Intel Releases DAOS 2.2 Distributed File-System

[phoronix]

Phoronix: Intel Releases DAOS 2.2 Distributed File-System

Intel earlier this year more formally announced DAOS as its distributed parallel file-system designed for NVMe storage and aims to be more efficient than other parallel file-systems. Yesterday marked the release of DAOS 2.2 as the newest step forward for Distributed Application Object Storage...

https://www.phoronix.com/news/Intel-DAOS-2.2-File-System

[stormcrow]

What is it with Intel and these useless benchmarks they keep marketing (regarding the pictures)? They're as bad as Apple.

[elatllat]

I wonder how DAOS compares to ceph, etc...

[CommunityMember]

What is it with Intel and these useless benchmarks...

No organization that actually run high end distributed file systems use vendor benchmarks (those are intended for marketing to the press and tech n00bs (who almost always take the bait)). The organizations that operate such file systems have teams of people who do their own benchmarks for their own workloads.

[numacross]

I wonder how DAOS compares to ceph, etc...

DAOS seems to have some very specific requirements:

DAOS requires each storage node to have direct access to storage-class memory (SCM). While DAOS is primarily tested and tuned for Intel Optane^TM^ Persistent Memory, the DAOS software stack is built over the Persistent Memory Development Kit (PMDK) and the Direct Access (DAX) feature of the Linux operating systems as described in the SNIA NVM Programming Model. As a result, the open-source DAOS software stack should be able to run transparently over any storage-class memory supported by the PMDK.

The storage node can optionally be equipped with NVMe (non-volatile memory express)[^10] SSDs to provide capacity. HDDs, as well as SATA andSAS SSDs, are not supported by DAOS. Both NVMe 3D-NAND and Optane SSDs are supported. Optane SSDs are preferred for DAOS installation that targets a very high IOPS rate. NVMe-oF devices are also supported by the userspace storage stack but have never been tested.

Which limits it to very high end enterprise hardware, at least officially.

Ceph by comparison can work over wider array of storage types.

[stormcrow]

No organization that actually run high end distributed file systems use vendor benchmarks (those are intended for marketing to the press and tech n00bs (who almost always take the bait)). The organizations that operate such file systems have teams of people who do their own benchmarks for their own workloads.

Duh. But this is not a product aimed at n00bs nor even the tech press. It's aimed at people that actually handle these kinds of systems day in and day out. Hence the question. Is Intel really this out of touch with customers that they think these useless benchmarks have any impact on adoption or is this strictly aimed at "good copy" for ignorant shareholders? I'm leaning towards the latter. (And it's really not even going to move shareholders either.)

[zexelon]

Wait... arent they dumping optain??!! The first words in their documentation is "optimized for optain"....

Always interested in seeing more development in this area but i think I will stick to Glusterfs.

[F.Ultra]

Duh. But this is not a product aimed at n00bs nor even the tech press. It's aimed at people that actually handle these kinds of systems day in and day out. Hence the question. Is Intel really this out of touch with customers that they think these useless benchmarks have any impact on adoption or is this strictly aimed at "good copy" for ignorant shareholders? I'm leaning towards the latter. (And it's really not even going to move shareholders either.)

prob aimed at both shareholders and PHB:s, you have to consider that few enterprises have their technical staff make the technical decisions, most are done by the PHB:s who get sold on some glossy looking paper like these benchmarks.

[onlyLinuxLuvUBack]

Wait... arent they dumping optain??!! The first words in their documentation is "optimized for optain"....

Always interested in seeing more development in this area but i think I will stick to Glusterfs.

they meant: optimized for un-obtainium storage

[coder]

Speaking of Optane, I just noticed Intel lists the P5810X as Launched in Q4 2022!

Appears to be the same as P5800X, except for higher write IOPS and peak power utilization.

https://ark.intel.com/content/www/us...=201861,229114

I think this was probably meant to be the Axboe Special Edition.
; )