Similar articles

Abstract: In recent years, the advent of emerging computing applications, such as cloud computing, artificial intelligence, and the Internet of Things, has led to three common requirements in computer system design: high utilization, high throughput, and low latency. Herein, these are referred to as the requirements of ‘high-throughput computing (HTC)’. We further propose a new indicator called ‘sysentropy’ for measuring the degree of chaos and uncertainty within a computer system. We argue that unlike the designs of traditional computing systems that pursue high performance and low power consumption, HTC should aim at achieving low sysentropy. However, from the perspective of computer architecture, HTC faces two major challenges that relate to (1) the full exploitation of the application’s data parallelism and execution concurrency to achieve high throughput, and (2) the achievement of low latency, even in the cases at which severe contention occurs in data paths with high utilization. To overcome these two challenges, we introduce two techniques: on-chip data flow architecture and labeled von Neumann architecture. We build two prototypes that can achieve high throughput and low latency, thereby significantly reducing sysentropy.

高通量计算的兴起

[1]Bao YG, Wang S, 2017. Labeled von Neumann architecture for software-defined cloud. J Comput Sci Technol, 32(2):219-223.

[2]Barroso LA, Clidaras J, Hölzle U, 2013. The Datacenter as a Computer: an Introduction to the Design of Warehouse-Scale Machines (2^nd Ed.). Morgan & Claypool Publishers, USA.

[3]Dennis JB, Fosseen JB, Linderman JP, 1974. Data flow schemas. Int Symp on Theoretical Programming, p.187-216.

[4]Fan DR, Li WM, Ye XC, et al., 2018. SmarCo: an efficient many-core processor for high-throughput applications in datacenters. IEEE Int Symp on High Performance Computer Architecture, p.596-607.

[5]Greenberg A, 2015. SDN for the cloud. Keynote of SIGCOMM.

[6]Kapoor R, Porter G, Tewari M, et al., 2012. Chronos: predictable low latency for data center applications. Proc 3^rd ACM Symp on Cloud Computing, p.9.

[7]Kasture H, Sanchez D, 2016. Tailbench: a benchmark suite and evaluation methodology for latency-critical applications. IEEE Int Symp on Workload Characterization, p.1-10.

[8]Livny M, 2013. From Principles to Capabilities—the Birth and Evolution of High Throughput Computing. Wisconsin Institutes for Discovery, University of Madison-Wisconsin, USA.

[9]Ma JY, Wang HB, Zhang LX, et al., 2015. Supporting differentiated services in computers via programmable architecture for resourcing-on-demand (PARD). ACM SIGPLAN Not, 50(4):131-143.

[10]Xie Y, 2016. Technology-driven architecture innovation: challenges and opportunities. Proc 43^rd ACM/IEEE Int Symp on Computer Architecture, Architecture 2030 Workshop.

[11]Xu ZW, 2012. How much power is needed for a billion-thread high-throughput server? Front Comput Sci, 6(4):339-346.

[12]Xu ZW, Li CD, 2017. Low-entropy cloud computing systems. Sci Sin Inform, 47(9):1149-1163.

[13]Xu ZW, Chi XB, Xiao N, 2016. High-performance computing environment: a review of twenty years of experiments in China. Nat Sci Rev, 3(1):36-48.

[14]Yu Z, 2017. Labeled RISC-V: a new perspective on software-defined architecture. Proc 6^th RISC-V} Workshop, p.7.