How best to define the AP wireless network?

  • 1
  • Question
  • Updated 5 years ago
  • Answered
We currently have 2 AP141s servicing classrooms. They have worked fine for a year. However this year we have increased the iPad population and we can now have 130 iPads attempting concurrent connections. We connected about 60, then the remaining failed to connect to the SSID. I understand that since iPads only use 5.0 Ghz we are overwhelming the 2 APs, and we are going to add 4 more AP141s (still just 1 Radius. Would this be sufficient for up to 180 connections?). However, I think we have another problem as well. Now we can’t connect the iPads that previously failed even when they aren’t concurrent. We suspect that we are running out of IPs defined in our DHCP pool. We currently have 221 IPs available in the pool. With the additional iPads we now have more than 220 devices that require an IP. They won’t all be concurrent, but within a 24 hr time period there will have been more than 221 requests for IPs. What is best solution? Change the subnet mask or define a VLAN? The subnet mask would be a lot of work because we currently have multiple statics from that pool and we are using GPOs. We are leaning towards a VLAN for all wireless devices.
Thanks
Photo of Steve Hoyle

Steve Hoyle

  • 9 Posts
  • 0 Reply Likes

Posted 5 years ago

  • 1
Photo of Sarah Banks

Sarah Banks

  • 75 Posts
  • 4 Reply Likes
Steve, I think the "right" answer can vary, significantly. Without a lot of detail, I'd say that extending/increasing the VLAN pool would be an easy solution; the other thing at play, concurrency aside, is the length of the DHCP lease. Even if you don't have 221 devices on the network at the same time, if the lease is long, say, for example, 24 hours, you could certainly envision a scenario in which the 221st device that day won't get a DHCP address. However, there's nothing wrong with separating traffic out with a VLAN, either. I believe you've almost settled on this approach from what you mention above. :) Good luck!
Photo of Brian Powers

Brian Powers, Champ

  • 396 Posts
  • 92 Reply Likes
Steve,

iPads can use both frequency bands (2.4 & 5). Be sure to utilize this via band steering so as to not overload a specific AP. You can also to load balancing if these APs are within a short proximity of each other to balance number of devices on each AP.

As for extending the DHCP, what device is assigning DHCP? You can fairly quickly & easily change the subnet mask there to extend that DHCP pool. As for your statics, you could just exclude them from the DHCP pool or even set DHCP reservations for those devices MAC addresses, set them to DHCP and they'll always pull the same IP from the DHCP server.

A separate VLAN for the wireless clients wouldnt be much of an issue either, but would probably require the most configuration changes of any of the options (wired switch port VLAN assignment, modifying the wireless SSID user profile, etc.)

As for adding 4 more APs (a total of 6?), be careful doing this sometimes to much RF can be just as troublesome as to little RF. Careful channel planning, especially on the 2.4 Ghz side of things to avoid any CCI/ACI. I'd expect 2-3 APs should handle upwards of 220+/- users fairly easily, but this can vary depending on the clients requirements.

See Aerohives High Density deployment guide, here for some tips on calculating the amount of radios and APs needed based on device type and applications usage.
Photo of Crowdie

Crowdie, Champ

  • 972 Posts
  • 272 Reply Likes
Devin Akin has an excellent video on deploying iPads in educational facilities at http://blogs.aerohive.com/blog/wi-fi-....

I know that I have used Devin's recommendations at several schools.
Photo of Steve Hoyle

Steve Hoyle

  • 9 Posts
  • 0 Reply Likes
These are all excellent and I appreciate them. Sorry I didn't get back sooner but we have major problems. Incidentally, we did not go with VLAN, we changed the subnet mask, and we did exclude statics from the pool. All seemed to work well. However, I have 2 141s and they can't handle anything more than about 25-30 iPads. If it goes higher then they reboot. We look at reports and can't get a handle on the problem. The iPads are only using Facebook on the web right now. We even divided 130 users into 2 groups, and they still can't run concurrently. Some get slow, some reconnect intermittently, and some that are connected can't send messages. One user was connected and just froze. We looked at her report and it showed excellent radio and zero bandwidth. Some said to add more APs, but how would that help this problem? The 2 APs service the area, everyone has signal, but throughput is horrible. I also have gotten different explanations when I call Aerohive support (including the part about iPads only use 5.0 Ghz, therefore only half capacity, therefore only 25 per AP. Then we spoke to someone else who said iPads do use both bands and we should be seeing 100 to 200 users per AP in out environment. I only know that both Aerohive and I are about to be thrown out of this account.

I will look at the video Crowdie recommended - tomorrow, I'm worn out. Thanks again everyone.
Photo of Andrew MacTaggart

Andrew MacTaggart, Champ

  • 483 Posts
  • 86 Reply Likes
Photo of Brian Powers

Brian Powers, Champ

  • 396 Posts
  • 92 Reply Likes
Regardless what Aerohive support told you, the iPad has been dual band since its inception (http://en.wikipedia.org/wiki/IPad).

