Corundum developer meeting September 4, 2024: Sustainability of Corundum

[Alex Forencich]

As a reminder, the next Corundum developer meeting will be taking place
as scheduled on Wednesday September 4, 2024 at 9:00 AM pacific time
(16:00 UTC).

Now that SIGCOMM is behind us (where we presented the Realizing RotorNet
paper) and the research project at UCSD has more or less wound down,
it's a good opportunity to discuss how to make Corundum a sustainable
project in the long term.  Please tune in to the discussion if you can,
but I am certainly open to offline discussions as well.

Summary: Corundum is a large and complex project that likely will
require several people working full-time for continued maintenance and
development in the long term.  I think that the best way to make this
work is for the project to be self-sustaining, and unfortunately I don't
think that is going to be possible with the current permissive license. 
Therefore, I am planning a rewrite of the works (corundum and all of the
verilog libraries) in System Verilog under the CERN OHL strict license
(similar to GPL) + a CLA, with an option for a paid commercial license. 
A rewrite also provides the opportunity to make deeper architectural
changes, including supporting higher link rates including potentially
400G, higher PCIe rates including potentially gen 5 x16, improved
descriptor format and descriptor handling, descriptor inline data,
improved driver interface, improved time synchronization capabilities
including white rabbit and PTM, SR-IOV, RDMA, XDP, possibly VFIO, and
much more.


Calendar:
https://calendar.google.com/calendar/embed?src=6784b9ed971221ccfe78513753b3743d24db63ef75a657bdddade1f5e2fd9d2c%40group.calendar.google.com


Meeting information:
https://github.com/corundum/corundum/wiki/Corundum-Developer-Meeting

Zoom link:
https://ucsd.zoom.us/j/94871106964?pwd=dHg2UGU3WDBXOHU0NEdITkVYdGVSUT09

Meeting ID: 948 7110 6964

Password: rdma

--
Alex Forencich

[Alex Forencich]

As a reminder, the next Corundum developer meeting will be taking place

as scheduled on Wednesday October 2, 2024 at 9:00 AM pacific time (16:00
UTC).

For this week I am going to discuss the latest status updates on
Corundum, including the latest status of the OCP WR NIC and provide some
initial high-resolution measurements of PTP performance.

[Alex Forencich]

As a reminder, the next Corundum developer meeting will be taking place

on Wednesday November 20, 2024 at 9:00 AM pacific time (16:00 UTC).

For this week I am going to discuss the latest status updates on

Corundum, including the latest status of the OCP WR NIC, details about
the Sync-E demo from ISPCS/OCP summit, status on a port for the SE120
board, and more potential time-sync-related improvements.

[Alex Forencich]

As a reminder, the next Corundum developer meeting will be taking place

on Wednesday December 4, 2024 at 9:00 AM pacific time (16:00 UTC).

For this week I am going to discuss the latest status updates on

Corundum, including the latest status of the System Verilog rewrite,

details about the Sync-E demo from ISPCS/OCP summit, status on a port
for the SE120 board, and more potential time-sync-related improvements.

[Alex Forencich]

As a reminder, the next Corundum developer meeting will be taking place

on Wednesday February 5, 2025 at 9:00 AM pacific time (16:00 UTC).

For this week I am going to discuss the latest status updates on

Corundum, including progress on the System Verilog rewrite including the
new Taxi transport library, progress on high-resolution time
synchronization, and progress on supporting the Alveo X3/X3522.

[Alex Forencich]

As a reminder, the next Corundum developer meeting will be taking place

on Wednesday March 5, 2025 at 9:00 AM pacific time (16:00 UTC).

For this week I am going to discuss the latest status updates on

Corundum, including progress on the System Verilog rewrite including the

new Taxi transport library, new transceiver wrappers, and progress on
high-resolution time synchronization and consistent transceiver latency.

[Alex Forencich]

As a reminder, the next Corundum developer meeting will be taking place

on Wednesday April 2, 2025 at 9:00 AM pacific time (16:00 UTC).

