2 Level Condo with Concrete: WiFi Signal Drop v/s Smooth Handoff Handover

Discussion in 'Networking Issues' started by crashnburn, Jun 4, 2008.

  1. crashnburn

    crashnburn Network Guru Member

    2 Level Condo with Concrete: WiFi Signal Drop v/s Smooth Handoff Handover:

    So we have a 2 level condo made of heavy concrete.

    I have a Wireless Router (currently NetGear to be replaced) near the centre of the Bottom Level where all Wired Connections come from various places in the Condo.

    Obviously the signal drops and becomes weak at the Top Level. So I am thinking of adding another access point / router at the centre of top Level to ensure Strong signals there.

    I am trying to figure out how to ensure smooth handoff / handover between multiple WAPs / Router as if it were ONE single Wireless Network.

    I understand that WDS is used for Hardwire > Wireless > Wireless.

    Is there any way that I can have a UNIFIED "Single Name" Wireless Network that hands off the Wireless from a laptop or any such portable device to the STRONGER signal between the two Wireless Access points - WHERE each of them have their HARDWIRE pipes back to the center of the network.

    I am guessing the IP address assigned to the mobile device remains same and both the Wireless Access Points use a common underlying Wired Network + DHCP etc.

    What Router + Firmware + WAP Device + Networking Device combinations could help make this happen?
  2. HennieM

    HennieM Network Guru Member

    Yeah crash, as you mentioned on the WDS post, the http://www.polarcloud.com/tomatofaq#how_do_i_use_wds will help your issue if you want to stay wireless on the top floor. Clients will transfer from the one AP/router to the other, based on signal strength, just about without you noticing.

    Lots of post on this site w.r.t. WDS - "search" is your friend.

    If you want to wire to the top floor, you just put up another AP with
    the same SSID as the bottom one
    the same security settings
    on a different channel (space at least 3 channels)

    client handover will be smooth if you use WPA. If your use WPA2 it will be even smoother.
  3. crashnburn

    crashnburn Network Guru Member

    Do I have to make any particular settings differently or follow the WDS standard guidelines I see in other threads.

    THE THING IS >>> I am seeing that the 2nd or 3rd WAPs that people seem to be using leverage a Wireless Repeater / Bridge via the 1st WAP.

    I am wondering if the same will allow me to have the 2nd WAP SKIP the 1st WAP and go via its hardwire.
  4. crashnburn

    crashnburn Network Guru Member

    The question is would the second Router leverage the connected Hardwire (CAT6) or would it HOP OVER using the 1st Router
  5. HennieM

    HennieM Network Guru Member

    Perhaps we are not quite undertanding each other. These are your 2 options:

    Option 1: WDS, all APs on same channel, and all wireless security and parameters identical
    wiredPCs-and-stuff--wired--WDS+AP1- - -wireless- - -WDS+AP2
    Wireless clients connect to either WDS+AP1 or WDS+AP2.
    When walking between reception area for WDS+AP1 to reception area of WDS+AP2, wireless clients will smoothly connect to the closest WDS+AP.
    A client (wired or wireless) connected to WDS+AP1 can get to wiredPCs-and-stuff by going through WDS+AP1.
    A client (wired or wireless) connected to WDS+AP2 can get to wiredPCs-and-stuff ONLY by going through WDS+AP2, and then through the wireLESS connection between WDS+AP2 and WDS+AP1.

    Option 2: APs, on different channels (like 1 and 6), and all other wireless security and parameters identical
    Wireless clients connect to either AP1 or AP2.
    When walking between reception area for AP1 to reception area of AP2, wireless clients will smoothly connect to the closest AP.
    A client (wired or wireless) connected to AP2 can get to wiredPCs-and-stuff ONLY by going through AP2, and then through the WIRED connection between AP2 and AP1.

    Option 2 is your best option.

    If you want to mix the 2 options, i.e. have option 2, but the APs are set to WDS+AP, you can do, but that would be wasteful.
  6. crashnburn

    crashnburn Network Guru Member

    I really appreciate your detailed answer. But as i read it I am still a little unclear on a certain question.

    I come from a strong networking background (Ad Hoc Networks etc) but I havent kept up to speed on standards so the curiosity.

    This is the actual situation.

    If you notice, both the WL-APs, WR1 & WR2 are connected to the central Gigabit Switch via a HARDLINK. The Wireless signal between WR1 & WR2 will be WEAK ofcourse. The reason for having WR2 is because WR1 signal does not go through the concrete as well.


    - Laptop (Wireless) WR1 (Wired) Central Switch

    Single Wireless Link / Hop.

