High-Throughput Many-Core Processor Design

doi:10.11871/jfdc.issn.2096-742X.2020.01.006

Abstract

Abstract:

[Objective] With the rapid growth of new high-throughput applications such as cloud computing, the Internet of Things, and artificial intelligence, the main applications of high-performance computing have gradually evolved from traditional scientific and engineering computing to emerging data processing, which brought huge challenges to traditional processors. High-throughput many-core processors are becoming a new type of processor architecture dealing with such applications and therefore an important research direction. [Method] In view of the above problems, this paper analyzes the typical characteristics of high-throughput applications, and discusses the key design of high-throughput many-core processors from the three core aspects of data processing, transmission, and storage. The design includes real-time task dynamic scheduling, high-density on-chip network design and on-chip storage hierarchy optimization, etc. [Results] The experimental results show that the above mechanism can effectively ensure the service quality of tasks, improve the data throughput rates, and simplify the on-chip memory hierarchy. [Conclusion] With the urgent demand for high concurrency and strong real-time processing in the era of Internet of Everything, high-throughput many-core processors are expected to become the main processing engine in future data centers.

Key words: high throughput computing, many-core processor, data path

Ye Xiaochun,Li Wenming,Zhang Yang,Zhang Hao,Wang Da,Fan Dongrui. High-Throughput Many-Core Processor Design[J]. Frontiers of Data and Computing, 2020, 2(1): 70-84.