For this week I am going to discuss the latest status updates on

Corundum, including progress on the System Verilog rewrite including the

new Taxi transport library, new transceiver wrappers, new fractional
MMCM module for DDMTD, updates to the statistics subsystem, potential
new qJTAG interface, and progress on consistent transceiver latency.

[Alex Forencich]

As a reminder, the next Corundum developer meeting will be taking place

on Wednesday May 7, 2025 at 9:00 AM pacific time (16:00 UTC).

For this week I am going to discuss the latest status updates on
Corundum, including progress on the System Verilog rewrite including the
new Taxi transport library, new transceiver wrappers, new fractional
MMCM module for DDMTD, updates to the statistics subsystem, potential
new qJTAG interface, and progress on consistent transceiver latency.

I also have a short presentation on QUASR (the current UCSD project)
that I gave at a QuANET workshop two weeks ago that might be
interesting, as we're using Corundum for that project to interface with
the optics.

[Alex Forencich]

As a reminder, the next Corundum developer meeting will be taking place

on Wednesday June 4, 2025 at 9:00 AM pacific time (16:00 UTC).

For this week I am going to discuss the latest status updates on
Corundum, including progress on the System Verilog rewrite including the

new Taxi transport library, new transceiver wrappers, new synchronous
gearbox support for lower and more consistent transceiver latency,
initial plans for the "next generation" version of Corundum, initial
plans for a potential new (embedded) IP/UDP/RDMA stack, some Versal
architecture notes, and progress on bringing up the VCK5000 board with
Vivado for future Versal device support.

[Alex Forencich]

Thanks to everyone who tuned in this week!

Recording: https://youtu.be/PaffLR9B5so

Slides:
https://docs.google.com/presentation/d/1xPxU2v0gJPrxqsgVcANDOdzGC4RK9Kke/edit?usp=drive_link&ouid=114059947803429764675&rtpof=true&sd=true

Alex Forencich

On 6/3/25 11:18 PM, Alex Forencich wrote:
> As a reminder, the next Corundum developer meeting will be taking

> place on Wednesday June 4, 2025 at 9:00 AM pacific time (16:00 UTC).

>
> For this week I am going to discuss the latest status updates on
> Corundum, including progress on the System Verilog rewrite including
> the new Taxi transport library, new transceiver wrappers, new

> synchronous gearbox support for lower and more consistent transceiver
> latency, initial plans for the "next generation" version of Corundum,
> initial plans for a potential new (embedded) IP/UDP/RDMA stack, some
> Versal architecture notes, and progress on bringing up the VCK5000
> board with Vivado for future Versal device support.
>
>
>
>

[Alex Forencich]

As a reminder, the next Corundum developer meeting will be taking place

on Wednesday July 2, 2025 at 9:00 AM pacific time (16:00 UTC).

For this week I am going to discuss the latest status updates on
Corundum, including progress on the System Verilog rewrite including the

new Taxi transport library, new transceiver wrappers, new synchronous

gearbox support for lower and more consistent transceiver latency, new
32-bit MAC+PCS logic, initial plans for the "next generation" version of
Corundum, and status of the new Zircon IP stack

[Alex Forencich]

As a reminder, the next Corundum developer meeting will be taking place

on Wednesday August 6, 2025 at 9:00 AM pacific time (16:00 UTC).

For this week I am going to discuss the latest status updates on

Corundum, including updates on the Taxi transport library, results from
the latest transceiver time sync tests, initial plans for the "next
generation" version of Corundum, and progress on the new Zircon IP stack

[Alex Forencich]

For the September meeting, I'm going to push it until next week
(Wednesday, September 10th specifically) as I am quite busy with a dry
run at UMD this week.  But, if things go well I should have some
interesting stuff to report next week, and we might be able to do
something live-from-the-lab.  Stay tuned!

--
Alex Forencich