Network for Lighting Best Practice

[Ian H]

As we rush towards the Techno Fest that is Plasa.. I would like some thoughts on how best to approach a project....

 

Client has studio space with MA2 lite running conventionals and small number of movers via ETC2 net, now wants to add control over an additional two studios.

 

One will be all Led with no conventional dimming required and the other is a existing space with 6 48way racks controlled locally via DMX.

 

I believe ETC2net and ETC3net will coexist on the same network. but should that be at 10/100 or do we move up to 1ghz? what is the best network config for redundancy, and what network switchs are best currently. (POE will be required) what is best practice for monitoring the network?

 

Ian H

[Tomo]

Lot of questions there, I'm afraid that most are not answerable in a forum such as this.

 

ETCNet2 and Net3/ACN will happily co-exist.

 

Link speeds required depends entirely on the load - with <50 universes across all active protocols, 100Mb is fine with dumb switches.

- This does not include MA-Net as I don't know anything about how that works.

 

This isn't a hard limit - E1.17 ACN and E1.31 Streaming ACN traffic is variable and it can handle 'quite a bit' more.

(though I wouldn't want to get close to maxing the link out, if only to leave bandwidth for E1.33 RDMnet.)

 

Above this you start to need IGMP and want Gigabit links between switches. 100Mb stays fine for links to consoles and devices as no individual device is dealing with that many Universes.

 

On redundancy:

Exactly what do you need to be redundant?

 

- Redundant rings are very bad for real-time data as the packets often arrive 'out-of-order', leading to steppy fades. (It might be possible to solve this, however I've never seen a redundant ring system that worked properly.)

 

- Port trunking works well (redundant link, not redundant switch).

 

As for switches:

- Under 50 universes and no IT department:

Use good dumb switches and forget 'hot' redundancy. Have a 'cold' spare somewhere you can put in quickly.

Dumb switches start up in well under a second, the proper managed switches can take more than 5 minutes to boot.

 

For a big system (>50 universes) or if you really want 'hot' redundancy, then you'll need at IT department to look after it and a proper IT contractor to configure it. Configuration of managed switches is still a very skilled job, and small things make big differences.

 

So if you have an IT dept that want to look after the system, then ask them. They'll have a preferred switch and network management system.

 

Note that lighting networks are not the same as office networks. I've seen a couple of 'really clever' office networks installed for lighting, and they caused a lot of trouble.

[Ian H]

Thanks Tomo.

 

MAnet now seems to run on 1ghz infrastructure...I know MAnet and ETC are happy to share the same bit of copper, we have even squeezed artnet into the same system...

 

MA2net seems to prefer a switch to fiber now for runs over 300ft? anybody got any recommendations for good fiber enabled switches....

 

Client has a very strong IT team and their engineers just love running and terminating Cat5e and Cat6.

 

Ian

[Sully]

Ring based topologies can't be set up in a redundant manner with dumb switches that don't support spanning tree or link state tracking or port aggregation/channelling without generating a broadcast storm so no data let alone real-time data would be working very well in that state which could easily manifest as what Tomo describes. Link redundant wouldn't really buy you much in the way of extra reliability normally either since the thing that tends to cause any outages (for us) on desktop type switches is PSU or fan death. In an environment with non-fixed cable runs this might be a different matter but in fixed environments this should hold true unless the cabling is particularly rubbish.

 

For fiber based switches you'd be looking a device with SFP or GBIC ports, these are modular optic slots that accept modules for transmitting over different types of fiber (though you'll find you normally need to buy optics from the same vendor as the switch because the companies that make switches are weasels).

 

If you/the client has an IT team in-house you may find that if they're using managed kit using the same switches means they'll be familiar with the kit and they may already keep spares in house. The point about "office" networks is well made though, on a typical client access port configured by me 6 lines of config will be turning on measures to stop unruly ethernet client behaviours (and mostly aren't relevant if you run the lighting/sound networks isolated from the normal network badlands), custom configuration by someone who's aware of what's required or not turning any of those features on in the first place is necessary.

 

Is there any low-level information on manet and ma2net available I can read through actually? it's piqued my interest a little bit now and google's not turning those two up for me.

 

David.

[Tomo]

MA2net seems to prefer a switch to fiber now for runs over 300ft?

Copper (Cat5/5e/6) cannot be used for Ethernet runs over 100m.

 

You'll need to use fibre for that kind of run - it's pretty usual to use a fibre link between switch equipment racks (eg far end of studio to gallery).

