The vision of Edge computing is to bring computing towards the edge of the network. Our previous work, ENORM facilitates the hosting of services closer to user devices that frequently interact with the Cloud so that communication latencies can be reduced. To improve the Edge Computing serives supported by ENORM, DYVERSE further provides priority-based dynamic resource allocation while supporting multiple Edge applications on an Edge node.
This is a developing research project and some features might not be stable yet.
All source code, documentation, and related artifacts associated with the DYVERSE open source project are licensed under the Apache License, Version 2.0.
In addition to the iPokeMon use case supported in ENORM, a real time Face Detection application is also provided to test DYVERSE. The differences between these two uses cases are:
- FD requires both the Edge and the Cloud servers work together to process requests, whereas iPokeMon requires either the Edge or the Cloud server to process requests.
- FD does not require a database server, whereas iPokeMon requires a database server.
You will need the code provided in folder Application to use the DYVERSE prototype:
- Follow instructions of Edge Manager in folder EdgeManager to set up the Edge node.
- Follow instructions of Cloud Manager in folder CloudManager to request (and terminate) Edge service for FD.
- Follow instructions of FD Cloud Server in foder FaceDetection-CloudServer to set up the Cloud Server
- Follow instructions of FD Client in folder FaceDetection-EdgeServer to set up the streaming camera.
Please cite DYVERSE - DYnamic VERtical Scaling in Multi-tenant Edge Environments when using this project as follows:
N. Wang, B. Varghese, M. Matthaiou, and D. S. Nikolopoulos, "DYVERSE - DYnamic VERtical Scaling in Multi-tenant Edge Environments," arXiv:1810.04608, 2018.