ProSoundWeb Community

Please login or register.

Login with username, password and session length
Advanced search  

Pages: 1 2 [All]   Go Down

Author Topic: Dante and Spanning Tree running on redundant networks  (Read 1076 times)

Lonnie Barrios

  • Newbie
  • *
  • Offline Offline
  • Posts: 1
Dante and Spanning Tree running on redundant networks
« on: May 13, 2021, 04:35:46 pm »

I'm standing up redundant Dante networks for a performing arts center.  The consultant who designed this wants both audio networks (Audio A and Audio B, aka Dante Primary & Secondary) physically configured in a ring topology (switch A to switch B to... ...switch J back to switch A).  Spanning tree isn't going to like this and is going to block an interface somewhere on each ring (network).  If someone connects a device (or maybe a portable cabinet with multiple wireless microphone receivers connected to small switches) to the A network, STP will have to go through the election process for root bridge and determine if there is now a network loop.  I imagine this could take several seconds.  My question is how Dante will behave when it sees that the A network is having an interruption.  Will audio continue to play on the B network?  I'm thinking that if there were only one audio network, audio might be interrupted for sure.  We've never done a full redundant setup like this so I don't know how it will behave.  Perhaps the Dante devices will see an issue with the A network and start playing the stream that they also received on the B network.  If an interruption does happen because of STP when connecting a device to network A, it seems like connecting the switches in the portable to network A and B at the same time would definitely cause an interruption to audio.  Any ideas?
Logged

Russell Ault

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 1322
  • Edmonton, AB
Re: Dante and Spanning Tree running on redundant networks
« Reply #1 on: May 13, 2021, 04:53:24 pm »

{...} The consultant who designed this wants both audio networks (Audio A and Audio B, aka Dante Primary & Secondary) physically configured in a ring topology (switch A to switch B to... ...switch J back to switch A). {...}

What? Why?

{...} Any ideas?

Have you considered suggesting to your employer/client that the design of their PAC's Dante network should be left to someone who knows enough about networking to understand that Ethernet doesn't use a ring topology?

No good will come from turning a star into a ring.

-Russ
« Last Edit: May 13, 2021, 05:27:18 pm by Russell Ault »
Logged

Dave Pluke

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 1304
    • BIGG GRIN Productions
Re: Dante and Spanning Tree running on redundant networks
« Reply #2 on: May 13, 2021, 04:55:14 pm »

What? Why?

No good will come from turning a star into a ring.


Agreed. What century is this, anyway?

Dave
Logged
...an analog man in a digital world [tm]

Flying direct to nearly everywhere out of ATL

Russell Ault

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 1322
  • Edmonton, AB
Re: Dante and Spanning Tree running on redundant networks
« Reply #3 on: May 13, 2021, 05:26:50 pm »

Agreed. What century is this, anyway?

I mean, rings were super rare last century, too. As near as I can tell, the only widely-deployed system to use a physical ring topology is the PSTN's SS7. The only one I've ever seen in action is the (admittedly fairly brilliant) Optocore system that DiGiCo licenses. (Strangely, and despite the name, IBM's Token Ring network protocol doesn't actually use a ring topology at the physical layer.)

{...} Any ideas?

Here's another one that will allow you to get up and running quickly (and kick the innevitable headaches down the road, ideally for somebody else to deal with): casually "forget" to make one of the connections in the loop and then also "forget" to enable Spanning Tree Protocol in the switches. Problem, uh, mitigated!

-Russ

ETA: I just realized something: Welcome to the forums! Thank you for being part of the ~0.00001% of new users who actually followed the directions and used their real name as their Display Name!
« Last Edit: May 13, 2021, 05:29:39 pm by Russell Ault »
Logged

Andrew Broughton

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 1836
    • Check Check One Two
Re: Dante and Spanning Tree running on redundant networks
« Reply #4 on: May 13, 2021, 05:56:32 pm »

I mean, rings were super rare last century, too. As near as I can tell, the only widely-deployed system to use a physical ring topology is the PSTN's SS7. The only one I've ever seen in action is the (admittedly fairly brilliant) Optocore system that DiGiCo licenses.

Yamaha's Rivage TwinLANe and Console control networks are also ring topologies.
Logged
-Andy

"Well, my days of not taking you seriously are certainly coming to a middle..."

http://www.checkcheckonetwo.com
Saving lives through Digital Audio, Programming and Electronics.

