Senior Software Engineer at Grammarly, where he leads the Core services team. For the last six years, develops natural language processing systems, currently running them in the cloud for millions of users. As a long-time on-call duty warrior, interested in making software simple and reliable.
Engineer with 15 years of experience. Most recent areas of interest: databases, distributed systems and performance. At Grammarly Igor is responsible for users data, authentication, statistics.
If you run a production database, there comes a time when you want to distribute it across multiple machines. Maybe your load grows bigger than a single machine can handle. Maybe you need fault tolerance – continue working if one machine or even an entire datacenter goes down.
For years, traditional master-slave replication was the primary choice for scaling relational databases. In this talk, we demonstrate Galera Cluster – a technology for Multimaster replication, available for MySQL and MariaDB. For many organizations and workloads, it is a sweet spot between high availability and ease of use.
What to expect from this talk:
– A case study of using Galera Cluster as a production database at Grammarly for several years.
– The key concepts of Galera Cluster and synchronous replication. Is it real?
– Demos that show pitfalls and best practices when dealing with primary keys, locks and schema change.
All rights reserved JEEConf ©
