NeTS:Small:Content Retrieval Networks: Network Access for Disadvantaged Regions

该奖项是根据2009年美国复苏和再投资法案（公法111-5）资助的。即使在美国，许多地区也由于高昂的价格、容量限制、落后的基础设施或低人口密度而遭受互联网连接不良的困扰。由于DSL的距离限制和将光纤电缆引入低密度地区的高成本，农村社区可能得不到充分的服务。即使企业或学校可以通过DSL连接，当许多人共享连接时，其容量限制仍然可能提供糟糕的最终用户体验。 这项研究通过使用智能软件来增强网络连接来解决这些问题。 通过积极的数据缓存和非高峰预取，数据可以存储在磁盘上，并在本地检索，而不是总是从互联网上获取。对于有许多用户的环境，该系统可以将带宽消耗降低一个数量级，提供更高的有效最终用户带宽，或者通过购买更少的广域带宽来降低成本。 这项研究的预期成果是一个开源的系统，可以部署在互联网网关，这将透明地提高网络性能。可能的地点包括学校、小型企业和互联网服务提供商（ISP）。在学校里，当满教室的学生访问一个主题时，只有一个副本被下载，其余的从该高速缓存中获取。对于那些在白天更新的新闻网站，只有修改的部分会被下载，而不是整个白天。预取也将减少延迟，通过确定用户模式和预加载内容之前，它是需要的。

This award is funded under the American Recovery and Reinvestment Act of 2009(Public Law 111-5).Even in the US, many regions suffer poor Internet connectivity, due tohigh prices, capacity limitations, poor infrastructure, or lowpopulation density. Rural communities may be underserved because ofDSL distance limits and the high cost of bringing fiber optic cable tolow-density regions. Even when businesses or schools can be connectedby DSL, its capacity limits may still provide poor end-user experiencewhen many people share the connection. This research is addressing theseproblems by using intelligent software to enhance networkconnectivity. By aggressive data caching and off-peak prefetching,data can be stored on disk and retrieved locally instead of alwaysbeing fetched from the Internet. For environments with many users,this system can reduce bandwidth consumption by an order of magnitude,providing higher effective end-user bandwidth, or allow costreductions by purchasing less wide-area bandwidth. The expectedoutcome of this research is an open-source system that can be deployedat Internet gateways, which will transparently improve networkperformance. Possible locations include schools, small businesses, andInternet Service Providers (ISPs). In schools, when a classroom fullof students accesses a topic, only one copy is downloaded, andthe rest get it from the cache. For news sites that update during theday, only the changed portions get downloaded, instead of wholearticles. Prefetching will also reduce latency, by determining userpatterns and preloading content before it is needed.