Normally you don't run traces on a Production server unless you absolutely need to. However, the blck box trace is designed to be saved locally and the data is not sent across the network (network bandwidth is a major concern with traces). You usually only start a black box on a problematic server. The idea is to catch all the queries right up to the point the server fails or memory stalls. Doing this serveral times should give you a clue as to what query/queries ultimately lead to the problem. Then you have a starting point as to testing. In testing you try to recreate the symptoms again intinally. If you are able to then you break the query down to determin what wihin the query may be the cause. Other things to look at are the errors in the trace to see if any will give you an idea of maybe a bad setting or corruption in the Database. But again black box is intended to capture events on a server known to be problematic already and as such adding the trace is no real issue against performance.