This was my first time attending the Cassandra Summit where Robin unveiled its container-based, software-defined infrastructure platform for stateful apps such as Cassandra, Oracle, Hadoop and ELK.
Considering this was my first visit, the conference experience was quite amazing – here are some Cassandra Summit takeaways. It was a healthy mix of experts and those like me, engineers who wanted to learn more about Cassandra and DataStax.
The best part for me was, of course, the training. It provided a good overview of the latest DataStax 5.0 release features such as Graph database, Cassandra integration with Spark and Solr, etc. I have to confess that I could have done without the modified Chromebooks. The keyboard was too small and the laptop was really underpowered to do anything useful.
Key highlights from the conference that I captured:
- Cassandra popularity is growing. Per DB Engine ratings, Cassandra is now in the 7th position
- Jonathan Ellis is stepping down as the chair for Apache Cassandra after 7 years
- Excellent talks from Netflix, Target, Instacluster, Datascale, etc. regarding running Cassandra in production
- While 2.1 and 2.2 still remain the most dominant versions being used in production, there was a lot of talk about 3.0 – SASI, materialized views, optimized read path, etc.
At the Robin booth we showcased multiple features that tremendously simplify management of Cassandra clusters. These include the ability to:
- Provision C* clusters within Docker containers
- Run C* clusters with guaranteed read and write IOPs for predictable performance
- Dynamically change resources allocated to the cluster without requiring a configuration change or an app restart
- Snapshot (point-in-time copy) the complete C* cluster, including OS, DB binary, configuration, and data. Once you have an array of snapshots, you can rollback to any of these snapshots.
- Create snapshot-based thin clones of the C*. These are space-efficient clones, so storage is consumed only for changed blocks.
- Scale out a C* cluster by adding additional nodes
- Achieve C* node container auto-failover across hosts, with the same volume and with the same ip address.
You can see the demos here.
Based on my conversations at the show there is clearly a lot of interest in using containers, across both stateless and stateful apps. While stateless apps are well covered, massive gaps must still be addressed for solutions required for stateful apps. At Robin we are addressing this very need — for more information, see ROBIN Hyper-Converged Kubernetes Platform for NoSQL Databases on the Robin website.