Programming peer-owned open distributed systems

点击翻译按钮获取中文摘要

Peer-owned open distributed systems such as the Internet offer both challenges as well as opportunities forsolving problems.  An important opportunity comes from physical connectivity with virtually limitless amounts of diverse resources, potentially amounting to unbounded computational power; a significant (non-functional)source of difficulty in writing software for such execution environments comes from the uncertainty ofcommunication delays, which leads to challenges in distributed control and coordination.  There is a need to overcome the challenges as well as to have mechanisms for availing of opportunities.  Our approach is to encapsulate distributed computations represented as multi-agent systems, together with theresources that are available to them, in (distributed) ownership encapsulations called cyberorgs.  Theencapsulated resources can include system resources such as processor cycles and network bandwidth, orapplication resources such as access to a user's mailbox or display screen space, each defined in time andspace.  Ownership is guaranteed by contracts negotiated between buyer and seller cyberorgs.   To support engineering of such distributed systems, we separate resource ownership and control concerns fromfunctional concerns, allowing each to be coded separately.  It turns out that this separation can be achieved atboth system as well as application levels, resulting in a wide applicability of our approach.We expect important results to be produced in this work.  On the one hand, this work will enhance ourunderstanding of challenges in using peer-owned open distributed systems as reliable environments forcarrying out distributed computations, and approaches to overcome them.  On the other hand, it will offerprinciples, mechanisms and tools for engineering such systems and evaluating their performance.