2x wifi on one router to maximize throughput - how to?
-
You can use the ALIX with two wired interfaces and two miniPCI slots
–> two AP's.But what do you mean 2x wired for double throughput? You cant just add a second link and have the double bandwith. --> You get a loop which in the best case gets shut down (STP) and in the worst case leads to packet storms.
Also if you're running two WLAN's DONT use the same SSID. It just leads to problems. (you'll see lot's of in/out error on the status page).
Also note that all channels are overlapping except channels 1, 6 and 11.If you provide two 802.11g WLAN-nets a single wired connection should be enough.
2x 54 Mbit = ~108 Mbit which is close enough to the 100 Mbit of wired ethernet.But what do i know ^^"
Maybe someone else has a better idea :) -
The mainproblem when running 2 wlan cards next to each other is that clients will see them both with the same signal quality so you might end up with all of them hopping to the one on the lower channel, the one on the higher channel or whatever the wlan-nic-vendor of the client thought is the best idea when seeing 2 times the same wlan with the same signal strength. If you use 2 different ssids you usually can set one as preferred over the other but that would need somemanually configuration logic (setting half of the clients to prefer ssid1 and the other half to prefer ssid2). I don't think be able to get what you want without manually configuring things like that. Also keep in mind that when using b/g you at least need 4 channels difference between the 2 cards to not have too many interferrence between the radios.
-
…Also keep in mind that when using b/g you at least need 4 channels difference between the 2 cards to not have too many interferrence between the radios.
Or change the polarity of the antennas ;)
-
Doesn't this bring only a difference of 3 dBi ?
(–> half the power of the signal) -
Doesn't this bring only a difference of 3 dBi ?
(–> half the power of the signal)Normally - Yes - it is about 3dbi… but there is a bigger issue here...
If you run the radios on the same channel then you will effectively reduce the total throughput as only one radio can transmit at a time on the same channel! Compound that with the fact the client radios will also be wanting to use the "air-space" so the total throughput will be even less...
The better way to go would be to run the radios on different channels, with at least a three channel separation to reduce crosstalk interference since the radios are so close - and use "stacked" antennas (one over top of the other separated by at least one wavelength (about 12-cm or so) to reduce crosstalk as well. To keep all the clients from "selecting" one radio over the other you may want to consider using two different SSID names and telling half the clients to use one and half the clients to use the other - that will insure they are spread across the radios evenly.
As for OLSR - you could use it to "link" to remote repeaters then have the clients link to the repeaters - but now you are getting really complicated! :)
As for running two NIC wired interfaces - probably not going to gain much, if anything here as the radios will only give you about 18-mbit to 22-mbit effective throughput each. I have yet to see a radio's throughput match the advertising hype (grin). If you are running 10-mbit NIC wired interface cards then using both will make a difference but most NIC cards today are at least 100-mbits so will handle two radio channels without any problems.
gm...
-
A three channel separation is kind of the minimum one would need. The center frequency (carrier) of that channel isn't affected by the other channel's sideband then.
DSSS CHANNEL FREQUENCY OVERLAP CHARTThe only true parallel service is with channels 1, 7 and 13, at least where regionally allowed:
| Channel Number | Frequency GHz | North America | Europe | Spain | France | Japan |
| 1 | 2.412 | X | X | | | |
| 2 | 2.417 | X | X | | | |
| 3 | 2.422 | X | X | | | |
| 4 | 2.427 | X | X | | | |
| 5 | 2.432 | X | X | | | |
| 6 | 2.437 | X | X | | | |
| 7 | 2.442 | X | X | | | |
| 8 | 2.447 | X | X | | | |
| 9 | 2.452 | X | X | | | |
| 10 | 2.457 | X | X | X | X | |
| 11 | 2.462 | X | X | X | X | |
| 12 | 2.467 | | X | | X | |
| 13 | 2.472 | | X | | X | |
| 14 | 2.483 | | | | | X | -
Hopefully I was not mis-understood when I stated "a three channel" separation - by that I mean if you are using channel 3 for one radio the next channel minimum to use would be channel 7 (you don't count the channel you are using in that three channel separation count… :) )...
gm...
-
this is obvious - you would need channel seperation, so lets move on to trying to solve the problem of client saturation - having 2 classrooms with 30 laptops each causes quite a load, even with 1 laptop the advertised "hyp" of these cards is worthless - I can only get about 25mbps of actual data throughput, if a 2nd client starts any traffic this gets cut in half- so the wirespeed 108 is a joke - I am guessing I can minimize this by forcing a 1 or 2mb negotiation with my clients, that way everyone will have at least some amount of garunteed speed.
