abstract - LES摘要-莱斯.ppt

The phenomenon ofOpen Distributed Systems … Computing is undergoing a transition from monolithic systems, constructed according to a single overall design, into open and irregular conglomerates of distributed components. The composition of such open system is changing dynamically, and largly unknown. Examples of open systems. Heterogeneous multi-agents systems, like sensor systems. Distributed Enterprises Coalitions of enterprises (e.g., grids, virtual enterprises) … and their Problematics How does one design and built such systems, so that they would operate coherently as a whole? How does one ensure the security and fault tolerance of such systems? How can one reason about such systems? On Reasoning about Systems The conventional approach—according to a 2000 report by SEI of CMU, which views the key challenge facing Component-Based SE, the “holy grail”, they call it, as follows: “…to ensure that the properties of a system of components can be predicted from the properties of the components themselves” . But in open systems the components are unknown so, the real challenge is: to reason about a system without knowing its details. Reasoning about Open Systems We do it all the time We reason about the physical world, relying on the laws of nature; And about the social world, relying on social laws and norms. But, there are virtually no laws regulating distributed software systems. The “Governance” Approach This approach-- introduced in The Imposition of Protocols Over Open Distributed Systems, by Minsky, in IEEE Trans. on SE, 1991. --rests on the following simple observations: In the absence of natural laws of software, we need to formulate artificial laws, like we do for societies. Such laws need to be enforced, by preventing their violations, so they can be as dependable as laws of physics. The enforcement needs to be done in a decentralized fashion, for scalability. Potential Benefits of Laws,and their Pitfalls The benefits They could engender pr