Major Throughput Problems

Help Me,

We have hooked up a school district to out system and they are having major problems getting any kind of good performance. Most of the schools are on a 2.4 AP with 1500+ RSSIs and 0 to 3 Jitters. The SMs are not re-registering and other customers on the same AP are working well. The schools setup are as follows:

SM --> Cisco 1605 Router --> Swtich --> LAN

Canopy Tech suggested that I lock the Link speed and duplex to 10 Half for the ethernet to Cisco connection. This seemed to help a little but the performance is so chuncky that it seems worse than a dial up to the end user.

The schools are coming off of 384 and 256 frames and we can seem to make our Canopy stuff run better than the frames did in fact its working worse.

The IT guys for the district hooked the SMs to the eth1 port on the routers. So the eth0 goes to the LAN and eth1 goes to the SM. Can this work? A couple of the schools have serveral hundred windows machines. Are we flooding the Canopy network with alot of pakcets that dont belong there?

The entire school network is on 10 dot private IPs.

Any help would be appreciated.

Jon :idea:

Not sure about your throughput issue, but I would check for interference even though your SM is reporting high RSSI’s and low jitters. I’ve seen it before.

As far as the Canopy network being flooded with packets from the school’s LAN, it shouldn’t be, because you are terminating the SM into a router. Layer-3 devices will not pass broadcast traffic from the LAN to the WAN.

If IT has the SM terminated into eth1, and eth0 is the interface that goes to a switch(es) where the computers reside, then all you need in the Cisco is a default static route for 0.0.0.0/0 eth1. For example if the LAN computers were on a 10.0.0.0/24 subnet, the IP address for eth0 is most likely 10.0.0.1 - so the LAN computers default gateway is 10.0.0.1. When a packet arrives at 10.0.0.1 and its destination IP is intended for the WAN, the routing table will route the packet out eth1 (the SM) to the internet. A broadcast packet generated from the LAN will not be forwarded out eth1 to the rest of your Canopy network.

Are you providing the school with a public routable IP address for eth1?

I’m guessing that the Cisco router they are using is capable of doing NAT since you said they are on a Net-10 network.

The trick is that the schools are sending out 10. addresses. They want to be able to have the whole system reachable from within. The schools get NATed via a PIX box at our uplink site which then goes to copper down to our NOC. This is not the way we wanted to do this. Our orginal plan was to NAT to pulbic IPs at each router. The IT for the district insistes that we set it up this way.

I “think” I am following you on this one. Please correct me if I am wrong. So the school district obviously has multiple locations, and you have an SM at each location. All the districts aggregate at your uplink site where one single Cisco PIX box resides, and then you have a copper link from the PIX to your NOC, then out to the world.

What they want is that a PC at, say the high school, should be able to share resources with a PC at, say a junior high. But the way it is setup now is that each school has its own router. Is this correct?

I agree that your company’s intended setup would be best, but it’s always what the customer wants, right? I guess what you could try to do would be eliminating the Ciscos at each of the schools and just terminate straight into their Ethernet switches. You would be creating a huge Layer-2 network, and all of the schools’ PCs would still be behind the PIX if they are addressed correctly. The only downside is that each school’s default gateway (the PIX) would be 20-25ms away (if using non-advantage products).

I have tested this setup before because a potential customer of ours wanted something similar. It works, and I actually got about 1.3Mbps throughput on our T1 pipe when I emulated the situation. Since it would be a huge Layer-2 network, any type of broadcast traffic generated from the school would be seen by all your other customers, unless you use the VLAN feature of the new 6.1 firmware.

Am I anywhere close to reproducing your situation?

Missed the target – Hit the Tree!

You’re really close on our topography. We did find out that a Cisco 1900 switch at the uplink site was linking at 10 full while the both of the BHM’s at the uplink site said they were connected at 10 Half! Error City! We are going to upgrade that switch to a new Cisco 2950 and get much better controlls in place with IOS 12.1

The Schools IT is working on getting the 1605s set up right so we are not broadcasting everything to the whole network. We are also considering changing the PIX to a edge router instead.

Keep the input coming, I’m learning alot on this one!

First: Find out what exactly the 1605 is doing. Is it acting as a bridge or router and how?

Second: Microsoft broadcast traffic for a layer two network this size will kill everything. You can filter it out at the SM. This should be done regardless.

Third: Turn on Bit error broadcasting at the AP’s and test at all SM’s. Then check signal and jitter at the AP side by looking at the sessions page and verify canopy radio link tests are all 98% or better. Document those that are not.

Fouth: Take the school offline and test with no load.

Fifth: Sniff using Ethereal or simular tools to look for other problems.

Sixth: Use the 1605 or simular to create a routed backbone so that the broadcast domain is not so large.

Seventh: Smile and be happy!!! :smiley:

Thanks so much Bill for helping us out on this one! The system is much more stable and data rates are what they should be! I also glad to know that the problem was squarley in the School District’s equipment and not our Canopy Network. Thanks Again.