DASH Workgroup Community Update 5/28/2025

[Kristina Moore]

Hello DASH Community –thank you for your time last Wednesday. 

In our call we discussed the recent repository Merge issue(s) that are hopefully resolved today.  We had a separate Behavioral Model meeting last week to review Issues/PRs to ensure we are up to date, with AIs sent separately.  If you would like to receive those notes, please ‘r’ to me (kri...@microsoft.com). 

We had a larger discussion around discovery of the DST MAC address for our VM destination, and a question around ACL ratios for PL traffic.  We’ve requested help from the Azure SDN team to obtain clarity.  High Availability work is also underway with 3 PRs in progress. 

XFlow Research had some higher priority work this week, and will attempt to look at the contribution to counters for the dash-sonic-hld to differentiate DASH counters (DPU, ENI, Global for example) vs. Switch counters over the next couple of weeks.  @Prabhat Aravind suggested following existing PRs in the SONiC Utilities repo that added show commands as a reference for implementing DASH show commands. 

And lastly, with respect to asking for help with optimization of the ASIC DB to reduce memory footprint for DPUs; there has been previous optimization work done for App DB that could be leveraged if anyone would like to help out.  The scope of work is more than optimization of memory; it may be removal of DASH objects entirely from the ASIC_DB.  We do not have a prior PR to reference for this, and more discussion is likely needed here. 

Follow-up tasks:

• Azure VM Experiment: Use an Azure VM and test the ARP table entries to verify if all services' IPs are populated. (Mircea)
• SDN Team Inquiry: Connect with the SDN team to clarify the handling of ACLs for private link scenarios. (Complete)
• ASIC DB Optimization: Ping Prabhat to check for prior PRs related to ASIC DB optimization. (Complete)
• Follow-up on ACL Ratio: Follow up on the ACL ratio for private link scenarios to determine the realistic distribution. (Complete)
          

In Summary (full list below), since the last Community call we have:

3 PRs Completed (-3)

8 in To Do (- 1)

9 in Draft (+1)

26 in Progress (+/-0)

5 Awaiting Review (+2)

 

Just a reminder that we would encourage/invite Community members to present to the Community (test runs or progress, new scenarios, etc…), just ‘r’ to let me know, or generate a PR in the repo.

The DASH YouTube channel link is here to subscribe / access WG content (and click the bell to receive notifications). 

Thank you for your time/contributions – see you on 6/4/2025

 

Meeting Title:  SONiC-DASH-Workgroup Community Meeting #148

Attendees (16):

DASH Group to join: https://groups.google.com/g/sonic-dash

DASH-Test-Workgroup Group to join: https://groups.google.com/g/sonic-dash-test-workgroup

Linux Foundation list: https://lists.sonicfoundation.dev/g/SONiC-Dash

 

Abdul Rouff - NVidia	Kanza Latif - XFlowResearch	Philipp Keydel - Synogate	Satya Valli Rama - AMD
Bud Grise - XSightLabs	Kristina Moore - MSFT	Prabhat Aravind - MSFT	Veerappan, Senthilnathan - AMD
Don Ewald	Marian Pritsak - NVidia	Ramesh Raghupathy - Cisco	Wenchung Wang - Cisco
Guy Duryee - Cisco	murali Venkateshaiah - Cisco	Rimsha Syed - XFlowResearch	Yakiv Huryk - NVidia

 

