Cephalocon 2024
| title | author | length |
|---|---|---|
| RBD in Squid and Beyond | Ramana Krisna Venkatesh Raja, Prasanna Kumar Kalever | 35m |
RBD in Squid shows concrete fixes and measurable wins: a diff-trace API optimization cuts a synthetic QEMU 500 GB backup from ~560s to ~3s with an exclusive… | ||
| CephBox - A Home Cloud Personal Storage Solution | Karun Josy | 10m |
A DIY home NAS that runs a small Ceph cluster on Raspberry Pi hardware, automated install scripts and a web dashboard, claims to make a 24×7 running Ceph… | ||
| Erasure Coding: 5 Ways to Split a Squid | Jamie Pryde | 12m |
Ceph supports multiple erasure-code plugins and benchmarks show a clear performance gap: using the Ceph erasure-code benchmark tool on 80 MB objects, ISA-L… | ||
| Keeping Ceph RGW Object Storage Consistent | Jane Zhu | 35m |
Bloomberg runs Ceph RGW across tens of clusters and four data centers serving >2 billion S3 requests/day and 77 PB raw; they see orphan multipart parts… | ||
| Beyond Particle Physics: The Impact of Ceph and OpenStack on CERN's Multi-D... | E. Bocchi, J.C. Leon | 41m |
CERN runs 100 PB Ceph for block, object and CephFS and an OpenStack cloud that exposes bare metal, VMs and virtual networking across more than 450k CPUs; the… | ||
| Ceph NVMe-of Road Map | Orit Wasserman, Aviv Caro | 38m |
Ceph NVMe-oF over TCP gateway status and roadmap: upstream is in technical preview for Riff/Squid and mainline is merged for Tentacle where it is production… | ||
| Crimson Project Update | Matan Breizman, Aishwarya Mathuria | 34m |
Crimson is a full rewrite of ceph-osd using the Seastar C++ framework and a single-thread per core shared-nothing model that eliminates locks and context… | ||
| Pull Requests and Reviews for Good | Gregory Farnum | 38m |
Explains how to prepare pull requests so they merge: clear goal and design, small logical commits, tests, user and developer docs, and supportability that… | ||
| D4N in RGW | Pritha Srivastava | 21m |
D4N is an intelligent RGW cache that is being upstreamed into Ceph and runs as a Zipper API filter with four drivers: filter, policy, directory, cache backend… | ||
| Keynote: Powering Modern Enterprise Data Storage with Ceph | Vincent Hsu, Neha Ojha | 16m |
Claims enterprise Ceph can replace cloud costs and legacy VMware block at scale: customers with 1 exabyte under VMware and another planning to move 300 PB say… | ||
| Bringing a Ceph Based Enterprise Email System Into the Field | Danny Al-Gaaf | 38m |
Deutsche Telekom migrates its email storage to Ceph using the open source librmb library that writes each message as a RADOS object and stores mutable… | ||
| The SMB Report Card | John Mulligan | 31m |
Adds managed SMB to Ceph by landing an SMB manager module inside cephadm orchestration, exposing imperative CLI and declarative YAML via a new seph smb apply… | ||
| Enhancing Observability and Monitoring for Large Ceph Clusters at Scale | Filipp Akinfiev | 32m |
Proposes a four-layer observability stack with nearby and remote producers, a NATS messaging backbone, and lightweight consumers that perform on-demand and… | ||
| Supporting 3 Availability Zones Stretch Cluster | Kamoltat (Junior) Sirivadhna | 25m |
Shows an implemented Ceph upstream fix that prevents a specific data‑loss edge case where a replication size 6 pool with min_size 3 can accept writes from a… | ||
| The Art of Teuthology | Patrick Donnelly | 40m |
The Ceph QA pipeline replaces a combinatoric YAML-matrix (example: a Ratt suite would enumerate ~2.7 million jobs, impossible on ~200 nodes) with a graph walk… | ||
| Keynote: Opening Remarks | Enrica Porcari, Jakub Mościcki | 13m |
CERN describes operational numbers and real deployments: SEF provides block storage since 2013 and now underpins hundreds of services across three… | ||
| Migrating AI/ML Data Pipelines for Real Time Streaming Applications from HDFS to Ceph | V. Alinouri | 36m |
Etraveligroup migrates 20 AI/ML streaming pipelines from HDFS to Ceph RGW on NVMe, testing two pipelines (50k and 20k records/min, 1 GB throughput, ~1s latency… | ||
