The 'Ceph' Community Now Stores 1,000 Petabytes in Its Open Source Storage Solution

1,000 petabytes.
A million terabytes.
One quintillion bytes (or 1,000,000,000,000,000,000).

That's the amount of storage reported by users of the Ceph storage solution (across more than 3,000 Ceph clusters).

The Ceph Foundation is a "directed fund" of the Linux Foundation, providing a neutral home for Ceph, "the most popular open source storage solution for modern data storage challenges" (offering an architecture that's "highly scalable, resilient, and flexible"). It's a software-defined storage platform, providing object storage, block storage, and file storage built on a common distributed cluster foundation.

And Friday they announced the release of Ceph Squid, "which comes with several performance and space efficiency features along with enhanced protocol support." Ceph has solidified its position as the cornerstone of open source data storage. The release of Ceph Squid represents a significant milestone toward providing scalable, reliable, and flexible storage solutions that meet the ever-evolving demands of digital data storage.

Features of Ceph Squid include improvements to BlueStore [a storage back end specifically designed for managing data on disk for Ceph Object Storage Daemon workloads] to reduce latency and CPU requirements for snapshot intensive workloads. BlueStore now uses RocksDB compression by default for increased average performance and reduced space usage. [And the next-generation Crimson OSD also has improvements in stability and read performance, and "now supports scrub, partial recovery and osdmap trimming."]

Ceph continues to drive the future of storage, and welcomes developers, partners, and technology enthusiasts to get involved.
Ceph Squid also brings enhancements for the CRUSH algorithm [which computes storage locations] to support more flexible and cost effective erasure coding configurations.

I actually am curious what the estimate is

[NFN_NLN]

How many petabytes of data do pirates store in "torrents"?

Re:

[test321]

You are right that the comparison to ship attacks and aircraft hijacking did not make sense. However nowadays even the people who most dislike copyright law and actively fight it identify with the word, calling themselves pirates and getting some pride out of it; see there is The Pirate Bay (founded 2003); and there is the Pirate Party (founded 2010), registered across the globe.

Also, which other short word do you suggest for copying contents in a way that you are not allowed to do? Independently on your op

Re:

[zenlessyank]

I prefer the term 'backup'.

Excited For Ceph

[organgtool]

A colleague and I are in the process of building a Proxmox cluster which uses Ceph to host all of our VMs. We've never built something like this from the ground up but we're both very excited by the many advantages these technologies provide over running on bare metal. The performance of our prototypes have been pretty impressive and we can't wait to have the full-scale system up and running. I'll have to see if we should wait to upgrade to the new version of Ceph before continuing the deployment.

Re:

[ls671]

I have a proxmox cluster across several datacenters and I use hourly snapshots with zfs replication every hour, easier to get better performance that way, I have looked into ceph although, maybe some day if the use case arises.
 

Re:

[volcan0]

unless you need the scale out capabilities, I would recommend using a different storage system.

I looked into Ceph for home storage

[JamesTRexx]

Currently running two PCs with a bunch of disks in RAID1 each, and thought of using Ceph to replace it and have a copy of all data on each PC while still having the checksum data integrity. For now I copy the most important stuff daily with rsync from one to the other, but I like to see it part of the underlying distributed filesystem and be as hands-off as possible.

Unfortunately, Devuan stable at the time didn't have the correct version for it, which it has now, but setting it up has to be done manually and there's no init script as far as I can tell. Only support for systemd which is a dealbreaker for me.
Haven't checked up on alternatives yet, though.

Re: I looked into Ceph for home storage

[griffo]

Just use Proxmox on all your machines. Then you can leverage zfs dedup too, run Cephalon on all the machines and run whatever you need besides it. Levelonetech posted some videos that proxmox even utilizes e- and p-cores really well.

Re:

[JamesTRexx]

Thought about that since I've used Proxmox until 2016 (also as personal workstation), but it's based on Debian with systemd (unless that has changed), and ZFS is a memory hog when used as it should be. My PCs are simple (and old) builds.

It's a shame I have no good reason to run virtual machines any longer because I liked to play around and run tests at work with my Proxmox workstation updated from a custom Debian install.

That's 1 Exabyte

[Dishwasha]

1 Exabyte only? That's so cute. Datacenters literally shred Exabytes per day. The Chia blockchain has over 30 Exabytes of storage in use. I'm thinking this is seriously underreported.

Re:That's 1 Exabyte

[Nite_Hawk]

I'm one of the Ceph developers. Indeed, there are a number of large sites that don't report their numbers. Also, I'm pretty sure we hit 1 EiB being report a while ago. Telemetry was showing 1.44EiB as of a couple minutes ago:

https://telemetry-public.ceph.... [ceph.com]

Re:

[Dishwasha]

Doesn't CERN still use Ceph? That's got to be several Exabytes alone....

Re:

[Deaddy]

CERN's main storage is using EOS:

https://github.com/cern-eos/eo... [github.com]

Re: That's 1 Exabyte

[Guy Smiley]

There are single Lustre filesystems that are 700PB in size [hpe.com]...

Exabyte Tapes

[Roger W Moore]

1 Exabyte only? That's so cute. Datacenters literally shred Exabytes per day.

That's nothing, we had Exabyte tapes back in the 1990s.

Squid?

[zdzichu]

Ceph docs shows no such release (https://docs.ceph.com/en/latest/releases/). Is this and AI-hallucinated press release?

Re:

[echo123]

Is this an AI-hallucinated press release?

Per TFA, Squid has only been announced, not yet released.

Features of Ceph Squid include:

- Improvements to BlueStore to reduce latency and CPU requirements for snapshot intensive workloads.
- BlueStore now uses RocksDB compression by default for increased average performance and reduced space usage.
- Enhancements for CRUSH to support more flexible and cost effective erasure coding configurations.
- Much improved read performance and stability with Crimson. It now supports scrub, partial recovery and

Half my porn collection!

[BenJeremy]

Now, where do I store the other half?