The following approaches can be considered for server maintenance after the 3-year warranty:
- recycle servers after 3 years;
- replace servers;
- keep spare servers;
- purchase components.
One approach is to consider that after 3 years, a server is meant to be recertified. This is the kind of approach that is implemented by companies such as Facebook. All servers are thus phased out after three years and replaced by new ones with better energy performance. Phased out servers are sold to the circular economy for recertification.
Another approach is to replace servers after they crash: this might happen after 3 years, 4 years, 5 years, etc. Crashed servers are replaced by new ones with better energy performance. Crashed servers are sold to the circular economy for recertification.
A variation of this approach is to keep some spare stock of new servers and replace crashed servers with spare ones.
It is also possible to replace certain components such memory, SSD or CPU which are common enough to be easily found on the market.
Rapid.Space's opinion is that the best approach consists of replacing servers after the first crash after the end of warranty and then rely on the circular economy to optimize costs.