Difference between Client Bridge and WET modes ?

Discussion in 'Tomato Firmware' started by menot23, Dec 21, 2006.

  1. menot23

    menot23 LI Guru Registered

    I've read the Wiki, and a few posts here which briefly touch on this topic, but I need a little more clarification before I set up my network one way or the other. I've programmed routers/switches for a long time, so I'm well aware of the fundamental networking concepts here. I just need a little clarification on the specifics of CB vs. WET as implemented in Tomato (as well as DD-WRT and Thibor).

    As far as I can tell, Client Bridge acts as a wired<->wireless non-transparent ethernet bridge for one device. The one device limitation coming from an overwrite of the source MAC address of devices attached on the wired side of things (much like a router would do).

    Things are a little less clear for me involving WET. The Wiki says that it's for attaching multiple wired clients to a wireless network. Fine. How exactly is it different than Client Bridge mode ? Does the MAC address over-write still occur ?

    It was my understanding that if you had one wired device you wanted to attach, Client Bridge was the way to go. If you had more than one device, you had to go with WDS to avoid the MAC address over-write situation. I'd like to avoid WDS, because it halves your bandwidth.

    So, what's the story here ?
  2. dvaskelis

    dvaskelis Network Guru Member

    There's a good wireless bridging article on SmallNetBuilder (formerly Tom's Networking), with a WET section along with a later WDS article.

    WET mode is true wireless-to-ethernet bridging. The wireless router in WET mode attaches to an AP, and an ethernet network, and bridges the two. The WET mode wireless router cannot act like an AP and does not allow wireless station (STA) clients to connect to it.
  3. digitalgeek

    digitalgeek Network Guru Member

    may be a drawing might help illustrate this...

  4. menot23

    menot23 LI Guru Registered

    OK, sounds good. My proposed config should work then. I've two Tomato boxes (Buffalo WHR-HP-G54 and Linksys WRT54Gv3.1), both running Tomato 1.01 :

    - Buffalo configured as "AP". Connects to cable modem, acts as sole DHCP server, and has one wired client and one wireless client hanging off of it.
    - Linksys as "WET". Has more than one wired client hanging off of it.

    All wired/wireless clients reside on same subnet and can communicate amongst each other, as well as the internet. If this is the case, I'm good to go.

    My main concern was that it was a hardware limitation as far as the MAC address over-writing in Client Bridge mode. If that's not the case, I'm all set.
  5. HennieM

    HennieM Network Guru Member

    I don't get it?? What's the difference between WET and client bridge mode in digitalgeek's drawing? Are you saying that in client bridge mode, only ONE of the LAN ethernet ports work? I find that hard to swallow, as the LAN ports are usually bridged together.

    IMO (and I have not seen Tomato) there's only 3 bridging modes: WDS, WET=client bridge, and client.

    Client Bridge = WET: same subnet wireless and wired, but only 1 wireless peer
    Client: different subnets, but still only 1 wireless peer.

    However bugmenot23, if you clarify, please let us know.
  6. GeeTek

    GeeTek Guest

    I just tested every mode and combination. Conclusion - Tomato REALLY ROCKS !! This post may be a bit long. I have some good data that may be useful in deciding which mode is best for what you want to do. My primary interests were bandwidth and client connectivity in the various modes. My test setup was with 2 computers and 2 radios with Tomato 1.01.

    Test 1 ; PC1 connected to Radio1 by LAN port. Radio1 Mode = AP. PC2 connected to R2. R2 mode = Wireless Client, same SSID as R1. Results = Radio of R2 becomes WAN. WAN needs to be in same subnet of R1. R2
    LAN is natted into a different subnet, providing 254 addresses to be used by multiple LAN devices. R2 provides no wireless connectivity. PC2 to PC1 bandwidth loops at 24 Mb using Ixia Q-Check.

    Test 2 ; Change R2 mode to Wireless Ethernet Bridge. WAN becomes disabled. LAN ports transparently bridged to R1 LAN ports. R2 Supports multiple computers. PC to PC BW loops at 28.5 Mb. R2 LAN IP can be on any subnet.

    Test 3 ; Changed mode of R1 and R2 to WDS using each other's MAC in the WDS table. No SSID is transmitted by either radio. No wireless client connectivity. PC to PC BW loops 29.5 Mb. This mode would best be suited for a dedicated point to point bridge. If using Lazy WDS mode, only 1 radio can be set for Lazy. (Both Lazy = no connection after 45 seconds. May need more verification).

    Test 4 ; R1 = AP+WDS with R2 MAC in table. R2 = WDS with R1 MAC. R1 supports wireless client. Wireless client (PC2) BW loop to PC1 28 Mb. R2 no SSID xmit, no wireless client connectivity. PC2 connected to LAN of R2 BW loop to PC1 = 28.5 Mb.

