidation can free rack space or reduce colocation charges.
Best practices for a smooth refresh
- Collect at least a month of utilisation data, including peaks, before sizing new hardware.
- Standardise on a small number of configurations to simplify spares, firmware management and automation.
- Validate firmware, drivers and hypervisor versions against your vendor compatibility lists.
- Plan the migration in waves, starting with lower-risk workloads to build confidence.
- Securely wipe and certify the disposal of retired equipment to protect data and meet environmental obligations.
Common mistakes to avoid
- Sizing for today’s workloads without allowing for growth over the asset’s life.
- Ignoring memory, which is often the true constraint for virtualised and database workloads.
- Forgetting network and storage upgrades needed to take advantage of faster servers.
Frequently asked questions
How often should servers be refreshed?
Many organisations plan for five to seven years, but the right cycle depends on workload growth, support costs and security requirements.
Should we move to the cloud instead?
Evaluate both. Steady, predictable workloads are often cheaper on-premises, while variable or short-lived workloads may suit the cloud better.
A 90-day action plan
Days 1 to 30: collect utilisation data across compute, memory, storage and network for every host, and list support end dates, firmware levels and the applications each host runs.
Days 31 to 60: model consolidation options, request quotes for two or three standard configurations and calculate five-year costs including power, cooling, support and licensing.
Days 61 to 90: secure budget approval, order a pilot cluster and migrate a first wave of low-risk workloads to validate performance and runbooks.
Questions to ask hardware vendors
- What security features are included, such as silicon root of trust and signed firmware?
- How are firmware and driver updates delivered and automated?
- What are the measured power consumption figures under typical load?
- What lead times and spare part availability should we expect?
- What take-back or recycling programmes are available for retired equipment?
Key terms explained
- Consolidation ratio: the number of old hosts or workloads replaced by each new host.
- Performance per watt: the work completed for each unit of power used.
- Total cost of ownership: purchase price plus running costs over the asset’s life.
- IT asset disposition: secure, compliant retirement of old equipment.
The bottom line
Replacing infrastructure is an opportunity to reduce costs, improve security and simplify operations, not just to swap old boxes for new ones. Base decisions on real utilisation, build a financial case that includes energy, support and licensing, standardise configurations and migrate in waves. Dispose of retired equipment securely and responsibly. A well-run refresh leaves the estate smaller, faster and easier to manage for years to come.
Further reading on server refresh
For authoritative, vendor-neutral guidance on server refresh, see ENERGY STAR guidance on efficient equipment. You can also browse our free whitepapers.
