| Over the past two decades, web technologies have evolved dramatically and have changed what we see and how we interact with the web. The web browser today works akin to a software distribution platform, running web applications written in JavaScript and rendered with HTML and CSS. Web services, ranging from online social networks, to video hubs, to e-commerce sites, enable content sharing for billions of people on an unprecedented scale. However, despite these massive changes, the web is still implicitly built around the premise of a client-server architecture, where web browsers are clients and service providers play the server role. Therefore, popular service providers face significant monetary burdens, but frequently keep their services free for users. Instead, the service providers often monetize the content that users upload to support advertising. As a result, the traditional client-server architecture of the web that is widely used today has significant implications for the privacy, security, and economics of the entire web ecosystem.;In this thesis, we aim to rethink the basic architecture of the web by leveraging the advances in web browsers and cloud services. This thesis focuses on three fundamental aspects of web systems design: first, how content is being served? We address low-cost, scalable content delivery problem with Maygh, a system that builds a content distribution network from client web browsers, without the need for additional plug-ins or client-side software. The result is an organically scalable system that distributes the cost of serving web content across the users of a website. Second, can we provide users with greater control and privacy over their data? We give users control of their data and allow data to be shared by introducing Priv.io, a new approach to building web-based services that let users take full control of their data. Priv.io composes web services with user-provided storage and stitches applications together using web browsers for computation. And third, how can users run long-lived computations while their web browsers may go on and offline? To address the problem, we develop Picocenter, a hosting infrastructure that supports long-lived, mostly-idle applications in cloud environments. As a result, Picocenter enables new types of applications that give users better control over their data while maintaining low running expenses. Taken together, the techniques in this thesis relax the strict client--server architecture of the web to allow end users to contribute resources, thereby better matching today's workloads and enhancing user privacy. |