    1) Lets say you were going from WR1 to WR2 on the Wireless network (seamless handoff like a cellular phone changing cells).

    2) Now there are 2 ways to ROUTE packets from the Wireless device to the Central Switch (as well as the Internet which is only VIA the Central Switch).

    The question is does it use a Wireless Hop or Wired Hop after the transition?

    A) Laptop (Wireless) WR2 (Wireless) WR1 (Wired) Central Switch

    This one has multiple wireless hops. The path to the the CS is established with 2 Wireless links.

    B) Laptop (Wireless) WR2 (Wired) Central Switch

    This one has a SINGLE Wireless hop. The overhead of a second Wireless Hop is not present.

    This is particularly beneficial since we know the following:
    - WR1 (Wireless) WR2 - is a weak Wireless Link
    - WR1 or WR2 (Wired) to Central Switch - Wired Link is stronger / faster / reliable.

    Of course some areas the WR1 and WR2 signals will overlap. I am not sure if / how 802.1___ standards allow for selection of better signal in that scenario.

    So, does option 2 work the same as my Option B?

    How would it MIX? How would it be wasteful?
  7. HennieM

    HennieM Network Guru Member

    The bottom line is that infrastructure wireless = ethernet, just wireless. So think of the wireless signals as invisible cables being plugged into one AP or into another AP. Important: all talk here is for infrastructure mode - ad hoc not allowed... (and would not make any connection in this setup). Here some specifics:

    These 2 APs (WR1 and WR2) should be on different channels, like 1 and 6, and both APs should be in AP mode, so no wireless comms between WR1 and WR2.

    Your Laptop (if it has only one wireless card) can only speak on one channel at a time in infrastructure mode - see where I'm going here.... The invisible cable to WR1 is plugged out, and plugged into WR2.
    Note: This "invisible cable" is the 802.11 wireless hardware connection, not the IP connection. IP traffic does not notice this plug out/plug in save for the short "network delay".

    Even if the two WRs were on the same channel, your laptop's wireless driver, when set to infrastructure mode, will only connect to one SSID (i.e. one AP) at a time. When you have both WR1 and WR2 on the same channel, you'll have weak WRx-to-laptop signal, as the one WR's signal would show as noise on the other WR.

    This can happen only when you have WR1 and WR2 in AP+WDS mode:
    WR1 must have the MAC of WR2 as its WDS link, and vice versa.
    STP (Spanning Tree Protocol) must be on for both WRs, as you can have multiple ethernet paths or loops.
    The two APs (WR1 and WR2) should be on the same channel.
    (In a WDS enabled scenario, tricks are used to make the 2 APs' signals not interfere with one another).
    This is the "wasteful" scenario I mentioned.

    The WR1--wireless-WDS--WR2 link would have a much higher "path cost" than WR1--wired--GigaSwitch--wired--WR2. STP would sort out that the wired link will always be "the chosen one". The only time the wireless (WDS) link would work, would be when, for instance, the cable to one of the WRs is plugged out or there's no wired path for some reason.
    It's nearly exactly the same as having ethernet cables WR1--GigaSwitch, and WR2--GigaSwitch, and WR1--WR2. The only difference is, because the WR1--WR2 link is at 54Mbps half duplex, and not at 100Mbps (or Gb) full duplex, it will have lower priority (higher path cost).

    The "wastefulness" of this scenario comes in because your APs have to deal with 4 MAC addresses all the time (client, AP, WDS, destination), instead of only 3 (client, AP, destination). This adds a small overhead on all packets that travels wirelessly somewhere.
    Further, if your GigaSwitch packs up, or a cable to one of the WRs fails, you'll have more to worry about than maintaining that AP through WDS.
  8. crashnburn

    crashnburn Network Guru Member

    The original post you had mentioned 2 scenarios and a MIX as the 3rd scenario which would / could be wasteful.

    So from reading your last post I am understanding that with WDS enabled the network will choose the shortest / least cost path? Is that right? ( Yes / No? )

    But the overhead of maintaining and comparing between the multiple possible alternatives might be bad as well? (Y / N) ?

    Elaborate on the latter half that was underlined on the paragraph above.
  9. HennieM

    HennieM Network Guru Member

    // Oh! So you're saying that my changing the Channles and Disabling WDS we SEVERE that WR1 (WIRELESS) WR2 Link.

    - Correct

    // Ah! So even though the MAC / Physical Layer notices the minor change, the IP layer is abstracted? If I have it correctly. Maybe some packets / frames drop but its recovered...

