This is a (fairly naive) implementation of a JGit Git repository using Amazon DynamoDB and S3 as a store. It is built
on the JGit DfsRepository - and it is not particularly optimized. It is at a very early stage and has not been properly
tested. Nonetheless, I am hoping this is useful for others - and a bit selfishly, I am hoping people will report back
issues, contributions.
Usage is fairly straighforward - create a new instance of the AmazonRepository using the AmazonRepository.Builder,
and passing in a JGitAwsConfiguration setup the way you like.
The repository uses a Dynamo table called jga.Refs to store Git refs, a Dynamo table called jga.Configurations to store
a repository's config, a single S3 bucket called jga.Packs to store pack file contents, and a Dynamo table called
jga.PackDescriptions to store meta-data about the pack files. All these names are configurable.
You can configure the Dynamo table names, as well as the S3 bucket name by setting the appropriate properties on
JGitAwsConfiguration. Note that you can also configure the initial provisioned throughput used on the Dynamo tables
using JGitAwsConfiguration - by default they are all set up for 1 read unit and 1 write unit. Scaling this up is up
to you to do manually or using a tool such as Dynamic DyanmoDB.