Data Center Hardware Refresh: What to Keep or Replace

List processor models and cores, memory, boot and data drives, storage controllers, network adapters and GPU hardware. Measure the workload to find the limit; a faster processor alone may not fix memory or storage delays.

For GPU equipment, distinguish complete servers from separate cards or modules. List the included chassis, power, cooling and shared cluster connections before pricing or reusing it.

Dell PowerEdge R740 systems use DDR4 memory configurations; the R760 supports DDR5 RDIMMs and does not support DDR4. Check the processor, module type and population rules before carrying memory into the upgrade budget.

Some R740 configurations include NVDIMM-N or DCPMM persistent-memory modules. Identify those separately from ordinary DRAM because their retained data requires a retirement decision.

Check CPU features, hypervisor releases, firmware, adapters and guest operating systems across the source and target hosts. Confirm whether the server requires a manufacturer-customized hypervisor image.

vSphere EVC exposes a common CPU feature set within the same processor vendor family; newer instructions may remain hidden. Hyper-V compatibility mode cannot move a running or saved-state VM between Intel and AMD hosts.

Hyper-V’s dynamic mode requires Windows Server 2025 and VM configuration version 10.0 or later. A powered-off move can allow different processors in some cases, but check VM settings, operating-system and application support before relying on it.

List controller pairs, drive shelves, individual media, usable capacity and software versions. Measure growth, response time and read/write demand. A controller upgrade may retain existing shelves and drives rather than replace the whole array.

NetApp’s controller-upgrade methods depend on the models, ONTAP storage software version and chosen method; some require an outage. Check host changes involving the operating system, storage-adapter model, driver and firmware, multiple storage paths and the switch network. Confirm that the full combination supports both the move and any planned switch back. A matching protocol name alone does not prove compatibility.

Block storage presents volumes to a server, file storage shares files, and object storage stores data as named objects. Check how the application accesses its data before treating these as interchangeable replacements.

Record which local or external key manager protects the data, and keep the keys needed for recovery. A key that unlocks a drive can be different from the key that encrypts its data. Encryption alone does not prove that retirement erased the data.

Record product IDs, chassis, modules, ports, optics, cables, firmware, software, licensed features and support dates. Map traffic and shared services, and check software, warranty and support rights separately from hardware ownership.

Cisco’s MDS retirement guidance calls for working alternate storage paths, configuration backups and a rollback plan. Confirm switch roles and the site’s fabric design before planning each change.

Give the facilities team the equipment dimensions, weights and cooling requirements. Check rack and floor loads, power circuits, power distribution units, backup-power capacity, airflow and any required liquid cooling.

Confirm the delivery route, handling equipment and secure staging space. Account for old and new equipment running together while the technical team tests the replacement.

Record model, serial or asset tag, configuration, quantity, location and owner. List drives, persistent-memory modules and other storage parts separately when their data treatment or destination differs from the chassis.

Keep the decision to retain, upgrade, redeploy, sell or recycle, with its reason and approver. Name the receiving team for reuse or the quote and services agreed for equipment leaving the business.

Save the workload approval, test results, end of the rollback period and release date. List any equipment still needed for service or recovery and who can release it.

Specify the data, media, method and work location covered by your security policy. Identify backups and shared keys still needed for recovery, but keep passwords and key material out of the general equipment list.

Sanitization uses an approved method to make the target data impractical to recover. Record the media serial, method, tool and version, date, result and evidence that the work completed successfully without errors.

NIST SP 800-88 Rev. 2 separates two checks: verification reviews whether the work succeeded; validation decides whether it meets the data-security requirement. Keep that accept-or-reject decision and the operator, verifier and validator names required by policy.

Match the pickup and receiving lists to the released serials and included parts. Record who handed over and received the equipment, plus missing items, damage or media packed separately.

Keep reuse, sale or recycling results and the required certificates. Match the accepted equipment, charges and payment to the offer, and give finance the result for its asset records. Include software or support-right decisions that affect the next use.

List failed media, rejected erasure results, missing parts and disputed transfers by asset. Assign each issue an owner and next action. Name who keeps the final records and how long policy requires them to be retained.