p4-apps@lists.p4.org

P4 Applications Working Group

View all threads

Meeting Notes from 8/20/2020

LJ
Lee, Jeongkeun
Wed, Aug 26, 2020 4:57 AM

Attendees:
Andy Spooner, Chris Sommers (Keysight), Ramesh Sivakolundu (Cisco), Randy Levensalor (CableLabs), Mickey Spiegel, JK Lee (Intel)

<> P4 Apps library
TNA has been opened, the need to create a library or "app store" for P4 code is higher. Apps WG has a repository, time to think about logistics.

Artifacts for each app: runnable P4 + control plane code, test script, document.
Document to specify Barefoot SDE version (in cast of TNA), an inventory of artifacts, and sufficient description to enable the artifacts to be exercised.
Vetting process: manual review at the beginning. Automatic integration with SDE and testing will be needed.

ONF staff can help with actual hosting, maintenance.

Questions:

  • what if the app requires a modification of Barefoot switch.p4, which is not open? A patch to switch.p4 can be shared?
  • What about non-p4 apps, such as INT implementation in DPDK or eBPF under a license different from Apache2 that P4.org uses?

Action items: JK will follow up on the questions and share.

<> Broadcast/Unknown Unicast/Multicast traffic handling in INT
Ramesh: There has been not much discussion around BUM traffic. IOAM had a presentation document for capturing multicast branch ID as part of hop metadata.
Mickey: for BCAST, MCAST, how about ‘force’ bit to report the entire tree and branches?

AI: Ramesh will make a proposal to discuss in the next meeting.

<> INT-MX, source-only feedback
Mickey created a PR that clarifies the semantics of Domain-specific source-only bits. https://github.com/p4lang/p4-applications/pull/87

We walked through a few known use cases of source-only data and found the proposed categorization help clarifying their semantics.

We also briefly talked about UDP port number requirements for INT and Telemetry Report.

In the next meeting,

  • we will continue the BUM discussion,
  • follow up on PRs from Mickey
  • discuss INT with IPSEC

Thanks,
JK

Attendees: Andy Spooner, Chris Sommers (Keysight), Ramesh Sivakolundu (Cisco), Randy Levensalor (CableLabs), Mickey Spiegel, JK Lee (Intel) <> P4 Apps library TNA has been opened, the need to create a library or "app store" for P4 code is higher. Apps WG has a repository, time to think about logistics. Artifacts for each app: runnable P4 + control plane code, test script, document. Document to specify Barefoot SDE version (in cast of TNA), an inventory of artifacts, and sufficient description to enable the artifacts to be exercised. Vetting process: manual review at the beginning. Automatic integration with SDE and testing will be needed. ONF staff can help with actual hosting, maintenance. Questions: * what if the app requires a modification of Barefoot switch.p4, which is not open? A patch to switch.p4 can be shared? * What about non-p4 apps, such as INT implementation in DPDK or eBPF under a license different from Apache2 that P4.org uses? Action items: JK will follow up on the questions and share. <> Broadcast/Unknown Unicast/Multicast traffic handling in INT Ramesh: There has been not much discussion around BUM traffic. IOAM had a presentation document for capturing multicast branch ID as part of hop metadata. Mickey: for BCAST, MCAST, how about ‘force’ bit to report the entire tree and branches? AI: Ramesh will make a proposal to discuss in the next meeting. <> INT-MX, source-only feedback Mickey created a PR that clarifies the semantics of Domain-specific source-only bits. https://github.com/p4lang/p4-applications/pull/87 We walked through a few known use cases of source-only data and found the proposed categorization help clarifying their semantics. We also briefly talked about UDP port number requirements for INT and Telemetry Report. In the next meeting, * we will continue the BUM discussion, * follow up on PRs from Mickey * discuss INT with IPSEC Thanks, JK
JR
Jennifer Rexford
Wed, Aug 26, 2020 11:55 AM

<> P4 Apps library
TNA has been opened, the need to create a library or "app store" for P4 code is higher. Apps WG has a repository, time to think about logistics.

Artifacts for each app: runnable P4 + control plane code, test script, document.
Document to specify Barefoot SDE version (in cast of TNA), an inventory of artifacts, and sufficient description to enable the artifacts to be exercised.
Vetting process: manual review at the beginning. Automatic integration with SDE and testing will be needed.

ONF staff can help with actual hosting, maintenance.

