Runtime Systems for Large-Scale Software Services

点击翻译按钮获取中文摘要

Runtime systems are the engines that power modern IT services associated with economic growth in the near- and medium-term future, such as big data, finance, web services, or high-performance computing. Especially in the context of cloud computing, runtime systems are often layered, which exposes performance and robustness deficiencies. This research program is concerned with studying the software design of runtime systems in terms of simplicity and efficiency. Simplicity implies robustness (which is related to security), while resource-efficiency is a precondition for high performance and green computing. The general objectives of this research program are increasing the robustness, efficiency, and performance of large-scale software services. These objectives address key challenges still associated with large-scale computing. The proposed research seeks to investigate existing runtime systems and to propose novel approaches for the software design of next generation runtime systems. The methodology being used includes developing software prototypes that are being published as open-source projects and the systematic experimental evaluation of alternative software designs. In addition to evaluating the novel proposals discussed in this grant application, the research program will also improve the canonical knowledge of software design as it pertains to scalable performance. In practical terms, the proliferation of results through open source software can directly incubate academic research and teaching, as well as industrial development projects. Many of the improvements sought with this research will be only indirectly visible to end users. Systems will run faster, services operate smoother, and hopefully fail less. At the same time, services can be implemented and provided with fewer investments in special-purpose hardware, thus less capital expenses, and operated with less energy input. This aspect is important from a business point of view (reduced operating expenses), but also from a general societal point of view. Data centres consume significant amounts of electricity and an increase in efficiency will likely lead to a decrease in energy consumption, along with the corresponding benefits for the environment. The proposed research will benefit the Canadian IT industry, as well as any industry using large-scale software services, by training HQP, disseminating software artifacts, and contributing to the expertise that is locally available for knowledge transfer. Furthermore, the environmental benefits from reduced energy consumption for large-scale computing services will be beneficial to society in general.