| Keynote: Getting to Know Ceph | Josh Durgin, Neha Ojha, Dan van der Ster | 13m |
Ceph is presented as an open source, strongly consistent storage system that supports replication and erasure coding, provides rolling upgrades and online… | ||
| Ceph at 20 Years! Still the Best for Modern Storage | Dan van der Ster | 30m |
Ceph marks 20 years and claims continued primacy by combining a unified block/object/file stack with concrete systems: CRUSH consistent hashing for… | ||
| Ceph Manager Module Design and Operation, an in-Depth Review | Brad Hubbard, Prashant Dhange | 38m |
Ceph MGR architecture is described end-to-end: C++ core spawns a daemon server, Python modules load via P_module_load into per-module python subinterpreters… | ||
| Revisiting Ceph's Performance After 4 Years | Wido den Hollander | 23m |
Benchmarks compare Ceph 14.2 → 18.2 on Ubuntu 22.04 using 4K single-thread, QD1 workloads (60s runs) on three 1U servers with 6 OSDs (25 Gb, jumbo frames, BGP… | ||
| Exploring RocksDB in RGW: How We Manage Tombstones | Sungjoon Koh, Ilsoo Byun | 38m |
RocksDB tombstones in RGW cause list operations to stall: tens of millions of tombstones accumulate, iteration latency rises up to ~6,000× when non‑compacted… | ||
| Remote Replication in MicroCeph: RBD and Beyond | Utkarsh Bhatt | 20m |
MicroCeph Squid adds APIs to configure remote cluster awareness and perform RBD replication end-to-end using real Ceph tools: cluster export/import to share… | ||
| Improving Ceph Economics with QAT Hardware Offload | Philip Williams | 24m |
A four‑node Ceph cluster (each node dual Xeon 6548 with two QAT cores, 100 Gb networking, Ubuntu 24.04 LTS) uses Intel QAT offload and the minio warp workload… | ||
| CephFS: A Stroll Through History | Mattia Belluco | 24m |
CephFS timeline: Jewel declared CephFS stable (10th release, 2016) but lacks snapshots and multi-active maturity; Luminous adds BlueStore and enables… | ||
| Conditional End2end Tracing | Yuval Lifshitz, Deepika Upadhyay | 39m |
Conditional tracing uses Lua hooks on RGW to set a trace flag per request so OpenTelemetry spans are created only for targeted flows, avoiding a measured 5%… | ||
| Maximizing the Value of Your Rados Gateway with Ingress Strategies | Michaela Lang, Daniel Parkes | 43m |
Envoy sits in front of RADOS Gateway as an ingress proxy implementing distributed rate limiting via Envoy’s gRPC external rate‑limit service backed by Redis… | ||
| Follow the Advice, but Not Too Closely - Building the Cluster YOU Need | Gregory Orange | 11m |
A 20 PB object cluster built from 26 storage units (each ~60 × 22 TB drives, >1 PB raw) uses 24 active nodes with two hot spares reserved to avoid multi‑day… | ||
| The Challenge of Storing Small Objects on a Large Scale | Luis Domingues, Ján Senko | 33m |
Managing about 100 PiB across 6,000+ OSDs on Ceph exposes two hard facts: millions of tiny email headers blow up storage due to allocation and 4+2 erasure… | ||
| Keynote: State of the Cephalopod | Zac Dover, Redouane Kachach, Josh Durgin, Orit Wasserman, Ramana Raja, Venky Shankar, Travis Nielsen, Yaarit Hatuka, Matan Breizman, Mark Nelson, Nizamudeen A, Neha Ojha, Aviv Caro | 45m |
Roadmap-sized update details shipped and planned features: claim of 1 TB/s aggregate throughput on ~630 OSDs, release Squid and upcoming Tentacle with focus on… | ||
| Keynote: Taking Ceph to the Next Level | Mark Nelson | 06m |
Clyso describes concrete upgrades: a Nautilus cluster moves live from old HDD nodes to NVMe and upgrades to Quincy with minimal interruption and measurable… | ||
| Keynote: Welcome to Cephalocon 2024 | Josh Durgin, Neha Ojha, Dan van der Ster | 05m |
MCs announce the schedule and logistics for a two-day conference split across two buildings, with Keynotes here both days and plenary sessions upstairs and… | ||
| User Feedback: Shaping the Future of Ceph | Dan van der Ster | 34m |
Interactive BoF collects concrete Ceph deployment feedback: debates over adopting Cephadm versus DIY orchestration, custom containers and secrets/Vault… | ||
