SpecialVFX@Cloud

The goal of this project is to design and implement a service hosted in an elastic public cloud to support a hypothetical Special VFX Studio. The service executes computationally-intensive tasks such as generating photo-realistic images using ray-tracing and applying effects to images using image processing algorithms.

Cloud Computing and Virtualization project of group 4 - MEIC @ IST 2023/2024.

Authors

• 110817 André Páscoa
• 110860 André Jesus
• 110893 Nyckollas Brandão

@IST
Master in Computer Science and Computer Engineering
Cloud Computing and Virtualization - Group 4
Summer Semester of 2023/2024

Table of Contents

• SpecialVFX@Cloud
  • Authors
  • Table of Contents
  • Overview
  • Architecture
  • Usage
    • Requirements
    • Running locally
      • Load Balancer and Auto Scaler
      • Web Server
      • Workloads
    • Deploying to AWS
    • All AWS scripts
    • Test scripts

Overview

This project provides a simplified Java application designed to handle different types of requests:

• Render Image: Returns the rendering of a viewport received as input.
• Blur Image: Returns a blurred copy of the input image.
• Enhance Image: Returns a copy of the input image with highlighted edges.

These are exposed through a Web Server that receives HTTP requests. The application is designed to be deployed in a cloud environment, where requests are distributed to instances via a Load Balancer and the number of instances can be scaled based on the load via the Auto Scaler. Additionally, the application is instrumented to collect and store metrics about the requests and the execution of the workloads.

Architecture

The project contains the following components:

• app - The main application that exposes the functionality of the workloads.
  • raytracer - The Raytracer workload.
  • imageproc - The BlurImage and EnhanceImage workloads.
  • webserver - The web server exposing the functionality of the workloads.
• instrumentation - The instrumentation module that contains the code for collecting metrics with Javassist.
• mss - The Metric Storage System that stores the requests and metrics collected by the instrumentation module.
• loadbalancer - The load balancer + auto scaler.
  • The load balancer that distributes the requests to the different instances of the application
  • The autoscaler that scales the number of instances of the application based on the load.
• scripts - Scripts to automate instrumentation and deployment tasks;
  • aws - Scripts to deploy the application to AWS.
  • aws_test - Scripts to test the application, AWS or otherwise.

Usage

Requirements

• Java 17+
• Maven 3.6.3+

Running locally

1. Make sure your JAVA_HOME environment variable is set to Java 17+ distribution
2. Run mvn clean install in the root directory to build the application

Load Balancer and Auto Scaler

1. To run web server instances in AWS while using the load balancer and autoscaler locally, you need to setup the necessary environment variables:
   • AWS_ACCESS_KEY_ID
   • AWS_SECRET_ACCESS_KEY
   • AWS_DEFAULT_REGION
   • AWS_IMAGE_ID
   • BLUR_LAMBDA_ARN
   • ENHANCE_LAMBDA_ARN
   • RAYTRACER_LAMBDA_ARN
   • AWS_SECURITY_GROUP
2. Run java -cp loadbalancer/target/loadbalancer-1.0.0-SNAPSHOT-jar-with-dependencies.jar pt.ulisboa.tecnico.cnv.LoadAndScaleWebServer to start the load balancer and auto scaler.

Web Server

1. To store obtained metrics in remote DynamoDB, setup environment variables for AWS_ACCESS_KEY_ID and AWS_SECRET_ACCESS_KEY with DynamoDB permissions.
2. Run mvn exec:exec -f app/webserver to start the web server using the instrumented classes. Alternatively you can use the following command: java -cp app/webserver/target/webserver-1.0.0-SNAPSHOT-jar-with-dependencies.jar pt.ulisboa.tecnico.cnv.webserver.WebServer -javaagent:instrumentation/target/instrumentation-1.0-SNAPSHOT-jar-with-dependencies.jar=RequestAnalyzer:pt.ulisboa.tecnico.cnv.imageproc,boofcv,pt.ulisboa.tecnico.cnv.raytracer:output

