GreenLAC

The AWS Greengrass Lambda Autoscaler Core (GreenLAC) is a load balancer and reverse proxy for supporting serverless deployments of edge-core IoT applications with AWS Greengrass. GreenLAC enablers applications on resource-constrained edge nodes to leverage neighbours edge nodes and the core cloud for real-time processing of workloads.

Architecture

GreenLAC is an AWS Greengrass Component that will soon be available to the dev community. It uses the core component of Greengrass, Nucleus, to forward requests to local Lambdas managed by the Lambda Manager. Greengrass scales out Lambdas in the core cloud whenever its local resources reach saturation, and it can also send requests to other edge nodes.

Build Instruction

The build process of GreenLAC is the same used for any AWS Greengrass component.

Download

git clone https://github.com/pacslab/GreenLAC.git

Prerequisite Setup

AWS Greengrass Nucleus (Core)
AWS Greengrass LambdaManager
AWS Greengrass CLI

Build From Source

Please make sure Java 11 and Maven 3 are installed before building the package. To build package, run:

mvn clean package

You need to upload the greenlac.jar to an S3 Bucket on AWS.

Component Deployment on AWS

We have provided the JSON recipe for deployment here. Please make sure to change any configuration and parameters according to your deployment needs. Below you can see some of the many configuration parameters of GreenLAC

  "ComponentConfiguration": {
    "DefaultConfiguration": {
		"greenlac.endpoint.core": "https://gydrsqutj5.execute-api.us-east-1.amazonaws.com/default/230mb_model_1_image",
		"greenlac.scaling.policy": "core-edge-pc",
		"greenlac.scaling.utilizationThreshold.cpu": 50,
		"greenlac.loadBalancer.buffersize": 100,
        [...]
    }
  }

Log Debug

Logs are stored in the default Greengrass folder inside the device which is running the Greengrass core component. For GreenLAC log, check /greengrass/v2/logs/GrenLAC.log

2022-02-23T01:36:25.149Z [INFO] (Copier) GreenLAC-a: stdout. [30m2022-02-23 01:36:25,147 INFO  [http-nio-8080-exec-18] com.jaimedantas.greenlac.loadbalancer.Lambda: Sending request to: https://gydrsqutj5.execute-api.us-east-1.amazonaws.com/default/230mb_model_1_image. {scriptName=services.GreenLAC-a.lifecycle.Run, serviceName=GreenLAC-a, currentState=RUNNING}
2022-02-23T01:36:25.159Z [INFO] (Copier) GreenLAC-a: stdout. [30m2022-02-23 01:36:25,156 INFO  [scheduling-1] com.jaimedantas.greenlac.monitor.Metrics: CPU: 2.5 % - Memory: 7.998381370355275 %. {scriptName=services.GreenLAC-a.lifecycle.Run, serviceName=GreenLAC-a, currentState=RUNNING}
2022-02-23T01:36:25.184Z [INFO] (Copier) GreenLAC-a: stdout. [30m2022-02-23 01:36:25,181 INFO  [http-nio-8080-exec-20] com.jaimedantas.greenlac.controller.Controller: Request to ml endpoint - 87e86207-218f-4b9c-afb3-5974b811bede. {scriptName=services.GreenLAC-a.lifecycle.Run, serviceName=GreenLAC-a, currentState=RUNNING}

Citation

Please cite this work as followed:

@inproceedings{10.5555/3566055.3566075,
author = {Jaime, Dantas and Hamzeh, Khazaei and Marin, Litoiu},
title = {Green LAC: Resource-Aware Dynamic Load Balancer for Serverless Edge Computing Platforms},
year = {2023},
publisher = {IBM Corp.},
address = {USA},
booktitle = {Proceedings of the 32nd Annual International Conference on Computer Science and Software Engineering},
pages = {169–174},
numpages = {6},
keywords = {Edge Computing, AWS Greengrass, IoT, Serverless Computing, Function as a Service},
location = {Toronto, Canada},
series = {CASCON '22}
}

License

This project is licensed under the MIT License - see the LICENSE file for details.

pacslab / GreenLAC