Outage-constrained service capacity with stochastic resource harvesting

Providing a uniform and non-disrupted service is of paramount importance for numerous applications. However, such a goal becomes challenging when the resource for providing service is supplied by a source which is stochastic in nature; examples include a media player whose content is received from a remote database via a unreliable network connection, or a base station whose power supply is from a renewable energy source which harvests energy. Counting the time epochs at which the stored resource is insufficient for supplying the target service utility as outage, a general problem framework of characterizing the outage-constrained service capacity, which is the maximum constant service utility subject to a constraint on the outage rate, is formulated. The outage rate, namely the frequency of outage occurrence, is shown to be closely connected with the mean value of the first ladder epoch of an induced random walk, and consequently a lower bound on the outage-constrained service capacity is established. It is observed that, asymptotically zero outage rate is achievable whenever the resource expenditure does not exceed the expected resource harvesting rate, and when allowing a nonzero outage rate, it is possible to have the resource expenditure strictly exceed the expected resource harvesting rate. Finally, a discussion is given on the possibility and challenge of extending the analysis to establish the information capacity of a channel with energy harvesting, which hinges upon a conjecture on the asymptotic behavior of a channel with input substitution by an adversary.