There are almost as many different configurations of campus clusters as there are campus clusters, and because of that it can be hard to determine what the best way to run Pegasus workflows. Below is a ordered checklist with some ideas we have collected from working with users in the past:
If the cluster scheduler is HTCondor, please see the HTCondor Pool section.
If the cluster is Globus GRAM enabled, see the Globus GRAM section. If you have have a lot of short jobs, also read the Glidein section.
For clusters without GRAM, you might be able to do glideins. If outbound network connectivity is allowed, your submit host can be anywhere. If the cluster is setup to not allow any network connections to the outside, you will probably have to run the submit host inside the cluster as well.
If the cluster you are trying to use is not fitting any of the above scenarios, please post to the Pegasus users mailing list and we will help you find a solution.