Mobile Edge Computing (MEC) Benchmark Suite

A list shown the applications used in the research projects.

Artificial Intelligence

CLIP with Jina AI: Microservice CLIP application in Jina AI framework. This repo would cover the generation of deployment files on Kubernetes.

Apache Traffic Control: The on-premise CDN system. This repo covers the configuration files of Kubernetes. To running an Apache CDN, you in k8s, you don't need to create an additional DNS server.

NextEPC: Open sourced LTE testbed. Containerized and k8s-friendly. In our project, we only use the RAN core network of NextEPC.
SrsRAN: Same as NextEPC. In our project, we only used the eNB and UE parts of SrsRAN.

Magma: Fork from official repo to build on-premised AGW and Orc8r by Helm charts. Currently integrated with eNB function of SrsLTE for 4G testbed. Repo is under processed.

Secure MPC: Benchmarks for EMP-Toolkit's AGMPC, a maliciously secure BMR-style protocol tolerating an arbitrary number of corruptions.

Robotic OS: A popular application used in robotic and IoT industry. This repo containerized the simple pub-sub example.

IOTCOMMS: An framework used for creating IoT application with a variety of different workload configurations. This framework aims to caputure the diverse nature of IoT applications in every stage - the workload generator, the tranport layer protocols and the could server side components.

TensorFlow Model Builder: This repo shows some reference steps of formatting official trained TensorFlow models into protobuf files.

Video 360: A framework developed by Fraunhofer Heinrich Hertz Institute. JavaScript implementation of MPEG-OMAF viewport-dependent video profile with HEVC tiles. We add containerized deploying methods as for our MEC benchmarking.

Object Detection TensorFlow Serving This ML model serving platform based on TensorFlow Serving, we refered to this public repo and add our deployment for MEC benchmarking. Checking the directory of ./k8s for deployment details.
Streamer: A framework developed by CMU, used for building pipelines from video processing functional blocks. Streamer is designed to work with ML inference process, the analytics part. TODO: currently linked to official SAF version, since not sure the legacy version is allowed to share or not.
YOLOv3 Tensorflow: YOLOv3 application ported to python TensorFlow2. YOLOv3 is for real-time object detection developed by Joseph Redmon, a U of W student.