    - Correct

    // Ah! So WRs on the same channel would actually increase Collison. Ta da! Its so long since I investigated wireless interference scenarios

    - Correct. Two seperate APs know nothing about each other, so the one Txs whenever it wants to, and vice versa. The same goes for the clients connected to those APs.

    // What kind of tricks? Distance / Signal attenuation? Power limiting?

    - No. In WDS mode the 2 APs know about each other, so they sync/share or do something to manage. Dunno exactly what - my guess is they just avoid collisions the same way a normal AP avoids collisions with multiple clients.

    //Is this path cost automatically determined or has to be setup / configured into the routers / WDS system so STP can use it?

    - No configuration needed, except if you want to force something. STP or the firmware or something figures it out. STP works by selecting/electing a main or controlling bridge, and all STPs talk to this bridge to figure out the paths and costs. Dunno exactly how.

    // I guess that would be a good BACKUP / Secondary Path. Would it be worth having that all the time given other negative factors? if any?
    // I am assuming this is the Negative Factor. Under what scenario (s) would this Factor be an issue.

    - It is (a good backup). Let's say your average packet is 1000 bytes. Add 6 bytes for the extra MAC + a little processing overhead to half duplex comms and your wireless speed is down by 1 to 4%. On a small packet, say 64 bytes, speed could be down by 10 to 40%. The overhead is actually on the FRAME (not packet) level, so it's always there, whether it's "wireless management talk" or "IP payload talk".

    Also, WDS with STP sends STP packets every 2 seconds or so - more overhead.

    // ...you'll have more to worry about than maintaining that AP through WDS.

    - Your whole net may be down due to short, or one of your Gb switches may be on the fritz, or the house may be burning down, or ,or,...
  10. crashnburn

    crashnburn Network Guru Member

    Back up path :)

    Thanks a lot Hennie. I really appreciate your effort and time to discuss this.


    Now, I have to try to find a decently priced Gigabit "smart" switch (between Unmanaged & Managed - I believe) to start off.

    PS: Is there a good KIT I could buy that has tools / testers for
    CAT6 "Socket" Installation & "RJ-45" Crimping AND
    testing them for CAT6 standards.

    The wiring has been done CAT6 and has been terminated but I just want to be able to check bottlenecks myself if they even happen.
  11. HennieM

    HennieM Network Guru Member

    As mentioned, I don't know how the root bridge is selected - I think it's like "I was here first". You could force it if needed, but in a 2-WDS setup, it would not matter. It would start to play a role when you have multiple paths of equal cost.

    You don't need any additional equipment. If you turn STP on in the routers, they will talk the bridge talk (I think it's 802.3D protocol. Not sure by a long shot though ;). It's a software protocol anyway, spoken by most managable switches.

    On Tomato it's a question of (re)typing the respective MAC addresses to turn WDS on. On DD-WRT you can turn the WDS on or off without losing the MAC address.
    However, if you lose a link to one of the routers, how will you get to that router to turn the WDS on? If you want WDS, leave it on. Best to experiment a bit perhaps....

    Good kits: Dunno about that, just that it's pricey.... ;)

    If your wired net is Gigabit - ANY gigabit switch - the A/B/G wireless stuff will NEVER, by wireless only, create a bottleneck on the wired net. Even if you have just 100Mbps switches, it still won't. Wireless gives you about 28 Mbps at best. This is becasue (i) it's 54Mbps to start off with, and (ii) wireless is half duplex, and (iii) wireless has LOTS of overhead (ACKs, etc. on the PHY level). Even if 2 or more different clients are using both your wireless APs full blast the wireless will never consume more than about 2 x 28 Mbps on your wired net.
    On the wireless side however, the bottleneck is inherent: 2 clients on the same AP shares the ca. 28 Mbps (so 14 each with both at full blast), 3 gets about 9Mbps each, etc.

    If you want to monitor anyway, best would probably be to get a switch which allows this. The switches in home routers don't have such fancies to the best of my knowledge.

    If you are looking at draft-N wireless APs, you could get 270Mbps (at this time) from the wireless. It's still half duplex, but multiple channels could be used, so the effective throughput from draft-N client to draft-N AP could probably reach 150Mbps. I doubt that it would though - real world thruput is less than 100Mbps to the best of my knowledge.

    Compare that to wired ethernet, where it's possible to get 90%+ thruput, i.e. 90Mbps on a 100Mbps net.
  1. This site uses cookies to help personalise content, tailor your experience and to keep you logged in if you register.
    By continuing to use this site, you are consenting to our use of cookies.
    Dismiss Notice