 TL:DR DASH Community Notes 😊  

• Repository Merge Issues: discussed the recent issues with merging into the SONiC DASH and normal SONiC repositories, which have blocked PRs for the past week and a half.
• VM Destination MAC Address: Keysight raised a question about how the VM or on-premise computer knows the DST MAC address, and Riff explained that Azure uses a driver to rewrite the MAC address to a fake one when traffic is sent out from the VM.
• Question on MAC Address: Mircea asked how the VM or on-premise computer knows the DST MAC address, explaining that typically a Linux PC looks at the neighbor or ARP table for IP-MAC pairs.
• Riff's Explanation: Riff explained that in Azure, a driver rewrites the MAC address to a fake one when traffic is sent out from the VM. This process involves the kernel initiating an ARP, and the driver faking a response to populate the table.
• Gateway and Routing: Riff clarified that there is no actual gateway service doing the rewrite; instead, it is done in the kernel driver. He also mentioned that on-prem traffic is routed differently when it comes into Azure.
• Further Clarification: Mircea sought further clarification on whether the VM sees all services' IPs as being in the same domain or if there is a gateway. Riff responded that he could not provide a definitive answer without node access.
• Testing Azure VM: Mircea suggested renting an Azure VM to test how the ARP table is populated with IP-MAC pairs, and Riff agreed that this should work.
• Agreement and Follow-Up: Riff agreed that this approach should work, and Kristina offered to run this by the SDN team for further validation and support.
• HA PRs
• PR62 swbus-edge support of reconnecting to swbusd: Jing will be reviewing the this HA item
• PR679 Channel Tunnel Key: Marian is working to add a channel tunnel key to HA (Add dp_channel_tunnel_key attribute to HA set).
• Initializing H Manager D and DPU Actor Implementation, both being worked on by Cisco and Fred respectively.
• UDP Port Forwarding: Marian sought clarification on forwarding packets with different DST ports, and Riff explained that as long as the port is different, the forwarding will work fine.
• Clarification Request: Marian asked Riff about forwarding packets with different DST ports, seeking to understand the implications for UDP layer forwarding.
• Forwarding Mechanism: the forwarding mechanism relies on the DST port being set correctly, and internally, different protocols like VxLAN or NVGRE can be used without issues.
• ASIC DB Optimizations: mentioned that Prince Sunny from the SONiC Team suggested working on ASIC DB optimizations, similar to the work done in the app DB.  Prior PRs done for APP_DB involved using protobuf instead of strings to store key value pairs, but the scope of work for ASIC_DB seems to be more in the sense that we looking to optimize more than just memory; more discussions needed.
• Private Link Traffic:
• Question on ACL Ratio: Mircea inquired about the ratio of inbound and outbound ACLs for private link traffic, noting that most traffic is outbound.  On prem and VM scenarios will affect the ACL ratio
• Graceful Shutdown Design: the HLD for DP Graceful Shutdown was presented in the Linux Foundation call last Thursday, and Oleksandr confirmed that there were minor comments but overall agreement on the design.

 

Sticky for Links/Reference:
DASH Groups to join to receive Invites, Meeting Notes, and Comms	DASH: https://groups.google.com/g/sonic-dash     DASH-Test-Workgroup Group: https://groups.google.com/g/sonic-dash-test-workgroup   Linux Foundation list: https://lists.sonicfoundation.dev/g/SONiC-Dash	If anyone knows other people who would like info re: our community, please have them joins these groups for receive Comms, etc…
Links to Recording	Teams/Sharepoint: ​Recap: SONiC-DASH Workgroup Community Meeting Wednesday, May 28  DASH Community YouTube: https://youtu.be/kZBfHXiCC7Q  HA moved to SmartSwitch LF group on Thursdays YouTube Behavioral Model: 5/29/2025 Notes sent separately	5/28/2025 DASH Community Call; please request access via the link if you are not able to view/listen
Azure DASH GitHub Repo:	https://github.com/sonic-net/DASH
Test/Docs folder:	https://github.com/sonic-net/DASH/blob/main/test/docs/dash-test-workflow-saithrift.md	Ideal test workflow is here, converted to .md
SAI Thrift	SAI Thrift PR	Client server needed for testing
P4	https://opennetworking.org/p4/ and https://p4.org/working-groups/	Open source, domain-specific programming language for network devices, specifying packet processing for data plane devices (switches, routers, NICs, filters, etc.)
PINS	https://opennetworking.org/pins/
PNA consortium spec	https://p4.org/p4-spec/docs/PNA-v0.5.0.html	An architecture describing the structure and common capabilities of network interface controller (NIC) devices which process packets transiting one or more interfaces and a host system. Describes the structure and capabilities of the pipeline, and a user program, which specifies the functionality of the programmable blocks within that pipeline. For more information, see the P4 Language Consortium specifications
IPDK	Infrastructure Programmer Development Kit (ipdk.io) and https://github.com/ipdk-io/ipdk-io.github.io	IPDK is an open source, vendor agnostic framework of drivers and APIs for infrastructure offload and management which runs on a CPU, IPU, DPU or switch. IPDK runs in Linux and uses a set of well-established tools such as DPDK and P4 to enable network virtualization.
bmv2	https://github.com/p4lang/behavioral-model	The second version of the reference P4 software switch, nicknamed bmv2 (for behavioral model version 2). The software switch is written in C++11. It takes as input a JSON file generated from your P4 program by a P4 compiler and interprets it to implement the packet-processing behavior specified by that P4 program
DPDK	https://www.dpdk.org/	DPDK is the Data Plane Development Kit which consists of libraries to accelerate packet processing workloads running on a wide variety of CPU architectures.
Linux Foundation SmartSwitch	https://lists.sonicfoundation.dev/g/sonic-smartswitch/calendar

  

Thank you again for your participation…

Kristina Moore MBA, M.S., CISSP - Azure Core Principal PM / DASH & SmartSwitch
Office: 425-722-7720     Mobile: 425-876-2040     Email: kri...@microsoft.com
DASH Group to join: https://groups.google.com/g/sonic-dash    
DASH-Test-Workgroup Group to join: https://groups.google.com/g/sonic-dash-test-workgroup
Linux Foundation:  https://lists.sonicfoundation.dev/g/SONiC-Dash