TriggerPerf

TriggerPerf is a custom benchmarking tool designed to evaluate the performance of datastore triggers. It currently supports three AWS datastores: S3 Standard, DynamoDB, and AuroraDB. The tool automates trigger setup and configuration, leveraging three Lambda function: Setup, Write, and Receive to create a testbed for benchmarking various trigger mechanisms. Currently TriggerPerf supports S3 Event Notifications, DynamoDB Streams and AuroraDB UDF triggers.

Usage

1. Configuration

All configurations are managed via the config.yaml file. This includes: - Datastore Parameters: Define the datastore type and its specific settings. - Experiment Details: Specify key-value sizes, throughput, and other benchmarking parameters. - AWS Credentials: Include the necessary access keys and region information. - Define experiment details and AWS credentials in config.yaml. ```bash

other inputs: "s3", "etcd", "dynamo", "aurora"

datastore: "dynamo" awsacc_id: xxxxxxxxxx region: us-east-1 ```

2. Setup & Run

Run the setup.py script to initialize and configure the benchmarking environment: - Deploys the Setup, Write, and Receive Lambda functions. - Configures the triggers for the selected datastore. - Integrates AWS CloudWatch for logging and performance metrics collection. After setup, run the main.py script to start the experiments defined in config.yaml

bash python setup.py python main.py

3. Logs & Teardown

• Logs: Filtered logs are automatically saved to the /logs directory for later analysis. Run the /proc_logs/latency_data_processing.ipynb to generate graphs. bash python teardown.py
• Teardown: To remove all datastore and trigger configurations from your AWS account, run the teardown.py script:

bash python teardown.py

4. Extending TriggerPerf: Adding a New Datastore

To add support for a new datastore, follow these steps:

Step 1: Create a Driver File

• Add a new driver file inside the /drivers directory.
• Implement necessary functions that align with the functionality of existing datastore integrations.
• The driver should include functions for:
  • Setting up the target datastore.
  • Performing CRUD operations.

Step 2: Modify Lambda Functions

• Update ./aws_lambdas/lambdas/initial-lmd.py:
  Add functions required to connect and interact with the new datastore.
• Update ./aws_lambdas/lambdas/recv-lmd.py:
  Implement functions to handle the datastore's trigger mechanism.

Step 3: Maintain Consistency & Compatibility

• Ensure the new integration adheres to TriggerPerfs existing structure and conventions.
• This maintains seamless compatibility with the benchmarking framework.