The document provides an agenda and overview of the Ceph Project Update and Ceph Month event in June 2021. Some key points: - Ceph is open source software that provides scalable, reliable distributed storage across commodity hardware. - Ceph Month will include weekly sessions on topics like RADOS, RGW, RBD, and CephFS to promote interactive discussion. - The Ceph Foundation is working on projects to improve documentation, training materials, and lab infrastructure for testing and development.

1. 1
Ceph Project Update
Sage Weil & Josh Durgin
Ceph Month - 2021.06.01

2. 2
● Ceph background
● Ceph Month
● Ceph Foundation update
● Sepia lab update
● Telemetry
● The future
AGENDA

3. 3
The buzzwords
● “Software defined storage”
● “Unified storage system”
● “Scalable distributed storage”
● “The future of storage”
● “The Linux of storage”
WHAT IS CEPH?
The substance
● Ceph is open source software
● Runs on commodity hardware
○ Commodity servers
○ IP networks
○ HDDs, SSDs, NVMe, NV-DIMMs, ...
● A single cluster can serve object,
block, and file workloads

4. 4
● Freedom to use (free as in beer)
● Freedom to introspect, modify,
and share (free as in speech)
● Freedom from vendor lock-in
● Freedom to innovate
CEPH IS FREE AND OPEN SOURCE

5. 5
● Reliable storage service out of unreliable components
○ No single point of failure
○ Data durability via replication or erasure coding
○ No interruption of service from rolling upgrades, online expansion, etc.
● Favor consistency and correctness over performance
CEPH IS RELIABLE

6. 6
● Ceph is elastic storage infrastructure
○ Storage cluster may grow or shrink
○ Add or remove hardware while system is
online and under load
● Scale up with bigger, faster hardware
● Scale out within a single cluster for
capacity and performance
● Federate multiple clusters across
sites with asynchronous replication
and disaster recovery capabilities
CEPH IS SCALABLE

7. 7
CEPH IS A UNIFIED STORAGE SYSTEM
RGW
S3 and Swift
object storage
LIBRADOS
Low-level storage API
RADOS
Reliable, elastic, distributed storage layer with
replication and erasure coding
RBD
Virtual block device
CEPHFS
Distributed network
file system
OBJECT BLOCK FILE

8. 8
RELEASE SCHEDULE
Octopus
Mar 2020
14.2.z
Nautilus
Mar 2019
WE ARE
HERE
15.2.z
16.2.z
Pacific
Mar 2021
17.2.z
Quincy
Mar 2022
● Stable, named release every 12 months
● Backports for 2 releases
○ Bug fixes and security updates
○ Nautilus reaches EOL shortly after Pacific is released
● Upgrade up to 2 releases at a time
○ Nautilus → Pacific, Octopus → Quincy
● Released as packages (deb, rpm) and container images
● Process improvements (security hotfixes; regular cadence)

9. 9
CEPH MONTH - JUNE 2021

10. 10
CEPH MONTH
● Goals
○ More interactive
○ Bite-sized
● Format
○ 1-2 hrs
○ ~2 blocks per week
○ A few planned talks
○ Un/semi-structured discussion time
○ Lighting talks sprinkled throughout
● Etherpads
○ Add your questions, or ask them verbally
○ Add any discussion topics
● Week of June 1 - 4
○ RADOS
○ Windows
● Week of June 7 - 11
○ RGW
○ Performance
● Week of June 14 - 18
○ RBD
○ Dashboard
○ Lighting talks
● Week of June 21 - 25
○ CephFS
○ cephadm
https://pad.ceph.com/p/ceph-month-june-2021

11. 11
● It will be in March 2022…
● No location yet
○ Seoul?
○ North America? (Portland?)
○ ???
● Expected to be in-person
○ Possibly with hybrid elements?
● We are very interested in community feedback!
CEPHALOCON 2022

12. 12
CEPH FOUNDATION UPDATE

13. PREMIER MEMBERS

14. GENERAL MEMBERS

15. ASSOCIATE MEMBERS

16. 16
CURRENT PROJECTS
● Ceph documentation
○ Zac Dover, full-time technical writer
● ceph.io web site update
○ Spearheaded by SoftIron
○ Static site generator; github; no more wordpress
○ https://github.com/ceph/ceph.io
○ Planned launch next month!
● Training materials
○ Working with Linux Foundation’s training group
○ Building out initial free course material (w/ JC Lopez)
○ edX and/or LF hosted; can support both self-paced or instructor-led
○ Potential in future for advanced material, paid courses, and/or certifications
○ LF training group is revenue neutral; collaborative development process with community