Justice C. Bigler

  • SR Forums
  • Hero Member
  • *
  • Offline Offline
  • Posts: 2618
  • back home in Oklahoma
    • My homepage
Re: Dante and Spanning Tree running on redundant networks
« Reply #5 on: May 13, 2021, 06:07:07 pm »

I'm standing up redundant Dante networks for a performing arts center.  The consultant who designed this wants both audio networks (Audio A and Audio B, aka Dante Primary & Secondary) physically configured in a ring topology (switch A to switch B to... ...switch J back to switch A).  Spanning tree isn't going to like this and is going to block an interface somewhere on each ring (network).  If someone connects a device (or maybe a portable cabinet with multiple wireless microphone receivers connected to small switches) to the A network, STP will have to go through the election process for root bridge and determine if there is now a network loop.  I imagine this could take several seconds.  My question is how Dante will behave when it sees that the A network is having an interruption.  Will audio continue to play on the B network?  I'm thinking that if there were only one audio network, audio might be interrupted for sure.  We've never done a full redundant setup like this so I don't know how it will behave.  Perhaps the Dante devices will see an issue with the A network and start playing the stream that they also received on the B network.  If an interruption does happen because of STP when connecting a device to network A, it seems like connecting the switches in the portable to network A and B at the same time would definitely cause an interruption to audio.  Any ideas?
Will you post the name and phone number of this idiot so that I can call him and ridicule him for being a total and complete moron?


That's not how Dante networks work and will produce a totally non functioning system.
Logged
Justice C. Bigler
www.justicebigler.com

Scott Holtzman

  • Hero Member
  • *****
  • Online Online
  • Posts: 6351
  • Ghost AV - Avon Lake, OH
    • Ghost Audio Visual Systems, LLC
Re: Dante and Spanning Tree running on redundant networks
« Reply #6 on: May 13, 2021, 09:37:38 pm »

I mean, rings were super rare last century, too. As near as I can tell, the only widely-deployed system to use a physical ring topology is the PSTN's SS7. The only one I've ever seen in action is the (admittedly fairly brilliant) Optocore system that DiGiCo licenses. (Strangely, and despite the name, IBM's Token Ring network protocol doesn't actually use a ring topology at the physical layer.)


-Russ

ETA: I just realized something: Welcome to the forums! Thank you for being part of the ~0.00001% of new users who actually followed the directions and used their real name as their Display Name!


Russ - SS& in the US is a combination of associated signalling (A-links) between major switches and quasi associated using Signal Transfer Points (STP's) that run F links back to the CO.  Most A-links except in large spit tandem LATA's like NYC have been retired.


In any event it's not a ring network.  It is a hardened point to point, there always is another path (except when there is not and then shit breaks). 


I believe a lot of SS-7 has been layered in to IP now.  Interesting  mobile SMS uses the TCAP (a messenger function) to transport SMS messages and mobility management information (for automatic roaming), it was never designed for that Bell Labs just hacked it into the mobile switches and other vendors adopted then it became a ratified protocol.  GSM on the other hand, that doesn't use SS7 had formal link protocols for mobility management and test messaging.


With all that information you didn't want, I scratched my brain and SONET (a fiber ring topology, part of the national digital hierarchy) is a true ring network with healing.  It's basic speed goes all the way to OC-768 40Gb today.  OC is multiples of T1 1.54 carrier rates.  There isn't much switched circuit telephony anymore.  SONET transport remains ubiquitous at the carrier level for both packet and switched traffic.


Original token ring (before it used twisted pair) was absolutely a ring.  The token was a collision prevention system.  Only the node with token could talk.  MAU's were installed in a ring architecture then nodes (computers, terminals and printers) were hung off the MAU's.  The MAU's were the ring. 







Logged
Scott AKA "Skyking" Holtzman

Ghost Audio Visual Solutions, LLC
Cleveland OH
www.ghostav.rocks

Russell Ault

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 1322
  • Edmonton, AB
Re: Dante and Spanning Tree running on redundant networks
« Reply #7 on: May 13, 2021, 09:46:47 pm »

Russ - SS& in the US is a combination of associated signalling (A-links) between major switches and quasi associated using Signal Transfer Points (STP's) that run F links back to the CO.  Most A-links except in large spit tandem LATA's like NYC have been retired.

In any event it's not a ring network.  It is a hardened point to point, there always is another path (except when there is not and then shit breaks). 

