Understanding Media Flows in Microsoft Teams – My Notes

Understanding Media Flows in Microsoft Teams and Skype for Business

Awesome Session by Tim … I was more interested in the difference between Teams Media Handling vis-a-vis Skype4B hence not noting down the basics…. I would strongly recomend watching this video but if you are short of time these notes may give you a quick roundup….

• Teams uses REST APIs for signaling while S4B uses SIP
• The REST APIs use HTTPS  WebSockets!… (I’m wondering …how would REST work over WebSocket!)
• Introducing Two ‘Relay’ concepts
  • Media Relay
    • Born On Premise and Not scaleable for cloud
    • Designed for Single Location
    • Used by S4B till date
  • Transport Relay
    • Born in The Cloud!
    • Designed to be Multi-Location using anycast IP (https://en.wikipedia.org/wiki/Anycast)! – Offices will need Local Breakouts to make use of this 😦
    • S4BO moving to this model
    • Teams uses right from start
• Anycast IP
  • Same IP Used by Global DCs
  • Requests land on the DC Closest to the User
  • Regional Government constraints respected-> eg EU Govts users will land on EU DCs even if users are in US! But same not applicable for US Govt Users…
  • Generic EU Users will be forced to EU and US Relays only (Don’t Trust the ROW I guess)
• Client Authentication
  • No more MRAS … Only TRAP -> Transport Relay Auth Provider
• For S4B users who moved to transport relay , The MRAS media usr/psw/server address retreival process is same as old s4b but the server url will always start with TREdge.<online/us/eu…>.lync.com…  Also the list will only have one server name while media relay based s4bo users will have multiple server names ( 1 for internal and 1 for external)…
• In teams there is no internal user hence only 1 server address
• Teams does not generate customer facing logs and hence cant use the same methods to debug
• We’ll need to use a trusted MITM Proxy like charles to capture traffic and debug protocols
• In charles search for mdn_trap
• All details are in json 🙂             [note: I typed this too soon and realised later that its not so always 😦 ]
• Server address is an Anycast IP address and not FQDN
• During actual Media session creation the Anycast IP will assign an alternate IP which is used by the client for relay
• Teams will connect on 443/tcp and 3478udp … and will retain connection on port 3478/udp (slide shows 4478 but Tim apologises it as a typo in audio)
• In locations where S4B has been working the 3479-3481 may be closed and hence relay will fall back to 443/tcp…. Sadly with TCP calls will work but with relatively bad user xperience…SO if Teams is performing badly in ur office , u know why…
• + In Teams all P2P traffic is defaults to udp only…. If udp is blocked then Relay will be used to mediate over tcp
• Aaaah! SDP Still used in Team!… Still stuck in ancient unlabeled Messaging age 😦
• No more 50K wala giant hole in the firewall funda …Media is always in 3479-3481 (3479-audio, 3480-video, 3481-desktop sharing/VBSS -> video based Screen Sharing) [Added on 22Aug2018] Please note that this is optional if all users are on teams + S4B online. If you are on hybrid with users spread across On-Premise S4B and Teams then this 50000-59999 TCP/UDP hole is still needed. Please check the latest documentation here at https://docs.microsoft.com…
• Call quality Dashboard used by Teams also (hurrah!)
• Connectivity Testing Tool coming soon…