3 ms·

I'm curious, what does your document storage microservice look like?

• lobster_johnson 11y ago
  It's a very simple layer on top of PostgreSQL that stores basic JSON documents. Documents are organized into paths, which can be queried, and it supports things like tags and putting documents into timelines, and it does path/document-level permissions. It does a great job for basic document-oriented content storage, and we use it for many things, but it admittedly falls down for anything requiring inter-relationships between documents (ie., graphs), and it has some other issues. It doesn't support querying on arbitrary fields, so we rely on indexing all documents in ElasticSearch, which introduces a lag problem where the UI needs to explicitly wait until ES has been populated after a change. Also, it's written in Ruby and uses HTTP, so it's not fast. We're working on a "2.0" that changes it completely. It implements operational transforms/fine-grained revisions, supports multiple backends (such as Redis), fast synchronous updates to other data stores such as ElasticSearch, and uses a faster RPC protocol. Currently writing it in Node.js/ES6, but might rewrite in Go once we see that the design works. It will be open source, if you're interested.
  • darksecond 11y ago
    Thanks for the explanation. Sounds like a really cool project.