That's a pretty blanket statement without taking context into account. (i.e. it might be the right approach for you, but it's not the correct approach for everyone).

If your service is spread over hundreds of servers then a cattle style naming convention makes sense. However, if you've got just 5 servers then pet style naming convention can be far more beneficial.

e.g. if you've got 5 servers and they're named after hobbits then when on sysadmin asks "What's going on with Frodo?" the other sysadmins will immediately know not only the server but what's running on there, if they've recently made any changes, etc. If those servers were named in a more cattle style approach then the question becomes "What's going on with DCPR002?", which is a lot harder for a sysadmin to parse and relate to the server.

What about service names?