I don't see any way you can address this without measuring the bandwidth requirements of the applications in question.  Not so much because of the BW between the cloud hosting site and your campus, but also the hosting sites total BW costs, as well as the costs of application IO (storage, mostly).  A lot of this depends on the pricing details of the hosting site you choose, though.  (Amazon charges for almost everything, for instance, but some other sites merely cap BW usage.)

1 Gb is definitely not a high-bandwidth link when shared by applications serving significant numbers of people.  It's important to realize that web/mail type activity is relatively tolerant of congestion, but most intra-application activity is latency-sensitive.

It's also really important to make sure your decision-makers are not deluded by the intense salesmanship of cloud vendors.  Their hardware is *not* cheaper than the hardware any organization can host for themselves internally - especially if your IT needs are nontrivial.  (I like to say that past 2 racks of equipment, costs are nearly linear.)

Thanks Mark for remarking the costs associated with the bandwidth. I had a bad experience with Google Compute Engine because of that factor and it increases costs a lot.

Mark, do you know about papers defining a framework to help to estimate the costs associated with using public cloud infrastructures? I want a model or tool to help me to support my experiences in a more academic way.

Thanks,

No, I am not familiar with models or simulators.  I would strongly advocate collecting real, live data from your existing systems.  Linux systems, in particular, are very good about making system-behavior statistics available, for both bare-metal and virtualized systems.  I've never found Windows systems to be other than opaque, but thankfully also never had to work with them.  It's also often possible to get a very good idea of what's going on by just looking at raw device counters (such as network stats available via SNMP).