Slow Wireless Hotspot Implementation - observations and solutions

RouterOS Wireless Networking
1

We have deployed a wireless network for a hotel camp and it is unreliable and slow. I am so tempted to remove it but trying to find knowledgeable consultants to help us.

I have 2 consultants/vendors working on this project. I do not like their diagnosis of why the network is slow and keeps disconnecting. Bear me with as I explain the configuration.

  1. A data room hosts a CCR1036-8G v6.32. There are four 100Mbs Ethernet lines from ISP feeding into it.
    A test of their combined speed using Ookla http://www.speedtest.net/ shows around 340Mpbs. I called the ISP and they explained that they need to update their fiber optics in the future.

admin@Mikrotik>interface print

NAME TYPE ACTUAL-MTU L2MTU MAX-L2MTU

0 R ether1 ether 1500 1590 10226
1 R ether2 ether 1500 1590 10226
2 R ether3 ether 1500 1590 10226
3 R ether4 ether 1500 1590 10226
4 R ether5 ether 1500 1590 10226
The interface list of the output Ethernet (ether1)on CCR1036-8G shows an average Tx of 50Mpbs -130Mpbs even during peak hours.
The only time it soars to say 250Mpbs if i have the administrator PC directly connected to it to do a speedtest.net then Tx goes to that level momentarily. Why is that? Shouldn't it reach 200+Mpbs during peak hours if a majority is complaining about slowness and disconnection.

  1. Mikrotik CCR1036-8G output ether1 is connected to Cisco Nexus 930 switch where an administrator PC , and two RouterOS RADUIS servers are connected to.

  2. Two RADUIS services feed a 52 port Cisco switch with 5 cables which in turn has 16 Ethernets servicing the 16 building blocks.

  3. RADUIS hotpot server 1, has RouterOS installed on it. And has one internet Ethernet cable from CCR1036-8G through the Cisco Nexus 930.
    It services 4 blocks M, N, O, P .

NAME TYPE ACTUAL-MTU L2MTU MAX-L2MTU

0 R Local-Network ether 1500
1 R M ether 1500
2 R N ether 1500
3 R O ether 1500
4 R P ether 1500
5 R internet ether 1500

  1. Each block has a Linksys 12 port switch where two TP-LINK EAP120 AP is setup on each floor. Each block has 3 floor thus a total of 6 AP in each building. The floor is 40 m long and the AP are set at around the 12m mark from either end.

  2. Second RADUIS server services 12 blocks A-L. They are setup as VLANs. One cable is connected to 52 Cisco port(the first server has one cable for each block and no VLANs )

NAME TYPE ACTUAL-MTU L2MTU MAX-L2MTU

0 R Local-Network ether 1500
1 R internet ether 1500
2 R A vlan 1500
3 R B vlan 1500
4 R C vlan 1500
5 R D vlan 1500
6 R E vlan 1500
7 R F vlan 1500
8 R G vlan 1500
9 R H vlan 1500
10 R I vlan 1500
11 R J vlan 1500
12 R K vlan 1500
13 R L vlan 1500
The choice of 12 vs 4 blocks onto 2 hotpot RADUIS was experimental, considering the 4 blocks account for 40% of the users.

  1. HOTPOT profiles are set at 1Mpbs, 512kps and 256Kpbs. 70% of users are on the 1Mpbs.

Here are my observations:

  1. There are around 800 paid Hotspot users across the 16 blocks. about 30% of them are from the 4 blocks.
    At peak time there are 500 Active users showing in winBox. about 350 in the 12 block RADUIS and 150 in the 4 blocks.

  2. End rooms 1,2,3 and 16,14,15 do not have a good signal. Should we add more AP, say 3 on each floor?

  3. Connect and disconnect to the Network. Many Users complained about this. There is a general frustration with the network.

  4. the consultants told me that at around 420+ Active users the network deteriorates. They say that traffic crowding happens and attribute to the slowness of the network and the connect/disconnect problem.
    They explain that considering that i have 400Mpbs from my ISP spread across 1Mpbs for each user, the network cannot handle more. I find it hard to believe as constantly watching the CCR1036-8G's Tx it is around 120 Mbps, I'm yet to see it reach the 200+ Mbps!
    His answer was that all the users have to use Youtube at once for it to reach that level. I don't believe that because even with them randomly using the internet they complain about slowness and disconnecting and it is not even using the 340 Mpbs available to the CCR1036-8G.

