CSR: Small: Collaborative Research: Taming Mobile Hardware & OS Diversity for Comprehensive Software Analysis

如今，在智能手机、平板电脑和可穿戴设备等移动系统上运行的Android操作系统越来越多样化。这主要是因为Android是由许多不同的供应商和移动运营商在多个层面上进行大量定制的，原因包括硬件和品牌特定增值应用和服务的差异。Android的多样性给运行在这些移动系统上的软件的测试、调试和分析带来了重大挑战，包括跨不同系统测试一致的功能和性能、查找崩溃漏洞、分析恶意软件和识别安全漏洞。这个项目的目标是使运行Android的移动系统的整个软件栈的测试和分析更快，更实惠，同时需要小的工程努力，尽管移动软件和硬件的多样性。拟议的项目将有助于提高移动系统软件的质量和安全性，这对社会和经济都有重大好处。计划通过基于云的移动农场服务实现上述目标，该服务提供虚拟移动实例，或简称为虚拟实例。每个虚拟实例都是运行在功能强大的服务器上的虚拟机，完全类似于特定移动系统的硬件和软件配置。通过利用在服务器中运行的虚拟机，建议的解决方案将提供低成本，从而降低用户的价格。这些重要的挑战将在两个研究重点中得到解决。第一部分研究了在云中整合真实移动系统以降低整体基础设施成本的技术。第二个研究重点是通过利用软件和硬件技术（如推测、记录和重播、动态边界和远程直接内存访问（RDMA））来提高虚拟实例的性能。

Today, the Android operating system running on mobile systems such as smartphones, tablets, and wearables, is increasingly diverse. This is mainly because Android is heavily customized at multiple layers by many different vendors and mobile carriers, for various reasons including the difference in hardware and brand-specific value-added applications and services. This diversity of Android creates important challenges for testing, debugging, and analysis of software running on these mobile systems, including testing for consistent functionality and performance across different systems, finding crash bugs, analyzing malware, and identifying security vulnerabilities. The goal of this project is to make testing and analysis of the entire software stack of mobile systems running Android faster and more affordable, while at the same time requiring small engineering effort, all despite the diversity of mobile software and hardware. The proposed project will help improve the quality and security of software on mobile systems - a significant benefit to society and the economy. The plan to achieve the aforementioned goals through a cloud-based mobile farm service, which provides virtual mobile instances, or virtual instances for short. Each virtual instance is a virtual machine that runs on powerful servers and that fully resembles the hardware and software configurations of a particular mobile system. By leveraging virtual machines running in servers, the proposed solution will provide a low cost, thus lower prices for the users. The important challenges will be addressed in two research thrusts. The first thrust investigates techniques to consolidate the real mobile systems in the cloud to reduce the overall infrastructure cost. The second research thrust targets at enhancing the performance of virtual instances by leveraging software and hardware techniques such as speculation, record and replay, dynamic boundary, and Remote Direct Memory Access (RDMA).