CacheCow

CacheCow is a distributed in-memory cache service that consists of a graph of intercommunicating nodes. Each node assumes the role of a generic endpoint to process client requests that handles and redirects client requests and as a fragment of the distributed cache store.

Helpful Links: Google Drive, Design Document

Authors: Eliot Solomon, Halit Ozkaya, Mark Pepperl, Nadav Levanoni, Ryan Huckleberry

File Structure

• cache-node: Javalin project to run cache node instance
• monitor-node: React project to run monitor node instance
• performance-testing: Performance testing scripts to test cache metrics

Set Up

1. Create a new directory

mkdir CacheCow

2. Clone the repository into the new directory

git clone git@github.com:RiceComp413-Fall2022/CacheCow.git CacheCow

Running Locally

1. Download Node.js if it's not already installed

2. Set up the monitoring node

# Navigate to monitoring node directory
cd monitoring-node

# Install packages
npm install

3. Launch a local cache cluster

# Navigate to cache node directory
cd cache-node

# Run 3 node cluster locally
./gradlew run --args 'local 0 7070'
./gradlew run --args 'local 1 7071'
./gradlew run --args 'local 2 7072'

# Note: use the -s flag to run in scalable mode
./gradlew run --args 'local 0 7070 -s'
./gradlew run --args 'local 1 7071 -s'
./gradlew run --args 'local 2 7072 -s'

4. Check that the nodes are running

# Ping a cache node
curl -X GET "localhost:7070/v1/hello-world"

5. Monitor the cache

• Open localhost:3000 in any browser

Running on AWS

1. Download Python 3 if not already installed

2. Download the AWS Command Line Interface

3. Set up an AWS key pair

• Navigate to EC2 Dashboard on AWS Management Console
• Create a new key pair
• Download the pair and save the file rootkey.csv to the root CacheCow directory

4. Configure the AWS CLI

aws configure

# Respond to the prompts using your credentials in rootkey.csv
AWS Access Key ID []: <your-access-key>
AWS Secret Access Key []: <your-secret-key>
Default Region Name []: us-east-1
Default Output Format []: <enter>

5. Install Python packages

pip3 install requests "boto3[crt]" fabric

# Note: if boto3 install fails try this
pip3 install --upgrade pip

6. Launch a cluster on AWS

# Crate a cluster with 2 nodes
python3 pasture.py create 2

# Note: use the -s flag to run in scalable mode
python3 pasture.py create 2 -s

7. Delete the cluster

python3 pasture.py delete 2

Sending Requests

1. Store a key-value pair

curl -X POST -H "Content-Type: text/plain" --data "<value>" "localhost:7070/v1/blobs/{key}/{version}"

2. Fetch a value

curl -X GET "localhost:7070/v1/blobs/{key}/{version}"

3. Clear all key-value pairs

curl -X DELETE "localhost:7070/v1/clear"

4. Scale cluster (Note: cluster must be running in scalable mode)

curl -X POST "localhost:7070/v1/launch-node"

Performance Testing

There are multiple performance tests. Here, we will run long-tailed.py which uses a heavy-tailed lognormal distribution to simulate cache-aside performance. The test is best performed against a cache which can hold a maximum of 100 keys. The distribution parameters generate 995 keys, 408 of which are unique. This ensures that the cache handles eviction appropriately.

1. Move into the performance-testing directory

cd CacheCow/performance-testing

2. Run the test script: long-tailed.py

python3 long-tailed.py --url <domain name>

For example,

python3 long-tailed.py --url localhost:7070