The biggest thing I am trying to figure out is where this bottleneck is and how to alleviate it. -is it my cpu (500mhz alix) - I have heard for faster throughput pf needs faster systems. is it the PCI bus on the alix? Is it the bridge between each port (combo of pf rules and hardware). If pf is only able to pass traffic on a bridge at x speed I was hoping to double x by doubling the bridge, but this type of bridge doesnt seem to build right and routed doesnt either.
So it seems a sure thing that I can run 2 different SSIDs and bridge them to a single LAN interface - but is this going to help? - at this point we seem to have a solution, but I am looking back to the cause that this solution is addressing.
-
Please read up on wlan speed. 108 mbit/s or 54 mbit/s are theoretical values under optimal settings and even then it's the bandwidth of the layer1 but not of the real tcp/ip traffic. What you are seeing (25 mbit/s) is more or less what you can expect from such a link. It's not limited by the speed of the alix or freebsd's bridge implementation. Search for benchmarks of other wifi gear and you'll see that they perform similiar (unless you run 802.11n). It's just the way it is when using 802.11a/b/g and nothing pfSense specific. Also the more clients you have the more collisions will happen which means data has to be retransmitted. The "air" is a shared media. Actually 25 mbit/s is not too bad.
-
25mb for a single client, with 2 clients it drops down to around 5-6mb, by the time 20-30 are simultaneously transmitting - not too good. I dont expect high speeds, just need to keep people from complaining, there are a couple of small apps they access on a local web server, and some misc web browsing for the rest. I use untangle content filter so internet bandwidth is regulated pretty well, just need to provide decent access to it for laptops - so if I can provide 1mb throughput to 20-30 clients, this would be great.
In this particular scenario I was just attempting to save on hardware cost by running 2 radios out of a single box, and then having a third on the far side of the building to utilize all 3 open channels. It is a large school that has 2 laptop labs that are about 1500 feet apart, I have 2 more APs in the other lab as well. But theres a lot of activity, a lot of teachers and visitors bring laptops wanting wifi - I dont want to complicate things with a lot of different SSIDs, but that is looking like the answer.
-
25mb for a single client, with 2 clients it drops down to around 5-6mb, by the time 20-30 are simultaneously transmitting - not too good. I dont expect high speeds, just need to keep people from complaining, there are a couple of small apps they access on a local web server, and some misc web browsing for the rest. I use untangle content filter so internet bandwidth is regulated pretty well, just need to provide decent access to it for laptops - so if I can provide 1mb throughput to 20-30 clients, this would be great.
My experience is real-world. I am the Network Administrator for a library system in Florida and we have implemented a Wireless Network for patrons and guests to use. I have setup three AP's in each library location (17 libraries) running on different channels with the transmit power settings setup to minimize overlap while still getting a good signal for wireless clients within range of the AP. Even with this configuration we normally do not have more than 12 - 14 people on the system at one time as the radio-collisions are such that the throughput drops to the point where the individual client may only get about 150 - 200 kbit/sec link speeds.
To "increase" the throughput would require running all the radios in a round-robin CTS/RTS scheme - that would relieve a good deal of the radio-collisions since each radio would require permission to send from the AP. What this does is reduce the collisions and subsequent re-transmissions of the radio packets thereby increasing the overall throughput. The round-robin scheme is the same as used by the old Orinoco Karlnet setup - of course that is proprietary in nature so you would have to write your own "system" to control the radios.
In my configuration it is not cost effective (nor do I wish to take on such a task) since a patron or guest machine would have to be configured to run in this mode and most barely understand how to turn on their laptops (remarkable how many people have laptops but don't understand how to use them) and software would have to be loaded into the laptops to implement the radio control.
If you are looking to cover 25 - 30 laptops with wireless connectivity at 1-mbit/sec throughput speeds then you will need a good number of APs running low power so as to keep the total number of laptops serviced by each AP to around 4 - 5 … Just the nature of 802.11.
Hope this helps...
gm...
-
There is a user from this forum that has 2 boxes running @5 ghz (cm9) cards 700 m away he gets about 6 Megabytes/sec.
I have seen a benchmark with 2 ALIX 3C1 with Staros @ 5ghz(cm9) doing 9.4 Megabytes/sec (Atheros super A) 7km away.
If your looking 4 speed use 5ghz cards to do the backbone .
@ 2.4 ghz like the hoda said you get about 25 mbits (~3.5 Megabytes/sec)
