High Network Broadcast Traffic Rate w/ C3500

[Michael HeathParticipant]

Hi all –

Our C3500 console is originating around 22 UDP broadcast packets per second. Looking at the DLive for IT Managers document, the only thing I can see that should be causing any notable broadcast traffic is AHNet discovery, which should be 1 message per second.

I’d like to avoid clamping down on broadcast traffic too drastically, since the AHNet discovery offers some nice functionality. But I’d like to get to the bottom of whats causing this.

Does anyone have any ideas?

[SteffenRParticipant]

can you provide more information?

wich network…?
what else is connected to that network? how is it configured?

[Michael HeathParticipant]

This is the main AHNet connection; there’s a separate dedicated GigACE link to the mix rack.

It’s part of a large house of worship network, a /21 subnet. Currently there’s not much segmenting it off. But like I said, this is broadcast traffic originating from the console…..Why would there be so much?

[JayParticipant]

Steffan’s question on “what else is connected” was targeted at A&H, because it will affect the numbers below.

A MixRack, as of 1.72, is sending about 8 broadcasts a second, some advertising and some looking for devices. 4 are advertising its capabilities while the other 4 are searching for IP8, IP6, IP1 and GPIO. This doesn’t include ARP or any other normal network traffic.

A Surface (and Director) will send the same, with slightly different advertising. Depending on what other devices you have, they may also be doing advertising.

Connecting to the network on either MR or Surface will give you all the broadcast traffic from both devices. Do not connect both ends to the enterprise network as you will have a loop that is hidden to the enterprise switches.

Still, with all that, 22 broadcasts/sec (UDP + ~9 bytes) are nothing on a switched network. A windows laptop creates almost as much random broadcast traffic.

You really need to move the dLive system to its own VLAN on your network, but not for broadcast traffic. 2 issues:
(1) You are now carrying all broadcast and unfiltered multicast traffic across the GigaAce IP tunnel. This will directly impact the performance of the surface/scene recall/etc.
(2) The dLive security is through network isolation, not secure channels. Anyone on the network really has access to the system.

Jay

[Michael HeathParticipant]

Thanks, that clears up a lot.

I know that the broadcast traffic is not “a lot”, but as I was doing some cleanup and analysis it really surprised me to find that the C3500 was my single biggest broadcast device on the network. I just wanted to get a better feel on what was expected, and the readily available tech documentation made it sound like expected traffic was much lower than you described.

I definitely agree on the VLAN segmentation; I had just been slow because it requires some additional complicated configuration if, for example, I want to make sure iPads on our WiFi using the app can discover and connect fine. But I think I’m going to move forward with that this week.

I was surprised to hear about only having one end of the C3500-gigACE-Mixrack connected to the network. I think it may be setup this way. I definitely wasn’t seeing traffic that seemed to show a loop, but this may explain some performance issues.

Thanks for your help!

[JayParticipant]

Michael,

More than welcome. I was looking for 2 posts I made, but only found 1.

What I couldn’t find:

From discussions, reading and experimentation it looks like the MR and Surface are each based on an 8 port unmanaged switch without any features enabled (v1.72). If you count ports on MR (S Class), Surface is similar:
– 2 external ports
– 1 for GigaAce tunnel
– 1 for Mixrack hardware
– 3 for card slots
– (maybe other future connections internally)

This matters in an Enterprise environment because, as a switch, it isn’t passing CDP, LLDP, Spanning Tree, QoS tagging, VLAN Tags, Multicast subscriptions/filtering/querier or other management protocols. This means anything entering the dLive network will take bandwidth away from communication between the MR and surface, and this will create a loop between two Enterprise switches if both ends are on and spanning tree is non-existent.

Bandwidth info:

Bandwidth – GigaACE Network vs Direct

Jay

[Michael HeathParticipant]

I definitely didn’t realize the layer 2 topology was so simple and unseperated. Great to know.

One quick followup, after reading the post you linked: we have a Dante expansion card on our mixrack. We are only using the dedicated Dante port (as opposed to what they have labeled as, I believe, something like Dante+control). Is this port bridged to the other ports, or dedicated to Dante interface only?

Our Dante network is 100% separated from our house network for performance and reliability purposes. While one bridge point wouldn’t be the end of the world (it’s a different subnet and VLAN), I’d like to avoid it if possible. I know with most other Dante certified products, the Dante port is truly dedicated for Dante access.

[Author]

Posts