    Test 5 ; R1 and R2 set for AP+WDS. Both WDS tables have MAC of other radio. R2 is different SSID. Both radios transmit their own SSID and support wireless clients. PC2 connected to R2 LAN BW loop to PC1 = 28 Mb. Wireless client connected to R1 BW loop to PC1 = 28 Mb. Wireless client connected to R2 BW loop to PC1 = 14 Mb. There is the WDS "Half Bandwidth". In this configuration I changed R1 LAN connection to the internet switch. My online download speeds while using a wireless connection to R2 via WDS to R1 were EXACTLY the same as I get from R1 and from a hard line connection to the internet switch. WDS wireless repeating reduces BANDWIDTH by 50%. It does NOT cut the browser or download speed in half as some may think. As long as my internet download / browser speed does not exceed the available 14 Mb, there is no slowing of internet service.
  7. menot23

    menot23 LI Guru Registered

    Wow! Thanks for the very thorough topology testing, GeeTek! That Ixia gear kicks ass, doesn't it? :)

    Just a bit more clarification on the WET test, if you don't mind. How did you verify transparent bridging ? Examination of the PC and radio ARP tables, sniffing the wired or wireless traffic MAC addresses, or by some other means ?
  8. GeeTek

    GeeTek Guest

    Test 2, "wireless ethernet bridge" ? My understanding of "transparent" may not be quite corrrect. I simply connected 2 computers to the LAN of the bridged radio and verified that they could simultaneously ping the primary AP. Since the bridged radio does not do any routing (and also had it's LAN ip in a different subnet), I concluded that the connection was transparent. If you want to know for sure that both client MACs register with the host AP, I can test for that specifically. My radios are still set up. Let me know if that is what you are referring to, or clarify for sure what you mean by "transparent" and I'll be happy to run some more tests.
  9. u3gyxap

    u3gyxap Network Guru Member

    Yes, please confirm if the computers from the LAN side of the bridge are visible on the wireless side with their own MAC addresses. If they are visible all with the MAC address of the wireless interface of the router - then that is the same as Client Bridge.
  10. GeeTek

    GeeTek Guest

    Tomato 1, router only, no wireless, serving DHCP. Tomato 2, access point mode, LAN to LAN with Tomato 1. Tomato 3, wireless ethernet bridge to Tomato 2. Two computers connected to Tomato 3 LAN. With both computers set with static IP, Tomato 1 client list shows 2 MAC entries of Tomato 3 "wireless" MAC, but with the IP addresses of the 2 computers. Device name does not register the computer's name, it remains blank.. When the 2 computers pull an address by DHCP, Tomato 1 client list shows the individual MAC, the IP and Name of the 2 computers. Transparency appears to have a DHCP dependecy. Rebooted Tomato 1 before each half of the test.
  11. GeeTek

    GeeTek Guest

    Changed Tomato 2 and 3 each to WDS only with each other. Tomato 1 now shows both computer MACs when using static IP assignment, but not the computer name. DHCP obtain on the computers results in Tomato 1 showing MAC and Name of computer.
  12. menot23

    menot23 LI Guru Registered

    Couple of things going on here :

    1) The DHCP protocol specifies that the "client ID" is carried in the protocol header, where that client ID is typically the MAC address of the DHCP client in question. So, the MAC address used by the client for transport over the network really doesn't matter here. This accurately explains why your Tomato1 ARP cache only shows the Tomato3 wireless MAC address, yet the DHCP client table shows the individual PC MAC addresses.

    2) The DHCP protocol is transported over ethernet via broadcast (rather than unicast) frames. So, the MAC address used by the client doesn't really apply here either.

    So, what I think you've confirmed is that WET is behaving just as Client Bridge does. I suppose its a limitation/feature/bug of the hardware forwarding plane in the Broadcom chipset.
  13. menot23

    menot23 LI Guru Registered

    Yup. This conforms to what WDS is supposed to be doing.

    Unfortunately, that's a no-go item for me. The promise of a true transparent ethernet bridge operating over wireless was what I was after. WDS is a non-starter for me because it only serves to connect two wireless routers together, with no other wireless clients being able to attach to either router. I have other wireless devices that want to play with the rest of the network, so WDS obviously breaks this for me.
  14. GeeTek

    GeeTek Guest

    AP+WDS should accomodate you there should'nt it ? My rig is still set up, so if you want some more refined testing let me know what you are looking for. Both radios could be set for AP+WDS, they will bridge together, and both will support wireless clients as transparently as possible.
  15. menot23

    menot23 LI Guru Registered

    I suppose I could. I was just trying to avoid the whole WDS 1/2 bandwidth thing. Right now, I don't think that I'd be hitting that limitation on my network, so it's do-able.

    I'm pretty sure that I want to do some media streaming over my network in the not-so-distant future, and that 1/2 bandwidth bottleneck may present a problem with the application I have in mind.

    I'd rather set the network up once and let it rip, rather than have to do some kind of reconfig later on down the line.

    Thanks again for the REALLY helpful tests.
  16. GeeTek

    GeeTek Guest

    Certainly welcome. My testing was with ideal signal strength, the radios are only 7' apart. 14 Mb is the max you will get across a WDS connection. With the price of buffaloes around $30 on sale, you might consider using 2 radios back to back as a repeater with no WDS penalty.
  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