MOCHA

The Modelling and Optimising Complex Heterogeneous Architecture (MOCHA) project at the University of York aims to equip the Huawei 5G base stations with increased real-time capabilities (i.e., the response time can be bounded in the worst case), reduce latency and improve throughput, to serve the emerging applications in the domains of industrial automation, highly automated driving, and robotics under ultra-reliable low-latency communication (uRLLC) mixed with other types of traffic. The keys to achieve these objectives lie on both the scheduling and memory levels.

Considering a multi-core architecture as is in the Huawei 5G base station, resource management has the central role. Simple scheduling heuristics to keep the cores busy may lead to long average- and worst-case latency. The project proposes scheduling and allocation mechanisms towards more efficient resource utilisation and improvement on the worst case, together with static analysis to tightly bound the response time, as well as digital twin to explore the system and perform statistical evaluation.

This research project is funded by Huawei Technologies Co., Ltd with a funding value of £985,927, covering three years from December 2019 to November 2022.

Research Roadmap

The MOCHA project has a wide range of research novelties in the context of multi-/many-core real-time sytsems research:

• High performance and heterogeneous embedded architecture
• Digital twinning and high speed simulation
• High-level system and cache modelling
• Multi-DAG scheduling and analysis
• Memory- and cache-aware scheduling and allocation
• Analysis based on statistical data
• Empirical studies on real-world applications

MOCHA Toolchain

The MOCHA Toolchain (MOCHA-T) aims to provide a collection of simulation and analysis techniques based on the concept of Digital Twin to facilitate multi- /many-core scheduling research.

The MOCHA-T includes the following modules:

• Response time analysis (RTA) for multi-DAGs [2][4]
• Digital Twin for run-time model refinement [3][5]
• DAG workload generator (GitHub)
• Multi-core scheduling and allocation simulator [6]
• Data-driven execution time profiler [7]
• Dynamic analysis using probabilistic and statistical approaches
• Timing accurate general purpose I/O framework [1]

Project Members

• Prof. Iain J. Bate (Principal Investigator), University of York
• Dr. Wanli Chang (Co-Investigator), University of York
• Prof. Alan Burns (Advisor), University of York
• Dr. Xiaotian Dai (Postdoc RA), University of York
• Dr. Shuai Zhao (Postdoc RA), University of York
• Dr. Benjamin Lesage (Postdoc RA), University of York

Presentations

• (June, 2023) Presented two papers at RTNS 2023 conference
• (Oct, 2022) Presented at ISoLA 2022 conference
• (Nov, 2021) Presented at UK Systems Research Challenges 2021 workshop, UK [slides]
• (May, 2021) Presented at RTAS 2021 conference
• (Dec, 2020) Presented at UK Systems Research Challenges 2020 workshop, UK [slides]
• (Dec, 2020) Presented at RTSS 2020 conference [slides] [video]
• (Oct, 2020) Presented at DAC 2020 conference

Publications

1. Shuai Zhao, Zhe Jiang, Xiaotian Dai, Iain Bate, Ibrahim Habli, Wanli Chang. “Timing-Accurate General-Purpose I/O for Multi- and Many-Core Systems: Scheduling and Hardware Support”. Design Automation Conference (DAC). 2020
2. Shuai Zhao, Xiaotian Dai, Iain Bate, Alan Burns, Wanli Chang. “DAG Scheduling and Analysis on Multiprocessor Systems: Exploitation of Parallelism and Dependency”. In IEEE Real-Time Systems Symposium (RTSS). 2020
3. Xiaotian Dai, Shuai Zhao, Iain Bate, Alan Burns, Xing Guo, Wanli Chang. “Brief Industry Paper: Digital Twin for Dependable Multi-Core Real-Time Systems—Requirements and Open Challenges”. IEEE Real-Time and Embedded Technology and Applications Symposium (RTAS). 2021
4. Shuai Zhao, Xiaotian Dai, Iain Bate. “DAG Scheduling and Analysis on Multi-core Systems by Modelling Parallelism and Dependency”. IEEE Transactions on Parallel and Distributed Systems (TPDS). 2022
5. Xiaotian Dai, Shuai Zhao, Benjamin Lesage, Iain Bate. “Using Digital Twins in the Development of Complex Dependable Real-Time Embedded Systems”. 11th International Symposium On Leveraging Applications of Formal Methods, Verification and Validation (ISoLA). 2022
6. Shuai Zhao, Xiaotian Dai, Benjamin Lesage, Iain Bate. “Cache-Aware Allocation of Parallel Jobs on Multi-cores based on Learned Recency”. 31st International Conference on Real-Time Networks and Systems (RTNS). 2023
7. Benjamin Lesage, Xiaotian Dai, Shuai Zhao, Iain Bate. “Reducing Loss of Service for Mixed-Criticality Systems through Cache-and Stress-Aware Scheduling”. 31st International Conference on Real-Time Networks and Systems (RTNS). 2023