Alibaba Innovative Research (AIR) > Research and Development of Key Technologies, Applications and Systems for Carbon Neutrality
【CCF-AIR青年基金】Collaborative Scheduling and Deployment Mechanisms for Edge-cloud Applications based on Low-power Heterogeneous Processors

Research Themes

Research and Development of Key Technologies, Applications and Systems for Carbon Neutrality


In recent years, with the rapid development of the mobile Internet and the Internet of Things, the total amount of computing devices and data in the world has continued to grow. However, with the rise of various novel applications such as autonomous driving, cloud gaming, live video, AR/VR, etc., the inherent weaknesses of cloud computing in terms of latency, energy consumption, and security and privacy are gradually exposed, affecting the speed of development and evolution of these new applications. Therefore, offloading computing to the edge has become a major trend and also attracts more and more researches from academy and industry.

However, various edge application services such as cloud gaming, machine learning inference, video codec, etc., requires different hardware resource (CPU, GPU, FPGA, Mobile SoC, etc.), heterogeneous hardware architectures (X86, ARM, etc.), and complex edge business advantages (network latency, throughput, cost, energy consumption, etc.). In addition, the edge applications often show a significant tidal phenomenon, resulting in a low overall hardware resource utilization rate. How to take full advantage of the idle computing resources and reduce the task execution cost while ensuring the QoS of the edge tasks has also become a valuable problem.

The above characteristics bring huge challenges to edge computing: in order to make full use of the hardware resources on the edge side, it is necessary to adapt a variety of novel scheduling and deployment mechanisms for diverse applications and heterogeneous hardware.


  • Establish evaluation standards and systems for edge cloud applications on heterogeneous hardware, covering at least three types of typical applications and three types of heterogeneous hardware.
  • Optimize typical edge applications (cloud games, video encoding and decoding, machine learning inference, 5G, etc.) on heterogeneous hardware such as ARM arrays and ARM servers with 30% performance improvement and 50% energy consumption reduction.
  • Build a collaborative scheduling and deployment system for edge applications on heterogeneous hardware, improve the resource utilization rate by 20%.

Related Research Topics

  • Cloud computing
  • Resource allocation
  • Task scheduling
  • Software-hardware codesign
  • Edge-cloud collaboration

Scan QR code
关注Ali TechnologyWechat Account