It is widely believed that, with the introduction of such straightforward extensions like remote procedure call, distributed systems programming is no different than more classical forms of programming. However, in contrast to most traditional software, the logical aspects of distributed programs are inextricably connected to the physical aspects of the environment in which they execute. The net result can be overwhelming complexity that, in essence, reflects the inevitable complexity of the physical world.
Our only chance of success against such overwhelming odds is to counter-balance this inherent complexity by simplicity of system design.
In this talk, we first examine the essential characteristics of distributed systems that make them so challenging and then describe certain basic architectural patterns that can help us achieve the goal of simplicity. We also show how these patterns can be captured using the industry-standard Unified Modeling Language. '