Support for Scalable Transactional Applications on the Global Internet

Luiba ShriraCCR-9901699 Brandeis大学合作缓存技术已成功用于提高分布式文件系统和Web的性能。 这些技术通过用更有效地访问“附近”的缓存副本来代替对“远距离”数据的更昂贵的访问，并通过协调缓存的内容，提高了系统的整体性能和可扩展性。 所提出的工作的一个主要目标是设计，实施和评估合作的缓存技术的事务对象存储系统。 这些技术将使广域存储系统能够为应用提供事务的可靠性和一致性保证，灵活性是用户定义的对象，并且仍然是可扩展的。本工作的主要贡献将是一组新的广域环境下的事务合作缓存一致性协议，适应对象粒度的全局缓存协调协议，以及保护协作缓存中的对象的安全技术。 该技术将被纳入雷神物体存储系统。 这项工作包括早期的研究和开发以及原型设计。

Luiba ShriraCCR- 9901699Brandeis UniversityCooperative caching techniques have been successfully used to improve the performance of distributed file systems and the Web. By replacing more expensive access to a "far away" data with a more efficient access to a "near by" cached copy, and by coordinating the content of caches, these techniques improve the overall system performance and scalability.Cooperative caching techniques have bot been used for transactional object systems. A principal goal of the proposed work is to design, implement and evaluate cooperative caching techniques for transactional object storage systems. These techniques will enable construction of wide-area storage systems that will provide applications with the reliability and consistency guarantees of transactions, the flexibility is user-defined objects, and still be scalable.A major contribution of this work will be a set of new transactional cooperative cache consistency protocols for wide-area environment, global cache coordination protocols that adapt to object granularity, and security techniques that protect objects in cooperative caches. The technique will be incorporated into the Thor object storage system. This work encompasses both early research and development and prototyping.