Continuation of Belkin router bridging topology conundrum....

>

> > I probably should have looked at the drawing before > posting. It's NOT > > a star topology. It looks like this: > > > > All boxes are Belkin 54G wireless access points with > WDS enabled. > > > > [ Cable ] [Main #1] W [ #2 ] W [ #3 ] > > [ Modem ]====[router ]==D==[ WAP ]==D==[ WAP ] > > | S | S | > > | | | > > | | | > > | | | > > wired LAN [ PC #1 ] [ PC #2 ] [ PC #3 ] > > > > Speeds 2933 580 640 > > kbits/sec kbits/sec kbits/sec > > > > > > Ok, this is NOT a star. It's a bus. The first > directly connected > > router will get the full speed of the > system. That's > working. > > > > #2 WAP is acting as a wirless client via > WDS. There's > no store and > > forward involved. It should be getting the full > 2900kbit/sec. It's > > not. Something is wrong. > > > > #3 WAP is acting like a wireless client via WDS from > WAP #2 which is > > acting as a store and forward WDS repeater. There > should be a 50% > > drop in bandwidth to WAP #3. Therefore it should be > getting > > 1500Kbits/sec. Something (else) is wrong. > > > > > Jeff, > > I'm going to try fully populating those MAC addresses > into the bridging > tables, and then tell you what happens. > > Once again, thanks! > > -- > bill evans >

Jeff or anyone else who was reading that "Wireless Access point" thread:

As previously discussed, all units are Belkin 54g routers, F5D7230-4's to be precise.

I tried changing the bridging tables in the main router, or "Main #1" in Jeff's ASCII representation , and in #3. I added the MAC address for unit #1 to unit #3's bridging table, and vice-versa for unit #1, adding #3's MAC address to it.

On the last bandwidth test using the CNet bandwidth test, I got 1652.5 Kbps,

314.8 Kbps, and 688.8 Kbps for the three units respectively.

After I added the MAC entries to the proper bridge tables, my numbers changed to 1764.7 Kbps, did not complete, and 188.1 Kbps for the three units respectively. I had to retry the test multiple times to get units #1 and #3 to complete the test. I also received many various errors, lockups, and whatnot.

I have since changed the tables back, and the everything is back to the way it was.

Reply to
Bill Evans
Loading thread data ...

F5D7230-4's to be

address for

1652.5 Kbps,

numbers

Is this what you have configured?

AP1 has AP2s Mac AP2 has AP1 and AP3s Mac AP3 has AP2 and AP4s Mac AP4 has AP3 and AP5s Mac AP 5 has AP4s Mac

Reply to
Airhead

Reply to
Bill Evans

YES, at least eventually at the site where I'm installing them on Monday.

Right now I have three nodes setup here at the house. AP1 has AP2's MAC, AP2 has AP1's and AP3's MAC, and AP3 has AP2's MAC.

When I added the entry for AP3 to AP1, and AP1 to AP3, everything went to hell.

Reply to
Bill Evans

message

units

lockups,

So, is this working ok?

And you got rid of this, right

Reply to
Airhead

OK, except the nodes other than AP1 don't get full bandwidth.

Jeff Lieberman, who was kind enough to help me and I can't tell him how much I appreciate his help, felt that I would get better results by changing the topology so every AP was talking to every AP.

Correct.

Reply to
Bill Evans

Time to divide and conquer. When things happen that are apparently weird, it's often best to tear things apart and build it up piece by piece. It's difficult to troubleshoot an entire system as a whole. Back to the basics.

You may wanna install a test computah at AP1 to act as a test server instead of using the cable modem. That will allow you to use IPerf, Netstat Live, or do simple timed large file copies. At 802.11g rates in a closed room, you should be getting about 22-25Mbits/sec thruput. If something is messed up, it will show up more obviously at the higher speed, than to rely on the cable performance remaining constant through the test. It also will make your benchmarking web site happier by not pounding it to death.

Cut the system back to just AP1 and AP2. Turn OFF everything else. Notice I said OFF, not disable. Remove all extra WDS configuration info and configure AP1 and AP2 to only talk to each other. Convince yourself that you're getting full thruput between your cable modem to a LAN connected computah on AP2. Do it in both directions to be sure. Go no furthur until this is working. As I recall, it was working but it's best to be sure.

Next, add in AP3. Setup WDS only to/from AP2, not AP1. Same with the configuration of AP1. AP1 should NOT communicate with AP3 even if would allegedly go faster. The reason is that all this WDS stuff is working at the bridging level which relies on the Spanning Tree Protocol to prevent loops. I'm not convinced that it's working with your Belkins and want to be sure that it's impossible to create a loop at this time.

Now test the performance again at AP2. The idea is to be sure that there is nothing that you've in AP3 to screw up the performance at AP2. Then, test the performance to a wired LAN connected PC at AP2. It should be half of what you previous obtained through the WDS link.

Now, try the performance between wired LAN computahs connected to AP2 and AP3. It should be full speed (22-25Mbits/sec).

