CNS Core: Small: Managed Languages: From Non-volatile Memory to Persistence

不久，计算机主存储器可能会使用新技术——非易失性存储器（NVM）。当电源断开时，当前存储器的内容将丢失，而NVM将保留其内容。然而，计算机在处理器中保存了一些最近的信息。这些数据仍然会丢失。因此，NVM本身并不能使用户准确地返回到原来的位置。为了使完全恢复成为可能，程序需要谨慎地使用NVM。对于现代编程语言（如Java和Python）来说，做好这一点很难。这个项目将为这些语言的NVM提供一个通用的支持基础。非易失性的干净语言语义称为持久性。托管语言（带有垃圾收集和线程的语言）需要适当的持久化语义和健壮的性能实现。在存在线程和垃圾收集的情况下添加干净的持久性是很困难的，而现代多核机器由于其较弱的内存一致性模型而带来了额外的挑战。该项目将扩展一个具有灵活持久化语义的独立于语言的虚拟机。这将有助于避免让每个语言项目从头开始构建自己的支持。项目的方向是在容器和版本化对象的持久堆上支持事务。该项目将建立在Mu （http://microvm.org）上，这是一个现有的具有商业利益的虚拟机。持久性虚拟机将节省开发工作，并鼓励在应用程序中创造性地使用持久性。这将有助于软件的创新、可靠性和可信赖性。项目团队将继续从代表性不足的群体中招募学生，包括指导未来的申请人，以及访问历史上的少数族裔机构。该项目的产品将驻留在https://github.com/umass-moss-lab/persistent-mu的存储库中，该存储库将在项目结束后至少三年内继续可用。存储库将包含已发布的代码、数据等。非代码材料将在知识共享许可下提供，代码将在Apache开源许可下提供，并受马萨诸塞大学知识产权政策的约束。该奖项反映了美国国家科学基金会的法定使命，并通过使用基金会的知识价值和更广泛的影响审查标准进行评估，被认为值得支持。

Soon, computer main memories will likely use new technology, non-volatile memory (NVM). While current memories lose their contents when power is removed, NVM will retain its contents. However, computers keep some recent information inside the processor. That data will still be lost. So, NVM will not by itself enable users to get back to exactly where they were. To make full recovery possible, programs will need to use NVM carefully. Doing this well for modern programming languages (such as Java and Python) is hard. This project will provide a common base of support for NVM for such languages.Clean language semantics for non-volatility is called persistence. Managed languages (ones with garbage collection and threads) require both appropriate persistence semantics and a robust performant implementation. Adding clean persistence in the presence of threads and garbage collection is hard, and modern multi-core machines present additional challenge because of their weak memory consistency models. The project will extend a language-independent virtual machine with flexible persistence semantics. This will help avoid having every language project build its own support from scratch. The project direction is to support transactions over a persistent heap of containers and versioned objects.The project will build on Mu (http://microvm.org), an existing virtual machine that has commercial interest. A persistent virtual machine will save development effort and encourage creative use of persistence in applications. This will contribute to innovation, reliability, and trustworthiness of software. The project team will continue to engage in recruiting students from under-represented groups, including mentoring of future applicants and visits to historically minority institutions.The project's products will reside in a repository at https://github.com/umass-moss-lab/persistent-mu, which will continue to be available for at least three years after the project concludes. The repository will contain released code, data, etc. Non-code materials will be available under the Creative Commons license, and code under the Apache open source license, subject to the University of Massachusetts policy on intellectual property.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.