 

The majority of managed switches (and some dumb ones) accept "Mini GBIC" modules for fibre connections like these.

You'll need to pick Mini GBIC modules suitable for both the switch and the type of fibre used - there are several kinds.

 

Ring based topologies can't be set up in a redundant manner with dumb switches that don't support spanning tree or link state tracking or port aggregation/channelling without generating a broadcast storm so no data let alone real-time data would be working very well in that state which could easily manifest as what Tomo describes.

True, but not what I meant!

 

In redundant rings the latency is variable as some packets go one route while others go another.

For most 'normal' purposes this doesn't matter, however in real time streaming data like DMX you can get variable spacing between packets, and even packet #2 arriving before packet #1, so the device has to throw away packet #1 as it's now unusable.

If it's only the odd packet it doesn't matter, but it's often a noticeable proportion.

 

That's not to say it can't be done, just that the real-time environment is not the same as office and internet type environments so needs a lot more care.

- I've yet to see a redundant-ring system that genuinely worked for real-time streaming data, and in some cases they were less reliable as the switch boot time was so long!

 

As the client has some good IT people, they're the guys to talk to (at least initially) about which specific hardware they'd consider, if only because they'll be looking after it.

- It makes a lot of sense to use the same switches throughout for holding spares etc.

 

Incidentally, most PoE switches sound like jet engines. Don't put one on the studio floor and they can be pretty annoying in the gallery!

 

Unfortunately it's only really possible to give general pointers in a forum or over the phone.

You'll need to get a specialist to do the design and to configure the network.

 

You can of course always ask ETC and MA Lighting for any comments on the proposed system design and switches.

 

Is there any low-level information on manet and ma2net available I can read through actually? it's piqued my interest a little bit now and google's not turning those two up for me.

Those two are proprietary protocols owned by MA, so you probably won't find much public info about them.

If you ask MA they'll be able to tell you things like the TCP/UDP ports and multicast ranges they use, and maybe the bandwidth typically consumed, but I doubt they'll tell you much more without a formal agreement.

[Ian H]

Thanks for the conversation so far guys.

 

Currently my thoughts are towards a 1Ghz network centered on the lighting control room with pairs of cat6 running to wall mounted network racks in each studio containing a pair of unmanaged POE switches - with all nodes split across two switches via a simple patch panel .... I was on a cruise ship this summer where the show tech had correctly identified that one of the two POE switches had failed and re patched for continued service but had failed to arrange a replacement...awaiting the failure of the 2nd - oblivious to the fact that if the 2nd switch failed he had no show.....

 

One of the Lds has now requested a Wlan in each studio for Ipad access to the MA2 -- now how do you shut down and hide or restrict access to every inquisitive smart phone or Pda

 

So if you see somebody rolling out a length of cat5 through the Plasa bar on Tuesday its probably going to be me trying to test the manufacturers claims on interconnectivity....

 

Ian

[fliggygeek]

Just to clarify. you are all talking about gigahertz networks. The word you want is Gigabit. As far as switches go, if you have a competent IT department, buy Cisco. There is a damn good reason why cisco is used in enterprise time after time. Also as was pointed out. Gigabit ethernet and 10/100mbit ethernet cant run at full speed over 100m. Always use fiber for runs over 90 metres. with regards to fiber, as far as I know, you cant get unmanaged switches with GBIC ports on them. This wont be bad, because if you have managed switches you can do cool things like link every one of them up to each opther and use vlans and redundant paths to other units, all updated in realtime, running on a shortest path first for links. that way if one link goes down the data reroutes itself to the destination via other paths without losing a beat.

[mac.calder]

One of the Lds has now requested a Wlan in each studio for Ipad access to the MA2 -- now how do you shut down and hide or restrict access to every inquisitive smart phone or Pda

 

 

Look for a checkbox or option on the wireless access point called "Broadcast SSID" - SSID is the name of the network. Disabling broadcast means you need to know the SSID and enter it manually to connect. It is not a replacement for additional security like WPA or WEP (also options on most routers)

 

To be honest, I would not worry about the switches dying too much. At least not enough to run dual switches in each space. Instead I would have a spare sitting in a box on the shelf. I also doubt you need Cisco - your requirements are quite low as far as management features. I love HP switches - they are a lot cheaper than Cisco and still very good quality. Got one that has been running without missing a beat for over 2 years with 35 path port nodes, 3 NSP's and an MA.