Skip to content

Roadmap

okay edited this page Oct 23, 2017 · 1 revision

Sybil Roadmap

Q4 2017

Distributed queries

This quarter is about preparing for and implementing distributed queries, because no matter how great a backend is, it will need to shard horizontally at some point. And perhaps that point is now (because it will take months to a year to have it working and reliably tested).

To get to distributed queries we will need:

  • Distributed Ingestion: Kafka pipes setup (or another solution not involving JVM)
  • Distributed Queries: one master aggregator that queries and combines results from all nodes
  • High cardinality support: our leaves need to give manageable results to the master agg

Modular column storage

Other features under consideration are swappable and customizeable per column storage. Some example storage types would be "timestamps", "logs", "blobs", etc. The goal here is to specifically pick one new column type and implement it as a go plugin for sybil.

Organizational Goals

  • Published performance results on real workloads
  • Join an open source foundation
  • Ramp up new team members
Clone this wiki locally