  5. this network was supposed to service 3000 hotpot users, we are struggling with 500 active ones!! I see about 60 users with Tx reaching 1Mpbs, possibly using Skype or Youtube, however the remaining majority complain about slowness(not having an actual speed of 1Mpbs).

  6. AP were configured to have 16 SSID e.g. A-room-1 etc.. I want that changed. We are adding complexity for no good reason.
    Set limit of each router close to its max concurrent allowed per floor. Each floor will have 3 or 4 Wi-Fi networks not 16 as we currently have! What do you think?

  7. Some time the entire floor has problem connecting. Based on what i was told, some users have repeaters and router with DHCP set up on it. This way they can buy one hotpot user and have say 6 people use it. They show up as SSID other than the user A-Room-1 etc. So the consultant suspect that the DHCP service is contradicting with that of our RADUIS and CCR1036-8G causing problems for the entire floor. How can i validate that. I asked my staff to go hunt for those misfits.
    How can we make the CCR1036-8G and RADUIS servers oblivious to such rogue devices?

FINAL Question, does anyone recommend a Mikrotik, hotpot consultant in Europe and MENA.

Thanks

2

Too many factors to take into account. I think no one can give you a fast answer. The only advice i can bring to you is: Make small pieces of you big network and test each alone for finding the problems and/or bottlenecks. I’m sure your deployment can run fine.What AP are you using? Are you using CAPSMAN?

3

Yes we are using CAPsMAN license level 6.
AP is TPLink EA120 http://www.tp-link.com/en/download/EAP120.html
I was hoping we can shed a light on any of those points. Or even have customers share consultants who have experience with solving such problems.

For observation #2, do you think that adding more AP is better. What is the frequency, protocol 802.X and physical distance apart that one recommends?

For observation #4, Should winBox Tx reflect a higher value than the range of 50-120 Mbps. Even during Peak hours we are not getting close to 320Mpbs that available. I want to make sure that is not normal unlike what the two consultants are trying to convince me!?

I understand this is not easy and I’m not expecting an answer to all my problems. But any feedback and direction i can get from end users all add up.

Thanks in advance

4

If your AP’s are NOT Mikrotik then you can NOT use CapsMan. Your config becomes more complicated now if you can not manage your AP’s from a single point. Do not focus in your ISP available bandwith, I think your problems are in how the hardware is configured and deployed.

5

Your problem is on the wireless part for sure. Self interference maybe.
You should look for a Tplink consultant not Mikrotik. I don’t understand how can your problem be Mikrotik related, but wait, maybe there will be more replays to your post.

6

One thing I did not gather was if users connected via cable experience the same problems… I mean you said you used the admin PC in place of the CCR, but what if the admin PC is connected to the CCR, but over a cable? What about a bandwidth test from the CCR to another external router (with equal or greater ISP attached)?

Doing those will tell you if the Wi-Fi has anything to do with that, or if it’s purely the CCR’s fault.

Certain rules in the router could effectively drain the total bandwidth, which in turn means that even if you have 400 users online, they might not all get their 1Mbps. Combining several cables into one (“bonding”) can be one of them. Hotspot is also another. But to find out the actual effect of each, you’d need to do the above two tests, preferably without other users being connected alongside you (to minimize their effect on the CCR’s CPU, which BTW would also start to deteriorate near that number of users).

7

omg 16 ssid !!!

that cause too many overhead using 1mbps beacons you are wasting 50% of capacity

look at this

http://www.revolutionwifi.net/revolutionwifi/p/ssid-overhead-calculator.html

8

One ssid per ap is more than enough, too many are confusing and chechito might be right.

9

Thank you guys for the input. I will start by examining APs and removing unnecessary SSID.

10

having deployed several hotel wifi setups, I have found that 99% of connectivity/slowness issues are caused primarily by Interference and weak clients due to dead zones.

Best practices ive found so far.

  1. ban all personal wifi networks from the building.
  2. frequency plan so you dont self interfere (auto channel setting doesnt work for squat)
  3. two SSID’s 1 for 2.4ghz, 1 for 5ghz.
  4. NO WDS, either have ethernet to your AP’s or at the very least, dedicated wireless bridges to get to your ap’s.
  5. block bit torrent.
  6. do a wifi scan with a laptop in every nook and cranny.
  7. consider having a minimum signal strength set so bad clients cant drag you down.
  8. work with your neighbors to avoid interfering with one another. set signal strength only as high as you need.