+ Reply to Thread
Page 1 of 2 12 LastLast
Results 1 to 10 of 13
  1. #1
    Join Date
    Nov 2004
    Location
    Cork, Ireland
    Posts
    242
    Rep Power
    10

    Default poor CPE connections & high packet loss

    We have recently started experiencing problems with our CPE's connecting to our AP's &/or TX/RX issues.


    We have been looking through our network trying to find the solution without any luck therefore we are asking for some help.


    Our Access Points are all using StarOS, 90% is using v3 1.1.4 and the other 10% is using v2.11.0

    Our access points are using a mix of Wrap boards, 2 port WAR's & 4 port WAR's

    The are all using CM9 radios, and it is 100% running at 5GHz

    They are all using POE, with some using power direct from 12v battery, and some from mains (protected by UPS)

    They are all housed in Plastic boxes



    All the CPE's are OSBridge 5Xli or OSBridge 5GXLi using Atheros chipset AR5006XS




    The problem we are seeing is after we upgraded to v3 1.1.4 (this may not be the cause) the CPE's have started to suffer from poor download speed, and started to have alot of packet loss (increased to about 15% from about 1%)

    Also we are seeing some CPE's not connecting in at all or having 99% packet loss


    We have tried to upgrade some of the AP's to V3 1.1.5b (Beta5) and this doesn't fix the problem, but it did improve the packet loss for some CPE's


    Could that we are using plastic cases be causing the problems (are some of the CPE's connecting directly to the Radio instead of the Antenna, and causing problems for both themselves and the CPE's connecting to the antenna?)

    Could it be that the noise floor has increased overnight from giving all our CPE's a snr of between 7 to 20 to having some of them drop off and others not loose snr but to start having packet loss?

    Could it be that we have reached the maximum number of clients on our AP's (about 25 per radio with never more than 2 radios being used for clients on each board)?

    Could it be that the volume traffic throughput is too high for the hardware (consistantly about 4mbit/s)?


    We have tried to use 10mhz channels (the OSBridge firmware seems to work with the StarOS cloaking?) and this didn't fix the problem.

    We have also noticed on our backbone (StarOS V3 to StarOS V3) that if we do a Starutil -tx through the network the quality reading on one station can drop to 13 and then the link drops, but the signal on that link is -68 (SNR of 25)

    Can anyone give us some direction with these problems as they are crippling our network, and at the moment we have stopped connecting up new customers until we can get the network stable again (as we are spending all our time talking to annoyed customers.)


    Thanks
    Keith Nolan
    Last edited by knolan; 10-15-2006 at 05:08 AM.

  2. #2
    Join Date
    Mar 2004
    Location
    Nevada
    Posts
    359
    Rep Power
    11

    Default

    What are your customers distances (closest/furthest)?
    Are you routing or bridging?
    Using CBQ?
    How many total customers & APs, and what's your uplink (T1, DS3, Fiber, etc)?
    What's the CPU load on your APs(low/high)?

  3. #3
    Join Date
    Nov 2004
    Location
    Cork, Ireland
    Posts
    242
    Rep Power
    10

    Default

    The customers are between 300m & 6 miles (most are about 2 miles)

    The network is 100% routed

    We are using CBQ but only at the connections to our peering ISP's, and the problems are happening at the AP's where no CBQ scripts are implemented

    The network is small but growing, we have 21 Access Points (4 * 4 port WAR's & the rest a mix of 2 port Wraps & 2 port WARs) across 13 sites, The most clients we have on any WAR or Wrap board is 40, and the minimum would be 4

    The bandwidth to the internet is supplied via fiber, so it can grow to what ever we require, currently we have 5mb from 1 provider, we are looking at implementing BGP with a second provider within the next 3 weeks (again over fiber), and we have a 3mb DSL we using for backup (natting all our customers valid IP's behind 1 valid IP)

    The CPU load is low, usually about 3% to 20% depending on the traffic at the time. (Connection Tracking is off on all AP's - it is only on at our CBQ machines)

    Also InterBSS Relay is off on all AP's

    Our CBQ box is a P4 Server with 2gb ram, running StarOS V2.11.0


    Thanks
    Keith

  4. #4
    Join Date
    Apr 2005
    Location
    Conway, AR
    Posts
    548
    Rep Power
    10

    Default

    How many APs are being affected by this issue?

