Erasure Code in Ceph
•
2 likes•3,442 views
Erasure code saves storage space and increases durability. Most commonly used in its simplest form (RAID5), erasure code is generalized for networked storage. It was first introduced in Ceph with the Firefly release in 2014. In this session, we’ll explain, with a lively experiment, the erasure code logic. We’ll give you an overview of the Ceph implementation and its most common use cases and go on to explain the various erasure code plug-ins (jerasure, ISA-L, LRC). Finally, we’ll show you how to estimate the trade off between erasure coding and replication—a choice that comes down to space versus speed.
Recommended
More Related Content
More from Red_Hat_Storage
More from Red_Hat_Storage (20)
Recently uploaded
Recently uploaded (20)
Erasure Code in Ceph
- 1. Erasure Code in Ceph Loic Dachary @ Red Hat
- 2. Save Space
- 3. 5 minutes role playing game
- 4. XOR
- 5. 3 peta => 1.3 peta
- 6. Harder object mutations / recovery
- 7. Simple operations and tiering Replicated Erasure Coded
- 8. Promoted to replica on read Replicated Erasure Coded
- 9. Sam & David : internals
- 10. Janne & Andreas & Loic & Takeshi erasure code
- 11. Released May 2014 : Firefly
- 12. Why save space before shortage ?
- 13. Reliability Model ● After an OSD is lost ● Recovery ● Backfilling a new OSD
- 14. April 2015 : Hammer
- 15. Repair K=10, M=4
- 16. Locally Recoverable Codes LRC @ Red Hat
- 17. SHEC Takeshi @ Fujitsu
- 18. ISA plugin Yuan @ Intel SIMD aka SSE2, SSE3, SSE4 Only for Intel processors ~50% Faster
- 19. jerasure ARM Janne @ ARM
- 20. Infernalis
- 21. Hitchhiker Rashmi @ U.C. Berkeley