CAREER:In-Network Computation Meets Data Persistence

在当前这个以数据为中心的时代，由社交媒体、视频共享应用、传感器群、自动汽车等产生的数据正呈指数级增长。不幸的是，随着技术扩展速度的放缓，半导体行业在处理如此大的数据集时提供更好的性能方面面临着重大挑战。因此，必须重新设计当前的系统，以维持对指数增长数据集的计算需求。然而，计算的基本模型几十年来没有改变。在我们目前的冯·诺依曼模型中，数据位于较慢的持久存储器中，并被来回移动到更快的内存中，以便处理器进行计算。该项目的重点是实现数据在未来系统中存储和处理方式的范式转变。数据应该在靠近数据生成的地方处理，数据应该在使用它的地方持久化。这项工作的目标是从根本上重新思考如何在数据中心维护持久化状态，并在网络中公开持久化域，以获得更好的性能、吞吐量和尾部延迟。这项工作重新设计了可编程交换机和智能网络接口卡（Smart Network Interface Card，简称NIC），以操纵网络中的持久数据。该项目能够更快地执行操纵全局持久状态的应用程序，例如全局规模的深度学习训练。机器学习模型具有巨大的社会影响，它们被用于预测癌症、流行病等，这项研究将可能影响许多科学领域，从而实现新的创新。在这项研究中开发的数据集，工具和技术将公开提供，以造福整个社区。该项目的教育部分将侧重于通过暑期项目吸引本科生和高中生对科学和技术的兴趣。该奖项反映了NSF的法定使命，并通过使用基金会的智力价值和更广泛的影响审查标准进行评估，被认为值得支持。

In this current data-centric era, data generated by social media, video sharing applications, swarms of sensors, autonomous cars, etc. is growing exponentially. Unfortunately, as the technology scaling slows down, the semiconductor industry has been facing a major challenge in providing better performance while processing such large datasets. As a result, it is essential to redesign current systems to sustain the demand for computing over exponentially growing datasets. However, the fundamental model of computing has not changed over many decades. In our current Von Neumann model, data sits in the slower persistent storage and it is moved back and forth to faster memory for computation by the processor. This project focuses on enabling a paradigm shift in how data is stored and processed in future systems. Data should be processed close to where data is generated and data should be persisted where it is used. The goal of this work is to fundamentally rethink how persistent states are maintained in data centers and expose the persistency domain in the network for better performance, throughput, and tail latency. This work redesigns programmable switches and smart network interface cards (NICs) to manipulate persistent data in the network.This project enables faster execution of applications that manipulate global persistent state, such as global-scale deep learning training. Machine learning models have a huge societal impact and they are being used to predict cancers, epidemics, etc. This research will potentially impact numerous scientific fields to take a leap forward towards new innovations. The datasets, tools, and techniques developed in this research will be made publicly available to benefit the whole community. The educational part of this project will focus on attracting undergraduate and high-school students towards science and technology through a summer program.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.