In this thesis the focus will be on the speed of light, and how one may reduce its impact on performance. For each remote invocation, there is an associated overhead that corresponds to the latency between the respective nodes. If latency is high and packet-exchange frequent, there is potential for performance degradation. The incorporation of our graph caching scheme for the near-far cloud architecture aims to reduce the the amount of remote invocations. It does so by taking advantage of key mechanisms provided by the distributed-objects programming language known as Emerald.