| Architecting Cloud Storage for AI Native Applications | Nathan Goulding | 14m |
Vultr presents an AI-native cloud storage stack with 32 global data centers and >60 PB of Ceph-backed storage deployed, mixing S3-compatible object, resizable… | ||
| Lessons From Running Ceph as a Platform at Scale | Jason Burris | 25m |
Sony Interactive Entertainment runs 60 clusters, 250 PB, 30,000 drives, 16 billion objects and finds scale forces concrete solutions: use Ansible + Prometheus… | ||
| Keynote: Ceph Foundation: Board Priorities and Engagement for 2025 | Enrico Bocchi, Josh Durgin, Vincent Hsu, Neha Ojha, Matthew Leonard, Joachim Kraftmayer, Dan van der Ster, Wout van Heeswijk | 18m |
Board outlines a tight 2025 plan with a small budget and limited volunteer hours and commits to hiring a dedicated community manager, updating the charter, and… | ||
| Deploy Ceph in Kubernetes: Rook Deep Dive | Travis Nielsen, Subham Kumar Rai | 35m |
Rook runs Ceph inside Kubernetes using CRDs and CSI drivers to give RWO RWX S3 storage, claims six years of production use, and targets two recent Ceph majors… | ||
| Async I/O with Libcephfs | Dhairya Parmar | 08m |
Libcephfs adds an asynchronous API used by NFS‑Ganesha that replaces blocking reads with non-blocking callbacks, letting the RPC scheduler suspend tasks and… | ||
| Bringing fscrypt Encryption to CephFS Userspace | Christopher Hoffman | 33m |
Brings kernel fscrypt support into userspace by adding libcephfs and ceph-fuse implementations, new ioctls (set/get V2) and a crypto keyring so directories can… | ||
| Ceph Operations Retrospective: Lessons from Scaling and Managing a Multi-PB Cluster | Mattias Skohg | 34m |
Ceph Pacific cluster uses CephFS and RGW, grows from 4.5 PB to ~9 PB raw across ~300 OSDs and adds nodes of ~500 TB each, ingesting ~40–55 TB/week; metadata… | ||
| Ceph Telemetry - The Why, What, and How | Yaarit Hatuka | 33m |
Ceph Telemetry collects anonymized cluster data from 3,400 clusters totaling about 1.7 exabyte and ~200,000 OSDs, split into five channels: basic, crash… | ||
| Benchmarking: Repeatable & Comparable | Trent Lloyd | 40m |
Practical checklist for storage benchmarking: use fio only (other tools mislead), precondition block devices and pre-write or thick‑provision VMs, run long… | ||
| Advancing BlueStore with Real-World Insights | Adam Kupczyk | 26m |
A Python scraper tool captures a 73-hour RBD trace (4 OSDs, ~300M ops) and reveals concrete opportunities: coalescing disjoint writes would touch ~4.6M extra… | ||
| High Performance Erasure Coding - the Story so Far | Connor Fawcett | 14m |
Work to make erasure coding perform near replica‑3 claims concrete: a Python offline consistency checker and C++ io exerciser plus an online checker validate… | ||
| Want to Become a Ceph Ambassador? | Gaurav Sitlani | 09m |
Program started by Mike Perez recognizes active Ceph community members and requires three ambassadors per region to ensure no single point of failure. Current… | ||
| Volume Shadow Copies with CephFS: A Samba-Centric Approach | Shwetha Acharya | 36m |
CephFS exposes snapshots as Windows Volume Shadow Copies using the vfs_ceph_snapshots Samba module, tested against a 3-node CephFS cluster and a Windows client… | ||
| Keynote: Open Source Community Engagement: We’re in. Are you? | Matthew Leonard | 18m |
Bloomberg says it treats open source as a coordinated community action and describes a ten plus year Ceph journey that begins with consumption → participation… | ||
| Welcome Back & Remarks | Enrico Bocchi | 04m |
Housekeeping notes cover sponsors and logistics: platinum sponsors Bloomberg Engineering and IBM, gold and silver sponsors, and a sponsor showcase in… | ||
| Characterizing Crc Operation on Ceph | Rixin Luo | 15m |
Ceph CRC32C work is dominated by many very small operations: for RBD 4K random writes 80% of CRCs are smaller than 292 bytes and for 4K reads 83% are smaller… | ||
| Pulpito 2.0: A Tour of Pulpito Next-Generation | Vallari Agrawal | 14m |