I believe a lot of SS-7 has been layered in to IP now.  Interesting  mobile SMS uses the TCAP (a messenger function) to transport SMS messages and mobility management information (for automatic roaming), it was never designed for that Bell Labs just hacked it into the mobile switches and other vendors adopted then it became a ratified protocol.  GSM on the other hand, that doesn't use SS7 had formal link protocols for mobility management and test messaging.

With all that information you didn't want, I scratched my brain and SONET (a fiber ring topology, part of the national digital hierarchy) is a true ring network with healing.  It's basic speed goes all the way to OC-768 40Gb today.  OC is multiples of T1 1.54 carrier rates.  There isn't much switched circuit telephony anymore.  SONET transport remains ubiquitous at the carrier level for both packet and switched traffic.

Original token ring (before it used twisted pair) was absolutely a ring.  The token was a collision prevention system.  Only the node with token could talk.  MAU's were installed in a ring architecture then nodes (computers, terminals and printers) were hung off the MAU's.  The MAU's were the ring.

Well, I must be particularly tired today (I mean, I know I am, but it's clearly showing). I was doing background research using the Wikipedia article for "ring network" and I missed the fact that it has basically zero citations (and is also apparently rubbish).

We all make mistakes, and this one was mine. Sorry, everyone!

-Russ
Logged

Tim Verhoeven

  • Newbie
  • *
  • Offline Offline
  • Posts: 8
Re: Dante and Spanning Tree running on redundant networks
« Reply #8 on: May 14, 2021, 08:20:30 am »

I'm standing up redundant Dante networks for a performing arts center.  The consultant who designed this wants both audio networks (Audio A and Audio B, aka Dante Primary & Secondary) physically configured in a ring topology (switch A to switch B to... ...switch J back to switch A).  Spanning tree isn't going to like this and is going to block an interface somewhere on each ring (network).  If someone connects a device (or maybe a portable cabinet with multiple wireless microphone receivers connected to small switches) to the A network, STP will have to go through the election process for root bridge and determine if there is now a network loop.  I imagine this could take several seconds.  My question is how Dante will behave when it sees that the A network is having an interruption.  Will audio continue to play on the B network?  I'm thinking that if there were only one audio network, audio might be interrupted for sure.  We've never done a full redundant setup like this so I don't know how it will behave.  Perhaps the Dante devices will see an issue with the A network and start playing the stream that they also received on the B network.  If an interruption does happen because of STP when connecting a device to network A, it seems like connecting the switches in the portable to network A and B at the same time would definitely cause an interruption to audio.  Any ideas?

So yes, STP and Dante don't work together. Building rings also not the way to build a redundant Dante network. As other here has said, you build a star topology and making it redundant is done by doubling everything.
So you are actually creating 2 star topologies that are never interconnected, so no links between any of the network devices making up both topologies. Only the Dante end-clients (mixers, stageboxes, wireless microphones, laptops, ...) are connected to both networks. Of which one is designated the primary and the other the secondary. Make sure that all clients use the same network as the primary.

There will be not STP needed here, so no ports shutdown, no delays in connecting, etc. etc. Which is what Dante needs to work without interruptions.

The biggest downside of such a setup is of course cost, since everything needs to be doubled up. But it is the only way to assure a fully redundant (Dante) network. If you want to save money you can give up some of the redundancy by not doubling up the actual network devices (switches), but divide them virtually (using VLANs and other methods). But keep the redundant network connections. The idea here is that a switch on itself is less likely to fail then the cabling and connectors.

Another tip when building out a star Dante network is to try to keep the number of layers as limited as possible. Ideally you would only have 1 set of core switches forming the top layer and just a second layer of switches used to connect the clients (e.g. http://units.folder101.com/cisco/sem1/Notes/ch2-topologies/extendedstar.gif). But if you can get away without needing the second layer, even better!
Logged

Aram Piligian

  • Jr. Member
  • **
  • Offline Offline
  • Posts: 88
Re: Dante and Spanning Tree running on redundant networks
« Reply #9 on: May 14, 2021, 05:07:27 pm »

I will say that I've carefully utilized rings with RSTP enabled for a couple of specific purposes.  We have a lot of the Martin Audio CDD-Live series speakers, which all utilize the Primary-only Ultimo chips.  For flown or ground stacked rigs that utilize a lot of daisy-chaining of the Primary, it can be handy to close the ring and have a slight bit of redundancy in the event that one of those links along the way gets broken.  Sometimes, it just makes more sense from a physical cabling perspective. 

