E1.31 - Network Switch

[Dave H1]

Getting closer to stepping through the "pixel door" for this season......
 
A simple question as to using a network switch with the controllers.  I see in a previous presentation for E1.31(over on ACL forum), that a 10/100 Mbps switch is adequate.  Is that the switch everyone is using or does a 10/100/1000 Mbps switch have any advantage on overall performance???

Edited January 30, 2013 by Dave H1

[Dennis Cherry]

That is all I used last season.

 

AT&T 2Wire Router

 

Show computer plugged into one port.

 

E1.31 plugging into another port.

 

No changes done to the 2Wire router configuration.

 

Configured E1.31 in S3 as the default Multicast 239.255.0.1 thru 239.255.0.4.

 

In S3, disabled all LOR protocols as my display was 100% DMX.

 

That was it and ran without a hitch.

 

Even could use my wireless laptop for display testing purposes only and still had wireless internet access on the laptops.

Edited January 30, 2013 by Dennis Cherry

[thevikester]

In reading alot in regards to the switch, 10/100 is more than adequate to do the job that we are doing, I also read on the aussie form that even though your computer may support the gigabyte switch, the hardware or controllers do not, so save your money and go 10/100

[hdracer]

10/100 is all you need. Don't waste your $$ on the gig switches. Just make sure you count how many e1.31 controllers your gonna be using in total and buy the switch that has enough ports to support your needs.

[dougd]

Go to your IT department and ask them if they have any old 10/100 switches. Those guys don't throw anything away and they still have the old ones when they switched them out for gigabyte switches.

[LOR Staff]

I know this was posted here at one time, but perhaps it was purged along with a couple of users....  This is a repost of something we were discussing with the Beta of E1.31.  While the examples here are for Gb switches, the base concept is still the same:  some hardware is going to work better than other hardware.

 

I'm not saying that a 'cheap' switch won't work, or that an 'expensive' one always will.  There are simply too many variables for blanket statements when it comes to E1.31.

 

 

----------------

 

As more and more things work towards E1.31, we also need to remember that there are going to be a lot more variables involved when it comes to performance. Even specing identical computers, network cards, switches, and end devices you are still going to get 2 different throughputs on 2 different installations. That is not to say that you can't get a feeling for the maximums, just that you are going to be hard pressed to find an 'absolute'.

 

Now once you mix in different equipment (network card, switch), and the numbers are going to be all over the place.

 

We like to think that a 100Mbps network gives us 100Mbps of throughput, and a 1Gbps gives us 1Gbps. We also like to think that switch A @ 1Gbps is the same as switch B @ 1Gbps. Neither is true.

 

For example, look at these 3 24/26 port switches. I've selected business class switches since they will typically advertise their specs. The switches you use for your home network are going to be FAR lower in ratings:

 

Switch 1 - $85

Switch 2 - $290

Switch 3 - $2,100

 

All 3 of those are Gigabit switches, so you would expect to see the same performance on all of them. No way Jose. There is a definite reason why there is a $2,00 price gap.

 

Switch 1 has a throughput of only 8.8Gbps. This switch can be fully saturated with as few as 4 senders and 4 receivers. Latency is going to be an issue as well (see #2 below)

 

Switch 2 has a throughput of 52Gbps. That means the switch can not be saturated - 13 inputs can map out to 13 outputs @ 1Gbps (or any combo) bi directionally. The problem with this switch is going to be switching speed. Assume you have 1 sender, and the data from there is multicast and needs to go to the other 25 ports. Latency could be an issue here since the switch has to take a single packet and move it to 25 separate destinations. Oh and that first switch? I don't even want to think how bad it will be.

 

Switch 3 not only has full bandwidth, but super fast switch speed and processor power (remember, a switch is nothing more than a computer with specialized hardware to move data from one physical connector to another). It can be considered a backbone switch. -- It can quickly look INSIDE the packet of data and figure out the best and fastest way to route it, regardless of what the IP on the outside says.

 

Wireless complicates things even more. Again just because it SAYS it is a 56Mbps wireless (G) and Windows reports you have a 56Mbps connection, doesn't mean you'll get 56Mbps. Take a look at this chart. It shows relative throughput for a bunch of different wireless routers. See the differences?