Pulpito-NG adds one-click run termination from the web with a kill run button that only appears for the run owner or users with admin rights (about 200-team… | ||
| Hyperconvergence Using Proxmox and Ceph; A Journey Through Data and Time | Iztok Gregori | 32m |
A Proxmox + Ceph hyperconverged setup grows from 8 to 16-node hyperconverged cluster, hosting ~120 VMs on ~600 cores and 4 TB RAM with 150 TB HDD and 240 TB… | ||
| Archive Zone: Lessons Learned | Ismael Puerto Freire, Xabier Guitián Domínguez | 20m |
Ceph archive cluster for S3-style objects stores 75M objects using replication 3 across HQ and two sites with ~55 ms latency; initial hardware used flash + big… | ||
| State of CephFS: Three Easy Pieces | Venky Shankar, Patrick Donnelly | 43m |
CephFS status report focuses on concrete improvements: MDS journal minor segment support reduces subtree-map churn and speeds trimming; a manager plugin… | ||
| SWITCH: Operations, Data Management and Automation | Theofilos Mouratidis | 27m |
SWITCH runs three Ceph generations and claims OSv3 replaces brittle scripts with a GitOps pipeline using Terraform, GitLab CI/CD and an OSv3 Ansible collection… | ||
| Understanding Ceph One Performance Counter at a Time | Marcel Lauhoff | 32m |
Raw Ceph perf counters come straight from the daemon admin socket via perf dump and a supplied GitHub notebook with jq commands; an ad‑hoc web app extracts… | ||
| Keynote: Ceph Developers AMA: Meet the Minds Behind Open Source Ceph | Zac Dover, Redouane Kachach, Josh Durgin, Orit Wasserman, Ramana Raja, Venky Shankar, Travis Nielsen, Yaarit Hatuka, Dan van der Ster, Matan Breizman, Mark Nelson, Nizamudeen A, Neha Ojha, Aviv Caro | 48m |
mclock is now the default OSD scheduler but teams report rough edges, so defaults move from high client IO to balanced, recovery prioritization and device… | ||
| Get That Cluster Back Online, but Hurry Slowly | Gregory Orange | 31m |
Pawsey’s petabyte Ceph RGW cluster suffers a months-long degradation with one cluster unhealth for seven weeks, OSDs holding 50–90 PGs per OSD instead of the… | ||
| The ‘Scrub-Type to Limitations’ Matrix | Ronen Friedman | 36m |
A concrete matrix maps six scrub urgency levels to which limits each type honors; Squid implements many changes and Tentacles finalizes them. Key points… | ||
| Cost-Effective, Dense, and Performant Prometheus Storage via QLC | Anthony D'Atri | 40m |
An example deployment scrapes 8,000 targets at 6–10s intervals and uses two 15TB QLC NVMe drives for WAL and long-term storage, reports 0.09 drive writes per… | ||
| Ceph Made Easy: One Dashboard for Multiple Ceph Clusters | Nizamudeen A | 27m |
A dashboard extension that lets one Ceph cluster act as a hub cluster to monitor and manage many others is shown, using Ceph REST APIs for control and… | ||
| Backing up CephFS | Abhishek Lekshmanan, Roberto Valverde Cameselle | 40m |
CERN backs selected CephFS shares with restic plus an orchestrator, avoiding snapshots for performance reasons; a 2 PB dataset contains roughly 400 million… | ||
| Encrypting Data in Ceph: An Overview and Opportunities | David Mohren | 27m |
A clear proposal to add AEAD ciphers to Ceph and store authentication tags in OMAP for auth tags is presented with a working proof‑of‑concept in the user‑space… | ||
| Arming Ceph: CI Enhancement and UADK Integration | Rongqi Sun | 10m |
ARM CI gets environment fixes and faster pipelines with cleaned legacy code and uniform OS images, more than 20 tickets closed and flaky failures like the… | ||
| From ISCSI to NVMe-of: A Comparative Look on Storage Protocol Advantages & Challenges | A. Muthmann | 30m |
Direct comparison shows RBD usually wins for throughput/IOPS/latency, with NVMe-oF typically faster than iSCSI but slower than raw RBD except in small‐dataset… | ||
| Ceph Notification at Scale: Notification V2, an Enterprise-Level Feature | Krunal Chheda | 36m |
Ceph’s notification_v2 adds concrete fixes: topic ownership and permissions to prevent cross-tenant leakage, multisite syncing so a single topic/notification… | ||
| nothing matches | ||