Forum

Hello All,

I'm working on a wireless project and need some advice for a portion of the design and thought I would see if anyone has used a similar configuration.

Here's the background information:

Vehicles equipped with wifi pull into a yard every night and are parked. While in the yard, the vehicles upload data from 3 vehicle systems via wifi to central servers where the data is stored. There are 175 vehicles per yard. The 3 onboard vehicle systems (computers) will be hardwired via Ethernet to a ?device?, which will allow the 3 computers to share a single onboard wifi connection. In other words, I don't want all three onboard computers to have their own wireless NIC and antenna (and roof penetration).

To create the on-vehicle network, I'm looking at using an access-point in bridge mode or workgroup bridge mode plus a switch and maybe a router (if needed). The switch connects the 3 computers to the bridge and the bridge connects the vehicle network to the central fixed servers via wifi.

Here's the question:
The fixed access point in the yard would basically be a point-to-multipoint bridge. Could each fixed access point handle, say, 25 vehicles? In other words, is there any additional overhead, limitations, or inefficiencies with a point-to-multipoint bridged connections compared to a standard access point to client connection that I need to take into account?

Would using the work group bridge mode scale better?

Is there an alternate on vehicle arrangement that I have overlooked that might work better?

By the way, this customer uses Cisco gear, so I will look at Cisco gear first, but other vendors might be considered. Any help would be much appreciated. The fixed equipment will likely be a controller based 802.11n at 5 GHz. It?s the on vehicle piece I need help with.

How much data are you trying to move per vehicle? Vehicles will arrive at different time?

I don?t have all the information at this point. I?ll be collecting additional details in the coming weeks. My understanding at this point is that vehicles do not all arrive at the yard at the same time and would likely arrive over a 3 hour period.

One of the on-vehicle systems is a video system, and is estimated to send 300-500MB per vehicle per day over the WLAN.

However, before I begin the infrastructure design, I want to understand what the options are for the on-vehicle network equipment. I?m not aware of a standard 802.11 client station that can be used to allow multiple wired computers to share a wireless connection (however, this device may exist and I?m just unaware of it). This is why I?m looking at using an AP in a non-standard 802.11 mode (bridge mode or workgroup bridge mode) on the vehicle.

Can a single AP support connections from, say, 15 to 25 AP?s in bridge mode as efficiently as a single AP would support 15 to 25 standard 802.11 clients? If not what are the issues? How about work group bridge mode?
Thanks

We have a customer that does this with police cars to upload their in-car video for the day. There were some pretty strict requirements, for example the data had to be uploaded fast enough to occur only in the time it takes to change shifts as the same police car will be used again for the next shift.

I work for Ruckus Wireless so of course we used Ruckus gear. :)

We used our outdoor 11n dual band AP (7762) mounted on the buildings and our 5 GHz client bridge in the cars. 5 GHz was chosen in this situation (and should be used in yours as well) due to the high number of channels it will support and thus give you a high number of simultaneous data streams.

How many outdoor APs you would need is of course a factor of number of clients connecting and coverage area needed.

The way the system ended up working was when the vehicle entered the yard, the Ruckus system auto connected and the video software did the rest (started the upload).

Let me know if I can be of any help. Even if you would rather go with a different vendor, we'll try to get you hooked up. Thanks!

GT Hill

Something that you might want to try is installing an Aruba MSR4000, config all 4 radios for 5GHz, install MSR1200?s in the vehicles and config them only to connect to the MSR4000. These are all mesh AP?s so you could connect them to a small switch in each vehicle. Yes it?s pricey, but performance always is!

btw I work for an Aruba partner and was a beta participant for the AWMP program

GT,

Thanks for the feedback.

We will also have similar strict requirements, and time limits to get the data moved off the vehicles each evening. However, each yard is covered by a tall canopy that provides shade / protection for the vehicles. I will have access to a fiber backhaul, power, and AP mounting locations in the canopy. The AP?s can be direct connected to a PoE switch in the canopy, so I won?t have to waste any frequency creating a mesh. Also, with all the channels available using 802.11n at 5 GHz, I think we will have plenty of capacity available. So at this point I?m not too worried about the infrastructure design.

What I am concerned about is the vehicle side and the relationship between the wireless device on the vehicle to the AP in the canopy. I suspect each vendor will have a somewhat non-standard or proprietary solution since the 802.11 standard does not explicitly provide for having multiple computers, sharing a single client station (to my limited knowledge). And when I say proprietary, I don?t mean that in a bad way, I?m just saying that the bridge, workgroup bridge or client bridge that is on the vehicle will most likely need to be of the same vendor as the fixed AP in the yard. I?m guessing there some extensions /enhancements needed to allow the AP to send frames destined for the multiple MAC addresses of the multiple systems on the vehicle to a single wireless client station.

This is where I have some uncertainty. All other things being equal, can a single AP in the yard support just as may associations to AP?s in bridge mode, as it would support normal wireless clients associations? Are there any nuances or got-ya?s when using bridge mode, workgroup bridge mode, or a client bridge device? Or should I just do the design as if each vehicle was a standard wireless client.