In case it’s helpful, at Princeton we have a repository of P4 apps (including Tofino implementations) at https://github.com/Princeton-Cabernet/ https://github.com/Princeton-Cabernet/ .  We also have pointers to the other open-source P4 apps we know of, https://github.com/Princeton-Cabernet/p4-projects/blob/master/README.md https://github.com/Princeton-Cabernet/p4-projects/blob/master/README.md. I think it’d be great to have a P4 apps library.  We’re happy to help.

— Jen

Questions:
what if the app requires a modification of Barefoot switch.p4, which is not open? A patch to switch.p4 can be shared?
What about non-p4 apps, such as INT implementation in DPDK or eBPF under a license different from Apache2 that P4.org http://p4.org/ uses?
Action items: JK will follow up on the questions and share.

<> Broadcast/Unknown Unicast/Multicast traffic handling in INT
Ramesh: There has been not much discussion around BUM traffic. IOAM had a presentation document for capturing multicast branch ID as part of hop metadata.
Mickey: for BCAST, MCAST, how about ‘force’ bit to report the entire tree and branches?

AI: Ramesh will make a proposal to discuss in the next meeting.

<> INT-MX, source-only feedback
Mickey created a PR that clarifies the semantics of Domain-specific source-only bits. https://github.com/p4lang/p4-applications/pull/87 https://github.com/p4lang/p4-applications/pull/87

We walked through a few known use cases of source-only data and found the proposed categorization help clarifying their semantics.

We also briefly talked about UDP port number requirements for INT and Telemetry Report.

In the next meeting,
we will continue the BUM discussion,
follow up on PRs from Mickey
discuss INT with IPSEC
Thanks,
JK


P4-apps mailing list
P4-apps@lists.p4.org mailto:P4-apps@lists.p4.org
http://lists.p4.org/mailman/listinfo/p4-apps_lists.p4.org http://lists.p4.org/mailman/listinfo/p4-apps_lists.p4.org

> <> P4 Apps library > TNA has been opened, the need to create a library or "app store" for P4 code is higher. Apps WG has a repository, time to think about logistics. > > Artifacts for each app: runnable P4 + control plane code, test script, document. > Document to specify Barefoot SDE version (in cast of TNA), an inventory of artifacts, and sufficient description to enable the artifacts to be exercised. > Vetting process: manual review at the beginning. Automatic integration with SDE and testing will be needed. > > ONF staff can help with actual hosting, maintenance. In case it’s helpful, at Princeton we have a repository of P4 apps (including Tofino implementations) at https://github.com/Princeton-Cabernet/ <https://github.com/Princeton-Cabernet/> . We also have pointers to the other open-source P4 apps we know of, https://github.com/Princeton-Cabernet/p4-projects/blob/master/README.md <https://github.com/Princeton-Cabernet/p4-projects/blob/master/README.md>. I think it’d be great to have a P4 apps library. We’re happy to help. — Jen > > Questions: > what if the app requires a modification of Barefoot switch.p4, which is not open? A patch to switch.p4 can be shared? > What about non-p4 apps, such as INT implementation in DPDK or eBPF under a license different from Apache2 that P4.org <http://p4.org/> uses? > Action items: JK will follow up on the questions and share. > > <> Broadcast/Unknown Unicast/Multicast traffic handling in INT > Ramesh: There has been not much discussion around BUM traffic. IOAM had a presentation document for capturing multicast branch ID as part of hop metadata. > Mickey: for BCAST, MCAST, how about ‘force’ bit to report the entire tree and branches? > > AI: Ramesh will make a proposal to discuss in the next meeting. > > <> INT-MX, source-only feedback > Mickey created a PR that clarifies the semantics of Domain-specific source-only bits. https://github.com/p4lang/p4-applications/pull/87 <https://github.com/p4lang/p4-applications/pull/87> > > We walked through a few known use cases of source-only data and found the proposed categorization help clarifying their semantics. > > We also briefly talked about UDP port number requirements for INT and Telemetry Report. > > In the next meeting, > we will continue the BUM discussion, > follow up on PRs from Mickey > discuss INT with IPSEC > Thanks, > JK > > > > > _______________________________________________ > P4-apps mailing list > P4-apps@lists.p4.org <mailto:P4-apps@lists.p4.org> > http://lists.p4.org/mailman/listinfo/p4-apps_lists.p4.org <http://lists.p4.org/mailman/listinfo/p4-apps_lists.p4.org>