Schema design for server performance data

I need to capture operational data on a set of servers. To simplify things a little let us says that for each server I need cpu and memory utilisation, and whether 8 critical services are running. One way to do this might be:
exchange,server=“EX01” cpu=10.0,memory=45.5, service1=“Running”, service2=“Running”, service3=“Running” timestamp

Another way might be to have a separate measurement specifically for services such as:
exchangeservice,server=“EX01”,service=“service1” status=“Running” timestamp

Is one method preferred over the other?

If I am using the first method and I want to capture cpu and memory utilisation at a different rate to service status, it is acceptable to have something like:
exchange,server=“EX01” cpu=10.0,memory=45.5 timestamp1
exchange,server=“EX01” service1=“Running”, service2=“Running”, service3=“Running” timestamp1
exchange,server=“EX01” cpu=10.0,memory=45.5 timestamp2
exchange,server=“EX01” cpu=12.0,memory=45.5 timestamp3
exchange,server=“EX01” cpu=11.0,memory=45.4 timestamp4
exchange,server=“EX01” service1=“Running”, service2=“Running”, service3=“Running” timestamp5
exchange,server=“EX01” cpu=10.0,memory=45.5 timestamp6
exchange,server=“EX01” cpu=12.0,memory=45.5 timestamp7
exchange,server=“EX01” cpu=11.0,memory=45.4 timestamp8

Or is this terribly inefficient (as it would be in a relational database)?