Networking/MonthlyCalls/2014-06-09

Minutes of OCP networking project monthly call

June 9, 10-11 AM US/Pacific, 1-2 PM Eastern

Attendees

• Omar Baldonado (Facebook) - scribe for meeting
• Carlos Cardenas (UTSA)
• Marc Randolph (Dell)
• Lionel Pelamourgues (OptumSoft)
• Samantha Jian-Pielak (Canonical)
• Mike Andrewartha (Microsoft)
• Bob Thurston (Fidelity Investments)
• Mark Plante (Fidelity)
• Jason Forrester (Apple)
• Adam Casella (Apple)
• Glenn Sullivan(Apple)
• Rob Sherwood (Big Switch Networks)
• Hideaki nagaya
• Curt Brune (Cumulus Networks)
• mike andrewartha
• nitin jain
• samantha jian-pielak
• oded raz
• satomi
• Albert Chen
• Jason Forrester
• Janardhanan
• Prem (Sylvanus) Jonnalagadda (Broadcom)
• + another 10 more who were just phone numbers in Fuze!

Agenda

Review of Accton specifications

Walkthrough of Accton open rack spec (see wiki - v0.03)

• sample units available in another 2-3 weeks - sample units to:
  • Bob Thurston (Fidelity)
  • Thao Nguyen (Facebook)
  • Carlos Cardenas (UTSA)
• made it high in the back to stiffen it up, but then do we lose the ability to hot-swap power supplies?
• Bob Thurston: supportability of 12V power supply - if we all push together, then it would be good!
• Carlos: is that a standard OCP connector? difficult to do hot swappability of power supply based on where bus-bar is in the back? will that be an issue?
  • Based on the OCP Open Rack, without the ORSA‐1RU, it is nearly impossible to hot swap power supplies
• Gerard (Facebook): are you going to share schematic/2d files. Jeff - yes :)

Walkthrough of Accton Leaf switch spec (see wiki v0.06)

• changes in new spec v6 - modular CPU subsystem + freescale CPU module in addition to x86
• Peter: which freescale? P2041 - believe that is the quad-core (http://media.freescale.com/phoenix.zhtml?c=196520&p=irol-newsArticle&ID=1561029)
• Gerard: can we expose the SATA port exposed? Peter ... probably an msata drive near the CPU module
• Gerard: need to have connector info for the power supply? in spec already...
• Gerard: how much DRAM and flash can be added? max is 32GB (based on availability)
• Omar: schematics and 2d files gerbers - intent is to share all those too!
• Omar: permissive or restrictive license? They are leaning toward the restrictive "GPL-like" one
• Omar: availability? shipping to beta customers/partners now - end of july for certification
• Gerard: availability of driver/firmware to get the box up? intend to submit all open source software, including BIOS, grub/ONIE/ONL changes, ... also for freescale
• Peter: Linux distro testing - have tested with Ubuntu and centos on x86 already
• Gerard: can we see the suggested power module for this switch, especially related to 12V power so we can give feedback on it...
• Carlos: Is there a test plan for the switch itself? I'm thinking if other manufacturers take the specs and build their own switch based on spec. How can we ensure they built the switch according to spec?
• Peter: should we split the testing based on the control modules vs. the networking module itself?
• Albert Chen: target BOM cost? depends on purchasing power/negotiations with vendor
• Robert Chu (from Chat): As to the FPGA, how much will we get from Accton? I also mean the CPLD.

Discuss testing and certification process, specifically related to ONIE but even more generally

• Carlos: new ONIE test doc rev coming later this week (covers manual testing)
• Scope of testing: Install a NOS, uninstall a NOS, provide a rescue mode, ONIE update
• Jana (Dell) - what is the equivalent in the server world? They been using the checkbox from canonical. Typically this is done with BIOS/IPMI dependencies or on-board serial
• encourage commits as folks review
• Curt: it would be great (Omar: strongly encouraged?) to contribute back any ONIE-related code changes. Certainly, the image has to be freely available (Carlos), but the source would be great.
• Rob (Big Switch): pretty good luck with working HW vendors/component manufacturers to ask them to open source stuff, let's just try asking

Discussion of Next meetings

• Next F2F: Possibly July 21 at UNH - Bob Thurston is driving this, still need to confirm
• Next Incubation committee meeting July 3rd, to be confirmed (really before US holiday?)
• Next monthly project call, July 14 (2nd Monday of month, same dialin, same URL, same etherpad)

Action items from meeting

• Jeff Catlin: Can we reduce some of the height in the back to support swappability but still maintain rigidity?
• Jeff Catlin: Look into sharing the testing plans, starting at low-level
• Omar to confirm future meetings