Figures/Tables 21

Table 1

Fig.1

Fig. 2

Table 2

Fig. 3

Table 3

Fig.4

Fig.5

Fig.6

Fig.7

Fig.8

Fig.9

Fig.10

Fig.11

Fig.12

Fig.13

Fig.14

Fig.15

Fig.16

Table 4

Fig.17

References 22

[1]	IDC & DataAge.IDC: 2025 年中国将拥有全球最大的数据圈[EB/OL]. .
[2]	中国市场调研在线. 2017-2023年中国互联网数据中心( IDC)行业研究及投资前景预测报告[R]. 2018.
[3]	Netlib. LAPACK: Linear Algebra PACKage[EB/OL]. .
[4]	范东睿, 叶笑春, 包云岗, 孙凝晖 . 中国高通量计算机的自主研发之路[J]. 中国科学院院刊, 2019,34(6):648-656.
[5]	Ninghui Sun, Yungang Bao, Dongrui Fan . The rise of high-throughput computing [J]. October 2018 Frontiers of Information Technology & Electronic Engineering 19( 10):1245-1250.
[6]	李国杰 . 大数据对计算机系统的挑战[J]. 中国计算机学会通讯. 2013,9(12):33-35.
[7]	孙凝晖 . 计算机系统的技术挑战[R]. 2010, 杭州.
[8]	Sandeep Agrawal, Valentin Pistol, Jun Pang, John Tran, David Tarjan, Alvin Lebeck. Rhythm: harnessing data parallel hardware for server workloads [C]. In Proceedings on Architectural support for programming languages and operating systems (ASPLOS), 2014, 19-34.
[9]	Michael Ferdman, Almutaz Adileh, Onur Kocberber, Stavros Volos, Mohammad Alisafaee, Djordje Jevdjic, Cansu Kaynak, Adrian Popescu, Anastasia Ailamaki, Babak Falsafi . Clearing the clouds: a study of emerging scale-out workloads on modern hardware[C]. In Proceedings of the International Conference on Architectural Support for Programming Languages and Operating Systems ( ASPLOS), 2012.
[10]	Pejman Lotfi-Kamran, Boris Grot, Michael Ferdman, Stavros Volos, Onur Kocberber, Javier Picorel, Almutaz Adileh, Djordje Jevdjic, Sachin Idgunji, Emre Ozer, Babak Falsafi . Scale-out processors[C]. In Proceedings of the International Symposium on Computer Architecture ( ISCA), 2012.
[11]	Pejman Lotfi-Kamran, Boris Grot, Babak Falsafi. 2012. NOC-Out: Microarchitecting a Scale-Out Processor [C]. In Proceedings of the 2012 45th Annual IEEE/ACM International Symposium on Microarchitecture (MICRO). IEEE Computer Society, USA, 177-187.
[12]	Nikos Hardavellas, Michael Ferdman, Babak Falsafi, Anastasia Ailamaki. Toward Dark Silicon in Servers [C]. In IEEE Micro, vol. 31, no. 4, pp. 6-15, July-Aug. 2011.
[13]	Jongse Park, Hardik Sharma, Divya Mahajan, Joon Kyung Kim, Preston Olds, Hadi Esmaeilzadeh . Scale-out acceleration for machine learning[C]. In Proceedings of the 50th Annual IEEE/ACM International Symposium on Microarchitecture (MICRO). Association for Computing Machinery, New York, NY, USA, 367- 381. 2017.
[14]	Akshitha Sriraman, Abhishek Dhanotia, Thomas F . Wenisch. SoftSKU: optimizing server architectures for microservice diversity @scale[C]. In Proceedings of the 46th International Symposium on Computer Architecture (ISCA). New York, NY, USA, 513- 526. 2019.
[15]	Artemiy Margaritov, Siddharth Gupta, Rekai Gonzalez-Alberquilla, Boris Grot . Stretch: Balancing QoS and Throughput for Colocated Server Workloads on SMT Cores[C]. IEEE International Symposium on High Performance Computer Architecture (HPCA), Washington, DC, USA, 15- 27. 2019
[16]	Cansu Kaynak, Boris Grot, Babak Falsafi . Confluence: unified instruction supply for scale-out servers[C]. In Proceedings of the 48th International Symposium on Microarchitecture (MICRO). ACM, New York, NY, USA, 166- 177. 2015.
[17]	Christos Kozyrakis, Aman Kansal, Sriram Sankar, Kushagra Vaid. Server Engineering Insights for Large-Scale Online Services [C]. In IEEE Micro, vol. 30, no. 4, pp. 8-19, July-Aug. 2010.
[18]	Steven Cameron Woo, Moriyoshi Ohara, Evan Torrie, Jaswinder Pal Singh, Anoop Gupta. The SPLASH-2 Programs: Characterization and Methodological Considerations [C]. International Symposium on Computer Architecture (ISCA), Santa Margherita Ligure, Italy, 1995.
[19]	Dongrui Fan, Wenming Li, Xiaochun Ye, Da Wang, Hao Zhang, Zhiming Tang, Ninghui Sun. SmarCo: An Efficient Many-Core Processor for High-Throughput Applications in Datacenters [C]. 2018 IEEE International Symposium on High Performance Computer Architecture (HPCA), Vienna, 2018, pp. 596-607.
[20]	Luiz Andre Barroso, Jimmy Clidaras, Urs Holzle . The datacenter as a computer: An introduction to the design of warehouse-scale machines[J]. Synjournal Lectures on Computer Architecture, 8(3):1~ 154, 2013.
[21]	包云岗, 范东睿, 陈明宇, 徐志伟, 孙凝晖 . 计算机即网络理念与高通量计算[J]. 中国计算机学会通讯, 2016,12(3):10-16.
[22]	Jorda Polo, David Carrera, Yolanda Becerra, Malgorzata Steinder, Ian Whalley. Performance-driven task co-scheduling for mapreduce environments [C]. In Network Operations and Management Symposium (NOMS), IEEE, 2010, pp. 373-380.

	典型应用	程序特征	设计目标
传统高性能计算	科学与工程计算应用	任务单一,负载变化小,计算量大,计算局部性好	高速度（算得快）
新型高通量计算	互联网数据中心应用	任务多样,流式计算特征,数据量大,实时性要求高	高通量（算得多）

	共性特征	结构对比
高通量计算结构	规模庞大实时调度需求访问不规则流量压力大	存储层次优化	全局实时调度	低宽度高密度网络
城市交通结构	规模庞大实时调度需求访问不规则流量压力大	候车空间导流 /快速通道	交通指挥控制中心	高密度路网

配置项	参数值
任务表项	4096项（对应一个子环）
每个子环上的核数	16*2路SMT
每个核上的调度任务表项	256项
每个核可调度的任务数N_TASK	128个,所有任务等长
任务DEADLINE	任务长度*F,通过调节F的大小（64-128）来调节任务的deadline紧迫程度

Cache层	私有/共享	容量大小	行大小
L1	私有	32K	64 Bytes
L2	私有/共享	256K	64 Bytes
L3	共享	2048K	64 Bytes