ImageService

Group Members:

• Sai Pragna Etikyala (1213195137)
• Srinath Ganesan (1213222229)
• John Santhosh (1213074510)
• Akshay Muraleedharan Nair Santhy (1212981859)

Requirements:

• JDK 1.8+
• Maven
• AMI ami-c3a7b1a3 region: us-west-1
• AWS Account.
• Amazon SQS.
• Amazon S3 bucket.
• MongoDB running on port 27017.
• Tensorflow deeplearning model setup in the EC2 instance
• recognize-image.sh available in /home/ubuntu/
• ports (publically accessible) 8080, 27017

S3 Buckets

File Bucket name : cloud-computing-file-bucket Result Bucket name : cloud-computing-result-bucket

Folders : images jobs

MongoDB

Should be accessible on http://54.215.241.86:27017/ Database : images Collection : jobs

SQS

Request Queue : image-recognition-request-queue
Shutdown Queue : instance-shutdown-queue

Dependencies:

pom.xml

Steps to setup

1. Update the application.properties file.

EC2 properties.

amazon.ec2.ami.id= ami-c3a7b1a31

amazon.ec2.access.key.id=

amazon.ec2.secret.access.key=

S3 properties.

amazon.s3.base.url=https://s3-us-west-1.amazonaws.com

amazon.s3.bucket.name=cloud-computing-file-bucket

amazon.s3.bucket.image.folder.name=images

amazon.s3.bucket.job.folder.name=jobs

amazon.s3.bucket.output.file.name=result.txt

amazon.s3.result.bucket.name=cloud-computing-result-bucket

SQS properties.

amazon.sqs.request.queue.name=image-request-queue

amazon.sqs.request.queue.message.group.id=image-recognition-request-queue

amazon.sqs.instance.shutdown.queue.name=instance-shutdown-queue

amazon.sqs.instance.shutdown.queue.message.group.id=instance-shutdown-queue

MongoDB.

spring.data.mongodb.host=localhost

spring.data.mongodb.port=27017

spring.data.mongodb.database=images

Application Settings.

sleep.time.min=1000

sleep.time.max=5000

sleep.time.shutdown=5000

amazon.ec2.instance.max.count=18

3. Navigate to image-service.
4. Run mvn clean install
5. Run mvn spring-boot:run
6. Go to http://54.215.241.86:8080/ to ensure that the application is running. Should display Image Service is running.

Spins up the spring-boot application on http://54.215.241.86:8080/ Master Instance (Web-Tier) IP Address : 54.215.241.86

Rest Endpoints

1. GET : http://54.215.241.86:8080/

response : Image Service is running.

2. GET : http://54.215.241.86:8080/cloudimagerecognition/jobs

Response: All the job records in MongoDB.

3. GET : http://54.215.241.86:8080/cloudimagerecognition?input=some_valid_image_url

some_valid_image_url=http://visa.lab.asu.edu/cifar-10/0_cat.png Response : bobsled, bobsleigh, bob

Sample Mongo Document :

{ "_id": "69b87af8-c082-4f09-ba52-c367f90efde5", "_class": "model.Job", "url": "http://visa.lab.asu.edu/cifar-10/0_cat.png", "result": "bobsled, bobsleigh, bob (score = 0.06877)", "inputFilename": "0_cat.png", "status": "COMPLETE", "submitDateTime": ISODate("2018-03-29T14:42:09.557Z"), "completedDateTime": ISODate("2018-03-29T14:43:14.747Z"), "error": "2018-03-29 14:43:13.065865: I tensorflow/core/platform/cpu_feature_guard.cc:137] Your CPU supports instructions that this TensorFlow binary was not compiled to use: SSE4.1 SSE4.2 AVX AVX2 FMA2018-03-29 14:43:13.515949: W tensorflow/core/framework/op_def_util.cc:343] Op BatchNormWithGlobalNormalization is deprecated. It will cease to work in GraphDef version 9. Use tf.nn.batch_normalization()." }

Bash Scripts :

recognize_image.sh #!/bin/bash python /home/ubuntu/tensorflow/models/tutorials/image/imagenet/classify_image.py --image_file $1 --num_top_predictions 1

shutdown.sh #!/bin/bash shutdown -h now

Remote Slave Instance Startup Script :

#!/usr/bin/env bash" pip install awscli --upgrade /home/ubuntu/.local/bin/aws configure set aws_access_key_id AKIAIJD7XM7766A3MIUA /home/ubuntu/.local/bin/aws configure set aws_secret_access_key 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 sudo pip install --upgrade pip sudo pip install numpy source /home/ubuntu/tensorflow/bin/activate cd /home/ubuntu/image-service-slave java -jar target/image-service-slave-1.0-SNAPSHOT.jar