Of course, this all requires having confidence in the switch config and testing to make sure it works the way it should, but I'll take the 2-6 second drop as RSTP reconverges if it keeps the rig running overall.

Getting back to OP's line of questioning--the Dante chips would recognize both Primary and Secondary ports on the same network and be Very Not Happy. Others have outlined the correct procedure.
« Last Edit: May 14, 2021, 05:34:33 pm by Aram Piligian »
Logged

Riley Casey

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 1660
Re: Dante and Spanning Tree running on redundant networks
« Reply #10 on: May 15, 2021, 12:04:46 pm »

I'm not qualified to address the hardcore nuts and bolts of networking but it never hurts to remember that the initial design criteria for Dante redundancy was addressing the portable /  tour market where snakes routinely unspooled daily and essentially ran thru open top conduit or sometimes literally under the feet of thousands of audience members. A Dante network installed in a venue with cables permanently installed in EMT and with connectors that might be re-mated once a year is pretty robust to start with. Its certainly is a good idea to have back up everything and redundancy is only a bad thing when it leads to misconfiguration but it doesn't need to be a religion.


I'm standing up redundant Dante networks for a performing arts center.  ...

Helge A Bentsen

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 1220
  • Oslo, Norway.
Re: Dante and Spanning Tree running on redundant networks
« Reply #11 on: May 15, 2021, 01:58:30 pm »

I'm not qualified to address the hardcore nuts and bolts of networking but it never hurts to remember that the initial design criteria for Dante redundancy was addressing the portable /  tour market where snakes routinely unspooled daily and essentially ran thru open top conduit or sometimes literally under the feet of thousands of audience members. A Dante network installed in a venue with cables permanently installed in EMT and with connectors that might be re-mated once a year is pretty robust to start with. Its certainly is a good idea to have back up everything and redundancy is only a bad thing when it leads to misconfiguration but it doesn't need to be a religion.

Agreed.
I frequently run  Dante primary only, no issues.
But, if my cables are exposed to traffic or other mechanical issues, I run secondary as well.
Logged

brian maddox

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 2656
  • HeyYahWon! ttsss! ttsss!
Re: Dante and Spanning Tree running on redundant networks
« Reply #12 on: May 15, 2021, 04:34:17 pm »

Agreed.
I frequently run  Dante primary only, no issues.
But, if my cables are exposed to traffic or other mechanical issues, I run secondary as well.

At the risk of swerving the thread...

I've come to pick and choose very carefully where I run a secondary network for Dante and where I don't. The reason for this is that running a secondary network into any device enables cabling redundancy BUT also introduces a potential catastrophic failure point as well because as soon as they are in the same physical box, the potential for them to be "merged" into a single network and crash your entire Dante system becomes at least a possibility. A number of Dante devices, when initialized, revert to "daisy chain" mode. This is catastrophic and will crash your entire Dante network, often requiring a complete reboot of EVERY device on the network including ALL of your switches. I've had this happen with a Shure wireless unit [completely on it's own mind you]. I've had this happen with a Yamaha RIO [during troubleshooting another issue, but still a major PITA]. This is not a failure you want to have happen during a show.

This is also why I cringe when single switches are configured for separate VLans and the Primary and Secondary are run on the same switch. Yes, it should work fine. But if it should STOP "working fine" [due to someone fat fingering a switch configuration change or any number of other potential oopses] it's gonna be a Pretty Big Deal.

I heard recently that even NASA has admitted that overly obsessing over too many Fail Over redundant systems can, in and of itself, lead to failures. Just something to keep in mind.
Logged
"It feels wrong to be in the audience.  And it's too peopley!" - Steve Smith

brian maddox
bdmaudio@gmail.com

'...do not trifle with the affairs of dragons...

       ....for you are crunchy, and taste good with ketchup...'

Phillip Ivan Pietruschka

  • Jr. Member
  • **
  • Offline Offline
  • Posts: 56
Re: Dante and Spanning Tree running on redundant networks
« Reply #13 on: May 15, 2021, 08:16:53 pm »

Brian,

The Dante engineer I did L3 training with a few years ago was very strongly of the opinion that redundancy was principally of value for live production work where there was an appreciable risk of cables being damaged or the like. For installs where this isnít a plausible risk, and higher end switches are used he argued there was basically no point, or the cost & risk V benefit was marginal. It does make sense in broadcast and other facilities which require 24Hr uptime; and without that redundancy maintenance becomes fraught. Iíve heard similar sentiments discussed by broadcast engineers talking about 2110 installations.
« Last Edit: May 15, 2021, 10:39:24 pm by Phillip Ivan Pietruschka »
Logged

