There are many ways to calculate CVAs on a portfolio of interest rate instruments; with varying degrees of sophistication. How accurate do you need it to be? Be aware that any method you chose will be very sensitive to the credit curve (containing the probabilities of default during each piece of a partition of the full tenor considered) that you bootstrap from either CDSs, KMV credit spreads, or bonds issued by the counterparty.If you have access to a simulation engine (e.g., Libor Market Model, or Hull White) you could simulate the underlying forward curves and evaluate the exposures pathwise at specified times in the future, and then "combine" the expected exposure curve (average over all paths) with the credit curve in the standard way, assuming there is no wrong-way risk (for a simple intro, see, e.g.,
http://papers.ssrn.com/sol3/papers.cfm? ... id=1032522). This method can be used on a portfolio of just about any instruments whose value can be easily computed from the forward curve.It sounds like you are looking for a simpler method, essentially equivalent to an option on a portfolio of assets modeled as diffusions with constant volatilities: each underlying is a standard swap rate with a known volatility (as specified by the swaption market). Is this correct? It sounds like you'd need an approximate volatility for that portfolio, maybe as a weighted sum of the volatilities of the rates of the swaps included in the portfolio. I'm not sure off-hand what such a formula would be, but it seems basic enough that there may be something in Hull's book on this.