For example, Cisco has a workgroup bridge configuration example document in which they make the following statement

[i]If you configure APs and bridges to treat a WGB as a client device, you allow more WGBs to associate to the same AP or to associate with use of a service set identifier (SSID) that is not an infrastructure SSID. The performance cost of reliable multicast delivery?in which the duplication of each multicast packet is sent to each WGB?limits the number of infrastructure devices (which includes WGBs) that can associate to an AP or bridge. In order to increase beyond 20 the number of WGBs that can associate to the AP, the AP must reduce the delivery reliability of multicast packets to WGBs. With reduced reliability, the AP cannot confirm whether multicast packets reach the intended WGB. So WGBs at the edge of the AP coverage area can lose IP connectivity.[/i]

I?m not sure I quite understand the above statement, but it certainly gives cause for concern. If anyone could elaborate, I?m all ears.

Thanks
-PBK

Sirkozz,

Thanks for your input. A mesh router may be something to consider.

It could be a definite advantage having a router onboard the vehicle. I?m concerned that some of the systems on the vehicle could be poorly design or have unneeded service running that might create unnecessary L2 broadcasts. 175 vehicle times 3 systems per vehicle would create a very large flat network and associate broadcast domain. Would a mesh router terminate the L2 broadcast domain on the vehicle network like a traditional router would?

Also, see my post above about having wired Ethernet access to all the fixed AP?s in the yard. Does this impact your suggestion at all about the mesh configuration at all. Also, would the MST200 be a good candidate on the vehicle?temperature and vibration will be a big concern.

Thanks,
-PBK

Look I?m not a real big fan of indoor Aruba stuff, you know job security and all, just throw darts on a map and place the AP?s and the controller will do the rest. However in the mesh market the Azalea products bought by Aruba and upgraded are really good products and still provide job security. You can deploy the AP?s in either layer 2 or 3 mode. The real advantage is that the radios are software configurable, so that you can split the network in half and then half again. As to your concerns about vehicle mounting an MSR1200 I think you?d be ok, check out the data sheet.
The reason I mention the MSR4000 is that it?s a quad radio AP, meaning that you can create a mesh WLAN off of each radio, and then divide the vehicles by 4 and then have them join only 1 of the mesh WLAN?s. Size/configuration of the yard would determine what actual root AP?s you would need, MSR2000?s are dual radio AP?s, the half-brother of the MSR4000, still software configurable per radio.

IMHO Aruba mesh is the Cadillac of mesh, having a great deal of experience with Belair, Moto and Cisco.

Also it?s always been kinda funny to me that Aruba doesn?t troll the board?

PBK,

You could definitely use Cisco APs/WGBs for this kind of deployment. The thing you have to bear in mind with wireless, though, is that it is a shared spectrum - the more client devices you have connecting to a single AP, the less throughput each device will have. The limitation on the number of workgroup bridges allowed to connect to a single AP is only there to allow reliable multicast traffic. If you remove this limitation, more WGBs are allowed to connect to the root AP - perfect for your scenario as long as throughput is not an issue. I would definitely recommend capacity testing.

So now to the crux of the problem - how many clients/WGBs to allow per root AP. It really does depend on your throughput requirement. To my knowledge, the outdoor APs offered by Cisco are 802.11a/b/g (you should probably verify if any 802.11n bridges are available - I've heard there is something in the pipeline). With that in mind, the most throughput you can expect is 54mbps over the RF. At 20 client devices, this translates roughly to 1mbps/client (due to overhead of the 802.11 medium, noise, etc.), assuming all the clients connect at the 54mbps data rate. If you have clients connecting at lower data rates, the throughput for all clients will decrease, not just those connected at the lower data rate.

I think the best approach is to determine the required throughput, leaving yourself about 15% - 20% overhead in case there is noise/interference/growth in data, and then work the math backward from there to determine the number of root APs to install.

HTH

Tim

Tim,

I?ve been trying to avoid talking about the capacity in this thread, but here?s a rough back-of-the-napkin calc.

Let?s say I need to off-load 100 GBytes of data per yard per evening and I have 4 hours to do it. Assume I?ll be using 802.11n at 5 GHz w/ 2 spatial streams, 20 MHz channel, guard interval of 800ns, and a maximum data rate of 130 mbps. After protocol overhead, contention, interference, etc, I?ll arbitrarily assume an actual application data throughput of 30% the maximum data rate. By the way, I have no idea what this percentage should actually be.

So each AP can handle:
* M bits / sec :130 Mbps * 30% = 39 Mbps of actual application throughput
*G bytes / hour: 39 Mbps * 3600 seconds / 8 bits per byte = 17,550 M bytes / hour OR 17 G bytes/hour
* In the 4 hour window each AP can off-load 68 G bytes of application data

So, if the above is correct, two AP could move the required 100 GBytes of data in less than 4 hours.

However, I will likely need at least 6 APs per yard to provide proper coverage, at which point it would take less than 1 hour to move the data. Cost is not a big issue for the fixed side equipment, and there are plenty of channels available, so maybe I would us an AP w/ dual radios, or use 40 MHz channels, and half the download time again resulting in less than 30 minutes to download the 100 G Bytes.

Remember, this is meant to be a back-of the-napkin, rough order of magnitude calculation. However if anybody sees a gross mistake in the logic or math, please speak up. A proper analysis and design will be done when the times comes.

-PBK