Distributed sequential computing uses the collective memory of a network of workstations to reduce paging overhead. In contrast to the "page farm" approach, in which a stationary program remotely accesses data, distributed sequential computing moves the code to the data. In this paper, we show that mobile agents are the most natural and effective way to implement this approach. This is because mobile agents preserve the algorithmic integrity of sequential programs, while a message passing implementation requires a complete restructuring of the code.
