ITR: Exploiting Remote Infrastructure for Mobile Information Access

研究人员建议设计、实施、分析和评估机制，使移动客户端能够机会主义地利用固定基础设施中的计算、数据和通信资源，以促进信息访问。移动客户端通常面临许多资源挑战。这些问题包括无线网络质量的不可预测变化；远程服务可用性的巨大差异；重量和大小对CPU功率、内存大小和磁盘容量的限制；以及对电池功耗的担忧。我们的目标是利用远程基础设施，如临时缓存站点或计算服务器，来克服这些挑战。当这样的基础设施不可用，或者在电池功率或带宽方面太昂贵或太慢时，研究人员希望能够仅依靠当地资源继续运行。挑战在于提出可以在资源有限的移动客户端上有效实施的自适应机制、算法和策略，但从用户的角度来看，这些机制、算法和策略又可以平稳、无缝地运行。这项研究将包括实验和分析。一方面，研究人员计划设计、实施和评估包含上述挑战的创新解决方案的移动计算系统。这将给研究人员提供实践经验和洞察力，以阐明该领域的合理设计原则。与此同时，研究人员计划进行分析和建模工作，以更深入地了解适应中的基本权衡，并探索比实验可行的设计空间更广阔的区域。这项工作将被组织为三个主要推力：1.移动客户端的高效文件缓存管理2.为节省能量和带宽而卸载计算3.在第一个推力中，研究人员计划探索使用基础设施中的中间缓存站点来提高移动文件系统中缓存管理的效率。缓存对于移动信息访问的重要性得到了广泛的认可，但它在未来的移动系统中应该采取什么具体形式仍然是一个悬而未决的问题。在第二个推力中，研究人员将探索移动客户端的技术，以自适应地将计算卸载到基础设施中的服务器上。这种卸载对于改进性能以及延长电池寿命可能是有用的。在第三个推力中，研究人员将对其他推力中开发的机制的广泛政策进行分析性调查。未来的移动计算系统可能会在环境条件、用户偏好和应用程序特征方面遇到相当大的变化。对这一巨大的多维空间进行分析表征，并了解其对替代政策的影响，将对开发成功的系统至关重要。如果成功，这项研究将使移动计算离现实更近一步。虽然移动计算系统的商业部署对工业来说是一个巨大的潜在市场，但仍有许多技术难题有待解决。业界将今天广泛部署的商业现成系统扩展到移动环境的渐进式努力不会解决这些问题。研究人员提出了努力，结合对一组具有挑战性的关键问题的实验和分析，从而代表了一项高风险、高回报的冒险。

The researchers propose to design, implement, analyze and evaluate mechanisms that will enable mobile clients to opportunistically exploit computing, data and communication resources in the fixed infrastructure to facilitate information access. A mobile client typically faces many resource challenges. These include unpredictable variation in wireless network quality; wide disparity in the availability of remote services; limitations imposed by weight and size constraints on CPU power, memory size and disk capacity; and concern for battery power consumption. The goal is to exploit remote infrastructure, such as transient caching sites or compute servers, to overcome these challenges. When such infrastructure is unavailable or would be too expensive or slow to access in terms of battery power or bandwidth, the researchers wish to be able to continue operation relying solely on local resources. The challenge is to come up with the adaptive mechanisms, algorithms and policies that can be effectively implemented on a resource-limited mobile client, yet function smoothly and seamlessly from a user's viewpoint. The research will span both experimentation and analysis. On the one hand, the researchers plan to design, implement, and evaluate mobile computing systems embodying innovative solutions to the challenges described above. This will give the researchers the hands-on experience and insights needed to elucidate sound design principles for this domain. In parallel, the researchers plan to conduct an analytical and modelling effort to obtain a deeper understanding of fundamental tradeoffs in adaptation, and to explore a much broader region of the design space than feasible experimentally. This effort will be organized into three major thrusts: 1.Efficient file cache management for mobile clients 2.Offloading computation for energy and bandwidth savings 3.Modelling and analysis of adaptation tradeoffs In the first thrust, the researchers plan to explore the use of intermediate caching sites in the infrastructure to improve the efficiency of cache management in mobile file systems. The importance of caching for mobile information access is widely acknowledged, but what specific form it should take in the mobile systems of the future remains an open question. In the second thrust, the researchers will explore techniques for mobile clients to adaptatively offload computation on servers in the infrastructure. Such offloading may be useful for improved performance as well as enhanced battery life. In the third thrust, the researchers will analytically investigate a wide range of policies for the mechanisms developed in the other thrusts. The mobile computing systems of the future are likely to encounter considerable variability in environmental conditions, user preferences, and application characteristics. Characterizing this large multi-dimensional space analytically, and understanding its implications for alternative policies will be critical to the development of successful systems. If successful, the research will bring mobile computing one step closer to reality. Although commercial deployment of mobile computing systems is an enormous potential market for industry, many difficult technical problems remain to be solved. Incremental efforts by industry to extend today's widely-deployed commercial off-the-shelf systems to mobile environments will not solve these problems. The researchers proposed effort, combining experimentation and analysis on a key set of challenging problems, thus represents a high-risk, high-payoff venture.