How to Take a Backup in a Distributed System Without Anyone Noticing
summary
Describes how GridGain (built on Apache Ignite) handles creating backups of large, constantly modified distributed datasets—up to 200 TB across 160 nodes—despite Ignite lacking built‑in backup capabilities. Explains the problem of taking unobtrusive backups in a transactional, high‑write distributed system and outlines the practical approach/tools used to capture consistent backups without noticeable impact.
AI-written summary · about
chapters (10)
- 0:00Apache Ignite intro
- 2:53Backup issues in distributed systems
- 7:51Existing backup approaches
- 12:09Cluster deactivation for backup
- 14:00Local consistency without stop
- 25:50Backup checkpoint and global consistency
- 34:04Incremental backups and space savings
- 36:31Handling changing topology
- 39:42Backup without load stop
- 45:07Q&A
related talks
47mAn Algorithm for Incremental Backups in Apache Ignite
49mBeyond the Data Grid: Fast Data Processing with Apache Ignite
41mDistributed Java DBs Under the Hood: Main Components and Interactions Between Them
27mBackup and Restore: Discover Best Practices for a Distributed Database Backup Strategy
06mIGNITE: Schrödinger’s backups: how to avoid uncertainty
33mDistributed Database DevOps Dilemmas? Kubernetes to the rescue!
10mLightning Talk: Instant Backups and Restores
32mDistributed Database DevOps Dilemmas? Kubernetes to the Rescue