• Apache Cassandra
  • Technical
Presentation by Adam Zegelin at Cassandra Summit 2014 – San Francisco

Instaclustr’s Adam Zegelin presented at the September 2014 Cassandra Summit, on the topic,  Multi-Region Clusters — Cassandra Deployments Split Between Heterogeneous Data Centre

Speaker: Adam Zegelin, CTO at Instaclustr

In this presentation we discuss a method of provisioning and running an Apache Cassandra deployment spilt between multiple heterogeneous data centers which, rather than allocating per-node public IPv4 addresses or configuring mesh VPNs, uses Port Address Translation (PAT) for node↔internet connectivity and is self- configuring and discoverable via DNS Service Discovery (DNS-SD or wide-area Bonjour). While Cassandra has built-in support for AWS EC2 multi-region/data centre topologies (via Ec2MultiRegionSnitch, etc), the existing solution requires the wasteful allocation of public IPv4 addresses per-node. Additionally there is little support for topologies that are either a mix of or deploy completely on alternative infrastructure providers. Our solution uses a single public IP address per data center, is provider-agnostic, doesn’t introduce the configuration and management overheads of a mesh VPN between data centres, and allows nodes to automatically discover each-other.