The requirements look simple, but actually it’s rather complicated. A service may be available, but operations may fail and vice versa. There are also many potential problems: network failure, pool / service shutdown, wcf activation failure, operation errors ... The service may be available, but all operations fail due to a damaged database. WMI / Perf counters seem to be a good approach since you can have many details for different areas (service / operation / endpoint). But in addition, if your hosting of your service in IIS, the default application pool timeout is 20 minutes: the / WMI counters will be unavailable.
In conclusion, I would say that the best monitoring strategy is to aggregate all the tools / components available on the server side and on the client side.