Workloads

To run a workload directly, you can use the following commands:

• Ray Tracer: java -cp app/raytracer/target/raytracer-1.0.0-SNAPSHOT-jar-with-dependencies.jar pt.ulisboa.tecnico.cnv.raytracer.Main <input-scene-file.txt> <output-file.bmp> 400 300 400 300 0 0 [-tm=<texmap-file.bmp>] [-aa]
• Blur Image: java -cp app/imageproc/target/imageproc-1.0.0-SNAPSHOT-jar-with-dependencies.jar pt.ulisboa.tecnico.cnv.imageproc.BlurImageHandler <input-file> <output-file>
• Enhance Image: java -cp app/imageproc/target/imageproc-1.0.0-SNAPSHOT-jar-with-dependencies.jar pt.ulisboa.tecnico.cnv.imageproc.EnhanceImageHandler <input-file> <output-file>

Deploying to AWS

To deploy the application/system to AWS, you can use the provided scripts in the scripts/aws directory. The scripts require the AWS CLI to be installed and configured with the necessary permissions.

1. Configure the scripts/aws/config.sh file with the necessary parameters/environment variables.
2. If it's the first time deploying the system, run ./scripts/aws/create-and-launch-system.sh to create the necessary resources in AWS, including the lambda functions, the worker image and the instance running the load balancer and autoscaler;
3. If the system is already created, run ./scripts/aws/launch-system.sh to deploy the load balancer and autoscaler, because the image and lambda functions are already created;

To terminate the system, run ./scripts/aws/terminate-system.sh - this only terminates the instances and the load balancer. To terminate and delete all resources, run ./scripts/aws/terminate-and-delete-system.sh.

All AWS scripts

You can find all the AWS scripts in the scripts/aws directory, allowing you to also run tasks individually.

The scripts are:

• create-and-launch-system.sh - Creates the necessary resources in AWS and launches the system.
  • Registers lambda functions for the workloads
  • Creates an image (AMI) for the workers (webserver instances).
  • Deploys an instance for load balancer and autoscaler.
• launch-system.sh - Launches the system in AWS.
  • Deploys the load balancer and autoscaler.
• terminate-and-delete-system.sh - Terminates and deletes the system in AWS.
  • Terminates the load balancer instance and any running webserver instances.
  • Deletes/Deregisters the lambda functions for the workloads.
  • Deletes/Deregisters the image (AMI) for the workers (webserver instances).
• terminate-system.sh - Terminates the system in AWS, without deleting the resources (AMI and lambda functions).
  • Terminates the load balancer instance and any running webserver instances.
• function-register.sh - Registers the lambda functions and roles for the workloads.
• function-deregister.sh - Deregisters the lambda functions and roles for the workloads.
• create-image.sh - Creates an image (AMI) for the workers (webserver instances).
  • Launches an instance and installs the necessary software.
  • Tests the webserver software in the instance.
  • Creates an image from the instance.
  • Terminates the temporary instance.
• deregister-image.sh - Deletes/deregisters the image (AMI) for the workers (webserver instances).
• launch-vm.sh - Launches a VM instance.
• install-vm.sh - Installs the necessary webserver software in a VM instance.
• test-vm.sh - Tests the webserver software in a VM instance.

Others:

• launch-deployment.sh - Launches the deployment of the alternative system in AWS, using the load balancer and autoscaler services from AWS instead of our own version.
• terminate-deployment.sh - Terminates the deployment of the alternative system in AWS.

Test scripts

You can find all the test scripts in the scripts/aws_test directory, allowing you to run tests on the system. The scripts are:

• test-imageproc-01.sh - Tests the BlurImage workload.
• test-imageproc-02.sh - Tests the EnhanceImage workload.
• test-raytracer-01.sh - Tests the Raytracer workload.

These can be configured to your liking, and you can create new test scripts based on these too.