Erik Jerde

  • Classic LAB
  • Hero Member
  • *
  • Offline Offline
  • Posts: 1112
Re: Dante and Spanning Tree running on redundant networks
« Reply #14 on: May 16, 2021, 10:31:14 am »

So reading through this I think there may be a couple of mis-understandings from the OPs original.  It reads to me like some people think the IT recommendation is to have the primary and secondary networks as part of the same ring.  I believe that is incorrect and that the recommendation is still two separate primary and secondary networks with the switches daisy chained and the last switch in the chain linked back to the first.  STP would then kill one of the links and you'd have your tree.

Regarding the STP election, unless STP is completely eliminated there will always be an election when another switch is plugged in, ring or not.  So when that portable rack is connected and starts sending out it's BPDUs that will trigger an election ring or no.

I personally am loathe to disable STP because I don't trust people to plug stuff in correctly.  I've had to deal with production networks getting interconnected to office networks resulting in dueling DHCP servers and the resulting chaos because a construction worker thought he was helping.  That one was hard to track down because it was stuffed up in a ceiling.  I've never had STP problems because I always leave it enabled.

I would expect that a STP election wouldn't interrupt dante where there is a primary/secondary system setup.  Of course there's lots of primary only devices that could be taken out.  It should be easy to test what actually happens.

You could leave STP enabled on the main switches but disable it on all ports that are permanently plugged in to wall plates.  I'd still be cautious when there's multiple ports present right next to each other.  End users have a way of screwing things up and I could still see people plugging one switch into two ports.  If you had STP off on those ports then your network would get hammered worse than a STP election would do.

As to the ring design vis-a-vis dante, I'd skip that regardless of STP because it makes the network too deep.  IT needs to be instructed that the network should remain as flat as possible.  If redundancy is desired on the switch interconnects look into LAG groups.  I'm not aware of them being contra indicated for Dante though it's entirely possible I've missed that.
Logged

Russell Ault

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 1322
  • Edmonton, AB
Re: Dante and Spanning Tree running on redundant networks
« Reply #15 on: May 16, 2021, 04:05:24 pm »

{...} I personally am loathe to disable STP because I don't trust people to plug stuff in correctly.  I've had to deal with production networks getting interconnected to office networks resulting in dueling DHCP servers and the resulting chaos because a construction worker thought he was helping.  That one was hard to track down because it was stuffed up in a ceiling.  I've never had STP problems because I always leave it enabled. {...}

Huh, this is exactly the opposite of common practice here. All of the PACs I work at (several of which are even municipally-owned) have production networks that are so air-gapped from the rest of the building that the in-house IT department doesn't even know they exist. Heck, in at least one venue, the production networks are so air-gapped that other departments don't even know they exist (jokes about the nature of the stage carpenters' network are left as an exercise for the reader). Production network ports are only provided in locations where anyone who has access to them can be subjected to termination (or, at a minimum, buying everyone a round) should a show-stopping mistake occur, at which point I'd personally rather troubleshoot a broadcast storm (i.e. a persistent failure) than an STP election interruption (i.e. an intermittent failure).

For the OP, the other thing to remember with Dante and adding switches is that so many (almost all?) two-port Dante devices can also act as switches in some configurations, so it's possible to plug in something as simple as a single microphone RX and trigger an election.

-Russ
Logged

brian maddox

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 2656
  • HeyYahWon! ttsss! ttsss!
Re: Dante and Spanning Tree running on redundant networks
« Reply #16 on: May 16, 2021, 07:42:22 pm »

....
For the OP, the other thing to remember with Dante and adding switches is that so many (almost all?) two-port Dante devices can also act as switches in some configurations, so it's possible to plug in something as simple as a single microphone RX and trigger an election.

-Russ

You said what I said, but much more betterer. :)
Logged
"It feels wrong to be in the audience.  And it's too peopley!" - Steve Smith

brian maddox
bdmaudio@gmail.com

'...do not trifle with the affairs of dragons...

       ....for you are crunchy, and taste good with ketchup...'

ProSoundWeb Community

Re: Dante and Spanning Tree running on redundant networks
¬ę Reply #16 on: May 16, 2021, 07:42:22 pm ¬Ľ


Pages: 1 2 [All]   Go Up
 



Page created in 0.03 seconds with 18 queries.