cse546-project1

Cloud Computing Project 1

Group: CCP

Team Members

Name	ASU ID
Gaurav Kulkarni	1225477253
Parth Shah	1225457038
Shreyas Kirtane	1225453736

---

Project Report

---

Gaurav's Tasks:

• Researched about Amazon Cloudwatch metrics and alarm, Autoscaling groups and scaling policies
• Tested out different simple step and target tracking policies along with custom metrics
• Evaluated individual policies against the required time limit and scaling requirements
• Managed the documentation of the project

---

Parth's Tasks:

• Implement SQS method in web tier to push messages to request queue
• Implement SQS method in app tier to push results to response queue
• Add multithreading to app tier and web tier to poll SQS queues in background

---

Shreyas's Tasks:

• Implement Web Tier using Flask
• Incorporate image classification in app tier server
• Implement multiple reader single writer lock to poll result dict efficiently
• Implement S3 methods to download and upload files
• Create dockerfile, Makefile, configs, venv requirements file
• Figure out deployment

---

AWS Credentials

• user: demo
• default region: us-east-1

---

Usage:

chmod 600 web-tier-key.pem
ssh -i web-tier-key.pem ec2-user@[public-ip]

---

Web Tier details

• Base AMI: AWS-Linux
• user: ec2-user
• public-ip: assigned when started/launced (see ec2 console)

---

SQS

• request queue: requests

• request queue url: https://sqs.us-east-1.amazonaws.com/583586231865/requests

• response queue: responses

• response queue url: https://sqs.us-east-1.amazonaws.com/583586231865/responses

---

S3

• input bucket: cse546images
• resuls bucket: cse546results

---

Deploy on your aws account

1. Create queues and buckets
2. Create an ec2 instance name "web-instance" and install docker on it.
3. Configure aws credentials using aws configure or env variables
4. Clone the git repo and do:

   make run_wt
   

5. Create an ec2 instance named "app-instance" and install docker.
6. Configure aws credentials using aws configure or env variables
7. Clone the git repo and do:

   make run_at
   

8. Create an AMI out of app-instance.
9. Create a launch template using this AMI
10. Create an Alarm based on ApproximateNumberOfVissibleMessages on request queue
11. Create an AutoScalingGroup with a step policy
12. Create steps:
    • set 1 capacity for metric >= 1 and < 2.
    • set 2 capacity for metric >= 2 and < 3.
    • ...
    • ...
    • set 20 capacity for metric >= 20 and < +ve infinity
13. Create a simple scale in policy to scale in: configure to remove 20 instances when alarm in OK state
14. Attach alarm to the scaling policy
15. Run the workload generator on public ip of web instance