Thermal paste
Typical service life 3-4 years under datacenter conditions, less in warmer environments or under extreme load. Preventive replacement on systems 3+ years old avoids thermal throttling.
By manufacturer
Dell PowerEdge HPE ProLiant Lenovo ThinkSystem Fujitsu Primergy Supermicro IBM System x / Power Acer Altos
Maintenance · components
Component replacement: preventive, on parts that wear out by nature.
Hot-swap fans, RAID controller cache batteries, thermal paste, drives in predictive failure: these are components with a finite service life by design. Replacing them before the fault, within an agreed window, is vastly cheaper than dealing with the fault unplanned.
What wears out by design
Components with a stated service life.
Hot-swap fans
Nominal lifetime 50,000-70,000 hours. In practice the first drifts appear at 4-6 years. Symptoms: rising noise, RPM above nominal, fluctuation.
BBU / Flash backup batteries
The BBU (Battery Backup Unit) batteries on RAID controllers lose capacity over 2-4 years. Flash backup units last longer, but the supercapacitor degrades. Preventive replacement avoids losing write-back.
CMOS battery
The motherboard CR2032 typically lasts 5-7 years. Once flat, the BIOS loses its settings on every AC disconnect. A few minutes' work, negligible cost, and it avoids awkward discoveries after extended downtime.
Spinning drives (HDD)
Enterprise drives come with a stated MTBF, but the real indicator is SMART. An active predictive failure is the signal for preventive replacement. Swapping a drive predicted to fail costs a fraction of handling a rebuild after it drops out completely.
Enterprise SSDs
Service life is expressed in TBW (Total Bytes Written). Monitor the SMART attribute Wear Leveling / Media Wearout Indicator. Replace when it reaches 90-95% of the nominal value.
FAQ
The questions we get most often.
Drives in predictive failure: do I wait for the fault or replace right away?
Replace right away. Predictive failure is the drive's way of telling you "I am about to fail". A drive in predictive failure inside a RAID 5 array is especially dangerous: if it does fail and a second drive turns out to be degraded during the rebuild, the array is lost. Preventive replacement is almost always the rational choice.
Is a preventive replacement programme expensive?
For components that wear out by design, the cost is dominated by the part itself (thermal paste, battery, fan) — labour is marginal, especially when the work is batched with other maintenance operations. The real comparison is against the cost of unplanned downtime.
Do you replace components on EOL systems?
Yes. Fans, drives, thermal paste, CMOS batteries and BBU batteries are the components most readily available as certified refurbished, even for older systems. For very old specific models we can source compatible donor units.