Deployment

Although yams is meant to be run as a server, the project provides no strong opinions about how or where that server is run. Infrastructure deployments vary greatly between companies, and very few assumptions hold true.

The Go binary comprising the server can be easily built for cross-platform use, and the only real requirement is the ability to serve or front an HTTP service on an open port.

If you have requests or ideas for deployment patterns to support more natively, please raise an issue on GitHub.

At a Glance

An overview for a recommended, basic yams deployment would be:

• Use yams dump commands to populate data sources in S3
• Schedule these using your friendly, neighborhood cron equivalent for evergreen data
• Deploy on a single (large) EC2 instance or container
• Point yams server at the cached data in S3

Performance

The following tips should help in maximizing the performance of a yams deployment for environments where the extra edge is needed:

• Use CPU-optimized EC2 instances: access simulation is almost an entirely CPU-bound endeavor and yams is heavily parallelized: use as many cores as you can spare
• Split data sources into multiple files: while data loading will rarely be the limiting factor, it can become clunky as your sources approach GB scale. Individual data sources are loaded on their own goroutines, and so benefit from being partitioned by environment, resource type, etc
• Consider horizontal scaling: adding multiple instances of the yams server is an effective way to scale up. However, since every server independently refreshes its data sources it can introduce consistency issues if servers become out of sync