17. 17
CURRENT PROJECTS
● Reducing cloud spend with OVH
○ Build and CI hardware purchases for Sepia lab
○ We are now only hosting public-facing infra in OVH
● Lab hardware
○ Build machines
○ Expanding lab’s Ceph cluster (more storage for test results, etc)
● Windows support
○ Contract with CloudBase to finish initial development, build sustainable CI infrastructure
○ RBD, CephFS
● New marketing committee

18. 18
SEPIA LAB UPDATE
● More hardware from the Ceph Foundation
○ Expanding the lab’s Ceph cluster
○ More build machines (braggi)
○ More test nodes (gibba)
● Improved teuthology test infrastructure
○ Moved to a single process dispatcher (Shraddha Agrawal)
○ Replaced in-memory queue with limited features with postgres (Aishwarya Mathuria)
○ Enables larger scale test clusters
○ Ability to prioritize and use lab more efficiently
● Downgrade testing (WIP)
○ Downgrade within a major release (e.g. 16.2.4 -> 16.2.3)
○ Now feasible with cephadm

19. 19
ARM AARCH64 SUPPORT
● Hardware donated by Ampere
● CI builds for teuthology, releases
○ CentOS 8 RPMs, Ubuntu Focal 20.04
○ Container images (based on CentOS)
● Addressing some issues with bleeding edge of podman/quay and multi-arch
support

20. 20
TELEMETRY UPDATE
https://telemetry-public.ceph.com/

21. 21
TELEMETRY AND CRASH REPORTS
● Opt-in
○ Will require re-opt-in if telemetry content
is expanded in the future
○ Explicitly acknowledge data sharing
license
● Basic channel
○ Cluster size, version
○ Which features are enabled
● Crash channel
○ Anonymized crash metadata
○ Where in the code the problem happened,
what version, etc.
○ Extensive (private) dashboard
○ Integration into tracker.ceph.com WIP
● Device channel
○ HDD vs SSD, vendors, models
○ Health metrics (e.g., SMART)
○ Extensive dashboard (link from top right)
● Ident channel
○ Off by default
○ Optional contact information
● Future performance channel
○ Planned for quincy
○ Optional more granular (but still
anonymized) data about workloads, IO
sizes, IO rates, cache hit rates, etc.
○ Help developers optimize Ceph
○ Possibly tuning suggestions for users
● Transparency!
https://telemetry-public.ceph.com/

22. 22
IS TELEMETRY ENABLED?

23. 23
WHY IS TELEMETRY NOT ENABLED?

24. 24
IT’S EASY!
● Review and opt-in
● Enable SOCKS proxy
● https://docs.ceph.com/en/latest/mgr/telemetry/

25. 25
THE FUTURE...

26. 26
● Cephadm has brought end-to-end management of Ceph deployments
● Cluster management via Ceph dashboard
● Simple experience for non-enterprise deployments
○ Small/medium businesses, remote offices, etc.
○ NAS replacement
● Turn-key support for NFS, object
○ SMB coming in Quincy
OUT OF THE BOX EXPERIENCE

27. 27
NEW DEVICES
● ZNS SSDs
○ 3D NAND … dense, but the erase blocks are huge
○ Zone-based write interface
○ Combines capacity, low cost, and good performance
○ Key focus of Crimson’s SeaStore!
● Multi-actuator HDDs
○ Recent devices double IOPS in existing HDD package
○ Ceph treats them as two OSDs with shared failure domain
● Persistent memory
○ Will be well-supported (but not required) by Crimson
○ Recent support in RBD client-side write-back cache

28. 28
● Client-side
○ NVMeoF target that presents an RBD device
○ Alternative to iSCSI
○ Can be combined with new hardware (e.g., SmartNICs like Nvidia’s Bluefield) to present a
NVME device on PCI bus while running gateway/librbd code on the card’s “DPU”
○ Useful for “metal as a service” cloud infrastructure
● Server-side
○ Some discussion around Crimson “phase 2”
○ Enable primary OSD to write directly to replica OSD’s devices
○ Mechanism to reduce CPU cost per IO
NVMe FABRICS

29. 29
● Maturing
○ Rook
■ Key focus: Ceph orchestrator / dashboard integration with rook
○ Knative
○ Spark
■ S3 SELECT
○ Multisite
■ interop with public cloud
● New
○ Apache Arrow / Parquet
■ Data interchange formats for data pipelines
INTEGRATIONS / ECOSYSTEMS

30. 30
https://pad.ceph.com/p/r
NAMING THE R RELEASE

31. Questions?
Up next: RADOS
https://pad.ceph.com/p/ceph-month-june-2021