And I've seen many a deployment with more than 25-30 devices per AP (even more than 30 devices per radio e.g. 60+ devices per AP) work and work well.

As for why the devices are rebooting at those numbers, I'm clueless as I've not seen anything like that in the past. Please keep us in the loop as to if you come across a resolution or more information that may enable folks here to assist more.
Photo of Steve Hoyle

Steve Hoyle

  • 9 Posts
  • 0 Reply Likes
Spoke to Aerohive. I guess the consensus is we need more APs. I have also been told that 1 wired AP can support multiple APs in a mesh configurations, but was told today that every AP serving a classroom or two should each be hard wired. Any thoughts?
Photo of Brian Powers

Brian Powers, Champ

  • 396 Posts
  • 92 Reply Likes
One AP CAN support multiple APs in a mesh environment. What could bring one AP to its knees (and make it reboot possible) would be if it were serving a mesh environment & 60+ devices directly connected to it as well as all of the devices connected to the remote device.

Is this how you are setup currently with your two APs? One wired back and one using one of the radios as backhaul?
Photo of Steve Hoyle

Steve Hoyle

  • 9 Posts
  • 0 Reply Likes
We are one wired back and one using a radio as backhaul. We actually have had 10 users in the same room as the mesh AP (no other users) and they couldn't reliably connect. The mesh and the wired AP are about 70' apart and there are 4 standard sheet rock walls between them. Is this a problem? However, even when we had students in the room with the wired AP there were problems. We did seem to see improvement in our last test when we simply unplugged the mesh AP. We are not getting anywhere near the capabilities that others say this setup should support. We were on phone 1 hour with Aerohive and he went through our HiveManager and didn't detect anything out of the ordinary. I'm really stumped. The only thing we have tried so far on these APS is a couple groups of students only going to Facebook with their iPads. One AP was serving 20 and the other (the wired) was serving 40).
Photo of Brian Powers

Brian Powers, Champ

  • 396 Posts
  • 92 Reply Likes
Steve,

SSH into the wired AP and issue a "show acsp neighbor" command. What does it show as the RSSI between the two APs? Do the same on the other side an report back if you don't mind.
Photo of Steve Hoyle

Steve Hoyle

  • 9 Posts
  • 0 Reply Likes
Brian, these are the results of the 2 tests you asked us to run. Thanks for your help.
The first image is for the mesh AP (Science room) The second image is for the wired AP (Math room). Thanks for your help.



Photo of Brian Powers

Brian Powers, Champ

  • 396 Posts
  • 92 Reply Likes
Steve,

Do you have the 5GHz radios set to "Dual", where they are serving clients and acting as a backhaul link between the APs? Without getting into lots of details, if this is the case, than that radio is in the most simplistic terms spending 50% of its time serving clients and 50% sending the data from the clients to/from the APs to get the data on the wire and back to the clients. Bear in mind wireless is a shared medium and all the clients associated to the radios on Channel 149 and the radios communicating with each other on Channel 149 are all contending for the same chance to transmit/receive data.

http://www.revolutionwifi.net/2012/02...

http://www.wlanpros.com/wp-content/up...

Also, @ the RSSI that is showing between the two APs, you're probably looking at a data rate in the range of MCS 2-4 (teens to low 20s Mbps). This is assumed as Aerohive doesn't seem to document all data rates to RSSI values (and even if they did, it would be based on tests conducted in an anechoic chamber (which is not feasible in the real world).

All that to say that the RSSI between the APs could be improved, or just use the 5 GHz radio for the backhaul only and let the clients connect to the 2.4 Ghz band. Because if you're still working with +/- 130 iPads and we divide them up evenly, dumping 65 clients on one radio is not ideal, but its also not the worst thing that could happen. As I've seem higher numbers on some radios in a classroom environment. This would allow you a dedicated link between the APs and dedicated radios to serve clients.

Will this improve your situation? Hard to say. Would a wire to the 2nd AP improve the situation? I'd believe so.

If thats not possible and we're working with the only possible design, attempting the above could yield better results. You could also put directional 5 GHz antennas on the two APs and aim them at each other to help improve the connection between the them.

On a side note, things like this are what your vendor should be assisting you with (hence the V in VAR - VALUE Added Reseller). Not that I or anyone else on here doesn't want to assist, but having feet on the street to see things helps and can provide lots of valuable information.

That said, feel free to continue to update this thread with progress, questions and the like and we'll assist in any way we can.
Photo of Crowdie

Crowdie, Champ

  • 972 Posts
  • 272 Reply Likes
Steve,

The 5 GHz RSSI values from your screenshots are no low for a stable mesh connection. Personally I wouldn't make a mesh connection with an RSSI lower than -70 dBm.

