Johann Lombardi

[daosio @daos_io]

Check out the exciting agenda for DAOS User Group (DUG) which brings together members from across the DAOS community. The event is on Nov 16th and co-located with #Supercomputing25. No registration required - view agenda | https://t.co/gwIDhFoqJD
#DAOSIO #HPC #OpenSource

[daosio @daos_io]

Check out this video to learn how DAOS, the high-performance open-source storage stack, helps streamline data management for HPC and AI workloads on @argonne's Aurora exascale supercomputer. | https://t.co/1WIYveBYmt
#DAOS #DAOSIO #HPC #OpenSource

[Johann Lombardi @__jlo]

@JeffDenworth @glennklockwood I guess the next question is whether people usually laugh at your jokes :)

[Johann Lombardi @__jlo]

@JeffDenworth "Getting high" too often lately maybe? :)

[Johann Lombardi @__jlo]

@FearFactory at @rockhal_lu

[Johann Lombardi @__jlo]

@lustre_one's famous Lustre installation at home is going to have some serious competition soon ... maybe we need a new #io500 category ...

[Johann Lombardi @__jlo]

@adrianjhpc @tgamblin A major protocol change breaking wire interoperability.

[Johann Lombardi @__jlo]

@glennklockwood @adrianjhpc Lol

[Glenn K. Lockwood @glennklockwood]

Many parallel data servers with a single metadata server. Sound familiar? pNFS being twenty years old is not a good thing.

[Johann Lombardi @__jlo]

@echo5juliet @HPC_Guru @googlecloud @glennklockwood @hei_nyan Not really.

[Johann Lombardi @__jlo]

@_koyamaso Got it. In early io500 submissions that we did with optane for both data and metadata, we also noticed that ior hard write was better than ior easy write. But reads were still better for easy. https://t.co/enW6Sb5pUL

[Johann Lombardi @__jlo]

@_koyamaso I see. So it uses optane only for data? Interesting that your ior hard results are better than the ior easy ones ...

[HPC Guru @HPC_Guru]

Congratulations, @daos_io, on #1 and #2 on the @IO500benchmark production list

And to the folks behind #Aurora #storage for the stellar results : @intelhpc, @argonne_lcf, @HPE_Cray

#SC23 #HPC

[Adrian Jackson @adrianjhpc]

The DAOS object store at @argonne topping the @IO500benchmark production list with some huge I/O numbers

[Bilel Hadri @mnoukhiya]

A winner 🏆 in most of the categories @IO500benchmark is DAOS mainly on Aurora @argonne @argonne_lcf @intelhpc @HPE_Cray

Kudos. #SC23 #iamhpc

[Bilel Hadri @mnoukhiya]

The @IO500benchmark analysis update #SC23 #iamhpc

More and more DAOS, more submissions, over 200 entries!

[Johann Lombardi @__jlo]

@davidchapa @glennklockwood The latest version of @daos_io (i.e. v2.4) can run without optane. Each DAOS engine can keep an uptodate copy of the metadata trees on one or multiple local SSDs. See the presentation from Michael Hennecke at ISC'23: https://t.co/H2FBJ5Uigo

[Adrian Jackson @adrianjhpc]

No performance results before the @IO500benchmark is released, but it looks like @argonne will have some big performance numbers with DAOS this year (courtesy of @kevinha62412862 at #DUG23) #SC23

[Johann Lombardi @__jlo]

@glennklockwood A question to ask at the DAOS User Group that starts in a few minutes ...