Last, test the end to end performance from wired LAN connected computahs between AP1 and AP3. It should be half the full speed rate.

If you report your performance rates, please include the full units of measure (e.g Kbit/sec) and try to use the same units thruout (don't mix bits and bytes, kilo and mega, etc).

If these tests result in radically lower than normal performance numbers, then methinks something might be broken in your Belkin WDS implementation. I know it works with WRT54G routers, because we have a local person that uses such a multi-hop WDS system to obtain internet access to his remote location, which has no wired broadband service. I'm not sure which alternative firmware he's using, but I can ask.

I screwed up with suggesting a "mesh" type of WDS configuration. I hadn't thought about the potential problem with creating loops. Leave it so that WDS only talks to its immediate neighbor for now. The absolute worst performance drop would be to a wireless laptop at AP3, which would result in 1/4 the bandwidth. Methinks this is acceptable.

Also, I couldn't read the WMF file. Windoze complains that there's no program associated with the WMF file type. I have Office 2000 installed which should read it, but it's not configured. Also, some usenet servers drop attachments in text only newsgroups, so not everyone will see your attachments.

Gotta run. Good luck.

Reply to
Jeff Liebermann

point"

MAC

changing the

Jeffs a real trooper and not stingy with his knowledge. Just curious as to how you are checking the throughput on AP2 and AP3. Are you running from a wireless laptop?

Reply to
Airhead

Can you view either the WMF or DWG files attached? Those bandwidth numbers were my original numbers.

I have three computers here at the house for me to test my simulated network with, but no wifi NIC. About to order one, but haven't decided which one to get yet. Probably get a USB one for the versatility and ability to use the USB NIC as a poor man's cheap antenna via placement and possible use of USB extension cable, of which I have several. ;-)

Tested bandwidth with the CNET bandwidth test.

Reply to
Bill Evans

That's so strange, I thought Windows would view a WMF file itself, after all it's a Windows Meta File that the Windows clipboard itself uses to passes vectorized graphics through the clipboard.

On my machine, Paint Shop Pro is set to open all 83 or however many file formats it will understand.

Even though Windows doesn't have any default action assigned on your machine for an "open" or double-click on your machine, MS Word would likely let you see it with an "Insert/Picture" command but this really isn't important and is off topic and beyond the scope of this here memochatroom, but I digress.... ;-)

I'll implement your test in the morning, Jeff. Once again thanks.

Reply to
Bill Evans

Jeff,

Don't waste a lot of time working on this. Worst case scenario is that even if I can't get 100% of the throughput on the successive nodes after AP1, it'll work without lockups, errors, etc.; and hopefully be equally reliable and continue to work on site.

There's only one work computer at this site in that second building that will get the shaft on bandwidth, and the random walk ups that will attach to the network will only be there for lunch or dinner or at the bar.

I've reduced the test network to two AP's, and AP1 is consistently getting between 1600 and 1800 kilobits/sec, and AP2 is consistently getting between

500 and 600 kilobits/sec. I've tested it over ten times, each time doing a cold boot and immediately running the bandwidth tests.

I probably shouldn't mention this because it's not affecting the core wifi topology or testing on my end, but I've even inserting a 16 port wired hub/switch in between AP1 and my main computer here that I'm typing on now, the computer that's always been my test machine hanging off AP1. I've also got two other machines hanging off this switch (Dell PowerConnect 2016) and they're also consistently getting between 1600 and 1800 kilobits/sec. The only reason I added it was to test it and make sure it works, I found it in a drawer at this site and now that I know it works I'll use it on site to connect the existing eight wired computers and network printers in their main office.

Added AP3, and it gets between 1600 and 1800 kilobits/sec. Every time.

AP2 is still consistently getting between 500 and 600 kilobits/sec.

Copied 84.1 megabytes back and forth betwixt AP2 an AP3 with reboots in between. Each time took approx. 55 seconds.

And unless I made a rudimentary conversion error, that's around 1529 kilobits/sec.

????

Did I do my math wrong?

Copied 84.1 megabytes back and forth betwixt AP1 an AP3 with reboots in between. Each time took approx. FOUR MINUTES.

That's around ~400 kilobits/sec.

Strange. Not good.

I'll be looking at that problem tomorrow......

Can I use British units? I'd love to express an answer in stones/fortnight!

;-)

Sorry about mixing prefixes and bits/bytes on you. Being the one typing it, I hadn't considered that others reading it wouldn't know what I knew when I typed it.

I've been up since around 3 this morning, right now I'm loathe to even speculate on this problem. Copying between AP1 and AP3 is the only really bad problem.

I've already checked for any firmware updates. These Belkin boxes already had a recent (off the top of my head I think it was just a few months ago) firmware installed when I took them out of the box.

I seem to recall that the firmware update I'm thinking off added WDS functionality **ITSELF**, these boxes wouldn't do WDS without that last firmware update.

Once again thanks.

Reply to
Bill Evans

Cabling-Design.com Forums website is not affiliated with any of the manufacturers or service providers discussed here. All logos and trade names are the property of their respective owners.