When creating mesh connections the antenna selection is really important. I tend to head towards deploying semi-directional antennas on the AP-141 or AP-350. If you have deployed omni-directional antennas this could explain the poor 5 GHz RSSI values.
Photo of Steve Hoyle

Steve Hoyle

  • 9 Posts
  • 0 Reply Likes
Brian + Crowdie,

You guys have been a tremendous help to us. Not only do you know this topic thoroughly, but you take the time to clearly explain, and we are learning through this process. Incidentally, we did buy 141s because I wanted to begin with the best radio coverage possible. I have decided that mesh should be the last choice, only implement if necessary. The solution isn't worth the degradation if it is only to save money with less APs. It seems to me it is somewhat of a wash; you save money with a mesh but then degrade to the point that you don't support as many users, and would then probably need to add another AP in the mesh. I think manufacturers have over specified mesh capabilities so as to maintain a competitive marketing edge. I think I saw someone say that Meru used to always mention mesh, and now they never seem to bring it up in their advertising.

It was a problem getting additional wires from more APs to our backbone switch (some would be over 300' and we wanted POE, and it is an old concrete building, so running wires is a nightmare.) However, I'm going to locate an 8 port POE close to the classrooms. I'll use the one wire which now runs from the backbone switch to the AP, and connect that to the new 8 port POE. I'll then run from the POE to every AP (about 125' max). We will have a total of 4 wired APs for 6 classrooms, about 30 students for each classroom. No AP will have to provide coverage through more than 1 standard sheet rock wall. I'll post results here. Anything I should know about radios and channels for this 4 AP configuration? I plan on having each of the 4 antennas pointed about 45 degrees from vertical (APs are mounted on the ceiling in center of classroom) and will use both radios to cover the student devices. The layout is 3 adjacent classrooms, then a skipped classroom, then 3 more adjacent classroom. So I'll have 2 APs covering a group of 3 classrooms (90 devices). The classrooms are about 20x25.

Also, I don't think our VAR would probably do feet on the street since we are just beginning with Aerohive and have only bought the 2. However, our client wasn't going to buy more until this was resolved, and we weren't going to put this in any of other clients either till it was resolved.

Thanks again,
Steve
Photo of Brian Powers

Brian Powers, Champ

  • 396 Posts
  • 92 Reply Likes
Steve,

I couldnt agree more with your stance on mesh. It was a last resort we used when we were in areas where it just wasnt possible to get a cable to the location. I treat them almost as a point-to-point and always tried to dedicate radios for the backhaul only.

As for your changes you spoke of, I'd let ACSP take care of the channel/power settings on the access radios. I'd use something like inSSIDer or the likes to find the least used 5 GHz channel and manually set the backhaul links, but thats more of a personal preference and not mandatory as they will link up with their channel settings set to Auto.

And in my opinion, this is exactly what a VAR should be assisting with. If I had a nickle for all of the sites I visited with my old job meeting with potential customers and NOT winning any work from them, well, I'd have several pockets full of nickles! But its part of the job. When a VAR does this, it help avoid hurdles like this and just makes for an overall better customer experience.

Either way, glad you've come up with what looks like a remedy to your situation and feel free to keep us updated as to progress: good, bad, indifferent. Always happy to assist when we can.
Photo of Crowdie

Crowdie, Champ

  • 972 Posts
  • 272 Reply Likes
When I am using mesh for a reasonable distance I always use AP-350 access points rather than the AP-141 access points. Wireless clients will detect a higher RSSI from an AP-350 than an AP-141 all things being equal.

If we take a scenario where we have two access points with three brick walls between them and we want to make a mesh connection between the two access points we need to ensure that the non-cabled access point detects the cabled access point with a RSSI of -70 dBm or greater (greater is better).

If you stand next to the access point you will most likely get an RSSI of around -35 dBm from an AP-350 and -45 dBm from an AP-141. Here is the first issue - the AP-350 can lose 35 dBm and still hold the mesh connection while the AP-141 can only lose 25 dBm and hold the mesh connection.

We have three brick walls to pass through so stand on the cabled access point side of wall #1 and get an RSSI reading. Now stand on the non-cabled access point side of wall #1 and get an RSSI reading. You now have the RSSI loss caused by the wall. If that RSSI loss is 10 dBm and that loss is consistent with the other two walls:

* If the access points are AP-350s then the non-cabled access point will detect the cabled access point at -35 dBm less 10 dBm less 10 dBm less 10 dBm = -65 dBm (ignoring free space loss through the air) so the mesh connection will work.

* If the access points are AP-141s then the non-cabled access point will detect the cabled access point at -45 dBm less 10 dBm less 10 dBm less 10 dBm = -75 dBm (ignoring free space loss through the air) so the mesh connection will fail.

Add on to that the free space loss, which is considerably higher for 5 GHz transmissions rather than 2.4 GHz transmissions, and you can see my trying to hold a 5 GHz mesh connection over distance is extremely difficult.