Reltio Multi-Tenant Master Data Management DataStax Architecture Review

Reltio is a cloud-based MDM solution that unifies and cleanses multisource, complex data — CRM contacts and leads among it — into a single source of trusted information.

Reltio chose DataStax Enterprise Cassandra as the primary data store for its Cloud-native, metadata-driven model and operations, for its high performance columnar storage engine, fault tolerance, linear scalability and built-in multi-datacenter replication.

DataStax brought RhinoSource into Reltio to conduct a full review of that deployment on Amazon Web Services, covering the AWS architecture, Cassandra workloads, data model, code and cluster maintenance operations.

RhinoSource worked closely with the Reltio DevOps team and assembled the DataStax Enterprise operations runbook to lock in best practices. We also developed a statistical model showing that the optimum number of Cassandra vNodes — the num_tokens parameter — for maximising availability on large clusters while retaining some load balancing sits in the 2-8 range. That finding was later reflected in DataStax's own default guidance.

RhinoSource assisted with load and stress testing to determine the best Cassandra data model design and storage engine configuration, and made a number of recommendations to improve performance and reduce the volume of Cassandra data stored — and with it the total number of cluster nodes required — including the use of Solr Search indexes.

Finally, RhinoSource assisted with the redesign of a scalable multi-tenant data model and developed an automated data migration and single-tenant backup/restore process using Spark Scala jobs.