Greenplum Kontained: Coordinating Many Postgres Instances on Kubernetes
Goutam Tadi is a Senior Software Engineer at Pivotal Software working within the Research and Development Team for Data. As a long time big data enthusiast, Goutam is currently part of the GPCloud group which is architecting Greenplum to run on Kubernetes clusters.
Prior to this cloud native initiative, Goutam worked on packaging Greenplum to work on Ubuntu and helped to make the first open source binaries for Greenplum available to the world. Goutam also worked on the GemFire / Greenplum Connector which allows for easy data movement between GemFire (Apache Geode) and Greenplum. Goutam joined Pivotal from Tata Consultancy Services where he worked on a variety of big data projects, such as custom-coding writable formats for Hadoop.
Goutam earned a Masters Degree in Computer Science from California State University, Long Beach, and a Bachelor of Technology, Electronics & Communications Engineering, from Vignan Engineering College in Andhra Pradesh, India. During his free time, he enjoys watching movies, biking, hiking, a good chat over coffee, and - of course - coding.
Kubernetes provides a new way to deploy and run software in containers with flexible Services and global DNS discovery. This session will describe the new deployment of Greenplum on the Pivotal Container Service (PKS - the K is for Kubernetes), which provides a Kubernetes environment running on an enterprise-grade infrastructure powered by the BOSH open source project.
Container environments change the game for Greenplum. Persistent disks add a new degree of HA, and speed the creation of replacement containers. On-demand deployment of Greenplum scales to any number of segment containers with a single command. For segment failover, a DBA can specify that a mirror takes over, requiring no further Greenplum downtime for rebalancing, using affinity rules to determine how computing resources are distributed. For master failover, the Kubernetes predefined Service for port 5432 means that clients can be load-balanced to the standby master with no change in IP address.
What about storage in container environments? Software-defined storage can be configured to appear infinite, and to replicate data across widely distributed zones.
Learn more about the power of combining Kubernetes with Greenplum - the best enterprise class, open source, Postgres-based data analytics platform.
- 2018 April 20 09:50 EDT
- 50 min
- Liberty II-III
- PostgresConf US 2018
- Greenplum Summit