[LSF/MM TOPIC] ZUFS - How to get ZUFS ready for upstream

[Boaz Harrosh <openosd@gmail.com>]

From: Boaz Harrosh <boazh@netapp.com>

Hello fellow Linuxers

ZUFS has been proven to be very capable and very fast - though not as fast as I would like
and work is done to still improve on that.

And very stable. Its been in testing and QA labs for a year. And finally has been
released for paying costumers to help them with the extreme workloads they demand.
The release is based on RHEL 7.6. The Kernel + user-mode Server are open source.
The actual FS provider is closed-source.

So its been capable and reliable to be trusted with some type of real user data
and processing.

But is it secure? Is it capable to maintain the Kernel hardening.

I would please like to give a 5 minutes overview of the project. And then
leave the entire talk to discussion and investigation about the different
parts.

I would like to advise with the Linux good folks how to move forward and get
this project forward, to be ready for an upstream Kernel near us?
What do we need to test and prove for this to be considered good enough?

I would like to make a short pitch of why the LSF/MM guys should care about
ZUFS at all.
Well because what is introduced here is a very capable very new way to interface
user-mode from Kernel and/or app-to-app. Which by taking a deep consideration of
modern multi-core, multi-NUMA machines. Enables performance never seen before
from an application server.
We believe Filesystems are only the first to benefit from this and more subsystems
and projects could be extended to use the patterns / code introduced here.

Thanks for listening
Boaz