The rule of thumb for hardware End Of Life (EOL) is 3-5 years. This rule is based around finance orientation (depreciation). The real useful life question is more complicated and depends on a number of factors related to how well your IT organization is "managing" their IT assets. Some questions to consider:
+ Does your organization buy more along commodity lines or quality lines?
+ How often are you purchasing new assets: for growth and refresh? Are your purchases spread out fairly evenly year to year and within a given year?
+ How well is your environment virtualized?
+ How well is your environment automated - can I unplug old assets and plug in new assets without risk to the production environment?
+ Do you have a fairly standardized infrastructure across production, development, test, and DR?
You can increase the useful life of equipment through managing the above parameters of an IT infrastructure. Just because you want to refresh a production server doesn't mean it is end of life. Reuse as a development, test, or DR server is viable. The risk of reduced reliability can be mitigated through use of virtualization and clustering.
I think the biggest factor limiting the useful life in today's environment is cost. If you insist on keeping your equipment under maintenance (something that is more questionable in non-production use), eventually the cost of maintenance plus facilities cost (space, power, cooling) will exceed the cost of buying the latest technology.
So, I am sorry, there is no single answer for Design Life or Life Expectancy in today's world. I would say mission critical production life would be 2-3 years, secondary "important" use (development) would be 3-5 years, but supplemental standby (DR for non-mission critical appls or temp spares for secondary use) could go 5-7 years.