 

I can personally see the differences here in the office/lab. I have a 35/35 connection to the internet. On my wired computers, it's 4 hops to the outside world and I'm always at a solid 35 down 30 up. On the laptop, and other wireless devices, I'm only 1 hop away (well, actually 0 since my providers router is my wireless router). I have NEVER been able to pull more than 18/10. It's G wireless which should be 54Mps. Why? The throughput on their WAP stinks.

 

And then don't get me started on cabling, microwave ovens, channel conflicts, wireless phones, or french fries instead of pizza which can all cause you to have a bad time. (Thumper says "If you french-fried when you should have pizzaed, you're gonna have a bad time.").

[edvas69]

But the reality is for the most part a cheap switch will do, I used an old 10 Mbps switch in 2011 for 5000 channels and saw no issues, in 2012 I upgraded to a $20 second hand 100 Mbps managed switch with 10000 channels and network performance was peaking at 12.5% with 32 universes (not all universes were a full 512 channels) So unless you are planning on putting on big shows then a standard cheap 10/100 Mbps swith will suffice. The potential issue is actually with data flooding controllers using multicast with very large installations and this is where you would look at using unicast and a managed switch. But again for most this will not be an issue, only those displays that are pushing channel barriers should consider this.

If it gets to a point where you need to start closely managing your data throughput then its cheaper and easier to just add another netwotk interface card to the computer and then split the E1.31 network into 2.

 

But as Dev Mike has touched on switches are not all the same, but i dont think its something we need to spend a heap of money on for lighting and e1.31 which is just a single way transmission as the cheap ones will suffice nicely.

[Dave H1]

Thanks guys

 

I don't think I can go wrong (price wise) with this one.  May cost more in shipping than the unit

 

http://www.monoprice.com/products/product.asp?c_id=105&cp_id=10521&cs_id=1052104&p_id=7854&seq=1&format=2

[edvas69]

Thanks guys

 

I don't think I can go wrong (price wise) with this one.  May cost more in shipping than the unit

 

http://www.monoprice.com/products/product.asp?c_id=105&cp_id=10521&cs_id=1052104&p_id=7854&seq=1&format=2

 

 

Did i see right, $8.76 for a 8 port 10/100 Mbps switch, thats just insanely cheap

[TitusCarnathan]

I used a netgear 1000v2 wireless router and it worked flawless.

Edited January 31, 2013 by TitusCarnathan

[Dave H1]

 

OK
Should have asked this in the original posting,
Of the members that utilize network switches, do you keep them inside and run mutliple cables to the exterior or is it acceptable to locate the switch outside (protected of course) and branch out from there??

[bob_moody]

Just for clarification sakes, the comparisons made in the post shown by DevMike are not really apples to apples comparisons.

 

The first switch talks about 8.8 Gbps Switching Fabric. The switching fabric is basically the backplane where all connections are made and are more compared to switching capacity.

 

The switching capacities being shown are not the numbers for switching capacity in switch #2

The SG200-26 referenced has a "switching capacity of 38.69Mpps (Millions of PACKETS per second). The 52Gbps is the forwarding performance based on a 64 Byte packet.

 

Look at the Cisco SG200 switch and compare it to the equivalent 10/100 switch known as the SF200. basically the same switch but without the 2 extra 10/100/1000 ports . The switching capacity drops from 38.69Mpps to 6.55Mpps and the forwarding performance drops from 52GB to 8.8GB.

 

You pull the gigabit out of the mix and the first switch is actually a little better than the Cisco SF200. ( I know the reference was to the SG200 but the point is what addign Gigabit does to the ratings)

 

The third switch is an enterprise class switch which follows a whole new set of rules (Layer3, VLAN support etc).

 

It would be interesting to see one of the larger pixel displays pull thier stats from the sending controllers (I use the J1Sys D2 and the P2 and do not have a big display) and see packet traffic.

 

I think eddy's (et. al.) observations would be supported regarding the switches they use.

 

I'm not up to doing the math, but exactly how many pixel channels would it take to overrun 6.55 million packets per second?

 

Bob

[maccnh]

Best one i found so far and runs off of 5vdc to so you can get power off the E682 Aux plug to power it up  http://www.microcenter.com/product/257750/TE100-S5_5-Port_10-100_Fast_Ethernet_Switch

[TitusCarnathan]

 

OK

Should have asked this in the original posting,

Of the members that utilize network switches, do you keep them inside and run mutliple cables to the exterior or is it acceptable to locate the switch outside (protected of course) and branch out from there??

Mine ran fine outside I used one of the CG-1000 Enclosures