Wednesday, June 25, 2014

#MongodDBWorld Day 2 Keynotes: Elliott Horowitz What @MongoDB is working on

See the schedule here:

Elliott Horowitz Keynote

Design Principles

  • Developer Productivity
  • Horizontal Scalability
  • Operational Scalability

Administering a single server should be easy and so should a large cluster(s).

Three improvements are coming:


Where are we today after 7 years of development.

People working with MongoDB think about the fact that locking takes place at the Database level.
The crowd goes wild at Document level locking which will be released in 2014. Document Level locking will be a textbook implementation that will be familiar to relational database engineers.

It is released on github (v.2.7.3) BUT IS NOT READY FOR PRODUCTION.

Storage Engine

A consistent core that has stood the test of time for past 7 years. What is the right storage engine for the future of MongoDB?
There is no single right engine. So MongoDB is implementing a pluggable storage api.

The API has to be detailed enough to support ALL MongoDB features. The engine should support the operational scalability requirement.

Storage Engines should be able to be added to a single or few nodes per cluster. So you can create a pluggable storage engine optimized for different workloads. eg. Performance, Compression etc.

- In-Memory
- RocksDB (out of Facebook focused on compression)
- InnoDB (out of MySQL)
- TokuKV
- FusionIO (bypass filesystem for low latency)

The Engineering Team have built In-Memory and RocksDB storage engines.

This is experimental code that is also hitting github.


DevOps need tools to manage hundreds or thousands of nodes.

Mongo Management Service (MMS)

MMS will offer Automation. Each server will have an Automation Agent.

[Ed Note: I wonder if Puppet or Chef can access MMS Automation features]

A live demo defining a 10 node cluster.
Build a cluster (you can do this easily with Puppet or Sheff)

MMS can also automate upgrades.

Automation is integrated with the Monitoring capabilities in MMS.
MMS is also integrated with AWS allowing provisioning in EC2.

It is pretty impressive to see a 10 node cluster upgraded in just a few minutes.

MMS will also have an API which will allow access from your own orchestration platform of choice..

via WordPress