Skip to main content
← Back to research

M.Eng. thesis · Systems and learning

Open Resource-as-a-Service

Cooperative orchestration across heterogeneous compute and storage resources

Final thesis title
开放资源即服务
Degree
M.Eng. in Information and Communication Engineering · 2024
Institution
University of Electronic Science and Technology of China (UESTC)

Research question

How can heterogeneous compute and storage resources—distributed across different devices, locations, and providers—be exposed as a cooperative service without binding an application to one machine or one cloud?

Approach

The work combines a resource-service architecture with a learning-based scheduling model. OpenRaaS provides the system view; FogCom isolates the leader–follower resource-selection problem so that it can be studied under partial network observability.

From architecture to decision making

  1. 01

    Separate the resource roles

    The OpenRaaS architecture separates an application’s runtime environment, persistent files, and rendering or computation so that each role can be placed on a suitable node.

  2. 02

    Coordinate heterogeneous infrastructure

    A global coordinator combines compute, file-storage, and image-layer repository nodes to serve a request while balancing user-facing quality and provider-side resource use.

  3. 03

    Learn under partial observation

    FogCom abstracts the scheduling problem into leader and follower decisions. A PPO policy filters storage candidates, while the selected compute provider makes the final choice using information unavailable to the leader.

Interactive architecture · OpenRaaS

Assemble the resource path

Rotate the network nodes until compute, storage, and image resources all reach the coordinator. The puzzle turns the architectural roles above into one small orchestration problem.

Explore the full playground
OpenRaaS Mesh Resource orchestration · 1–5 min Open standalone

This network puzzle needs a little more screen width to remain playable. Open standalone

Public implementation trail

These repositories are public artifacts from the same research trajectory. They clarify the architecture and simulation model; they are not presented as a claim that every thesis experiment is reproduced by the current repository state.

Decentralized container-based resource-service platform

OpenRaaS

The collaborative system implementation, maintained in Zhuobin Huang’s GitHub account. Its public contributor list includes Yuhang Shen.

Open repository ↗ Leader–follower scheduling research prototype

FogCom

A simulation and PPO implementation for candidate-set control in collaborative fog computing.

Open repository ↗