    What have you done about these CPEs?

    Also we are seeing some CPE's not connecting in at all or having 99% packet loss
    I had a recent experience where one bad link (89% packet loss) caused an AP wide RF 'lock down'. Basically, my associated clients would start dropping off one by one until everyone on the AP went to 'N', at that time TCP/IP flow halted to the clients.

    I was also running 1.1.4. ACLing the trouble client did not fix it. I had to go disconnect their link, and all my problems went away.

  5. #5
    Join Date
    Nov 2004
    Location
    Cork, Ireland
    Posts
    283
    Rep Power
    10

    Default

    I would like to add some more information to what Keith has already posted.

    What we are basicsally seeing is this from a CPE side. When I do a starutil from my laptop to the AP, the Rx reading is perfect. We have it limited to 6Mbps both on the cpe side and the WAR/WRAP side. where we set our connection speed to just 6mbps and I get a recieve of 500KB/s + consistently but maybe jumping around by 30 or 40 KB/s. Near perfect. Problem is when I go and do the Tx. This can be anything from 1KB/s to the magic number of 500KB/s!! If varies all over the place really.
    Also, it doesn't really seem to matter what the cpe's SNR is. Be the CPE at an snr of 10 or double that at 20 or 30, it is the same effect.

    The whole maddening thing about this is that litterally this has "happened over night". I cant pin down exactly when I 1st noticed this. And it is happening on all our AP's. I was trying to, and please lonnie etc dont take this wrongly!, "blame" the upgrade of our AP's at the time to 1.14 as the issue. I was trying to figure out was it from the time we upgraded to 1.14 that we started seeing the issue? So, we thought bumping it up to 1.15 beta4 might do the trick but it hasn't helped.

    I cant fathom that it could be an interference issue. All our AP's are many miles apart from each other so it cant be a local interference issue that has suddenly sprung up out of the blue "overnight" as such.
    It is not a compatibility issue with our CPE's as we have been using them now for nearly 9 months with out a bother until recently. I have also ruled out firmware revisions on our CPE's also as "older" CPE's running older firmware that were working fine, are now seeing the same issue.

    I also cant buy into that it could be that we dont have our WRAP/WAR's in a metal "faraday cage" box which is the one staros recommended rule we haven't adhered to!!!! As I said, it wasn't an issue until recently.

    So, I have been trying to figure out "what has changed since" and I have come up with the following.
    1. We have more customers than before
    2. We have more traffic going over the network than before.
    3. We are now running v 1.14.
    4. Weather. It is now getting a small bit cooler, and wetter!

    Any suggestions as to what may be causing our current problem? There has been a suggestion by a fellow countryman that changing the power setting of the card from "def" to 18 might do the trick?
    As in when at "def", if a closely positioned CPE connects to the AP, the card automatically drops it power down to service the request but at the same time then if a CPE far away makes a request, it has to up it's power to service him and hence we get this "yo yo" effect of the Tx speed as the CPE doesn't know what power setting either it has to talk to the AP at?

    Eoin

  6. #6
    Join Date
    Dec 2002
    Location
    Frozen Tundra of Canada
    Posts
    2,485
    Rep Power
    14

    Default

    Quote Originally Posted by eoinok View Post
    Problem is when I go and do the Tx. This can be anything from 1KB/s to the magic number of 500KB/s!! If varies all over the place really.
    Well, you could be hitting hidden node. In V3, there are no RTS/CTS settings, so if one client is doing a lot of transmitting, it can REALLY impact everyone else. If that client happens to be a 1MB connection, then the problem is dramatically worse. In V2, you could substantially help that with RTS/CTS and FRAG settings, but AFAIK, those havn't found their way back into the V3 driver yet.
    The whole maddening thing about this is that litterally this has "happened over night". I cant pin down exactly when I 1st noticed this. And it is happening on all our AP's.
    Those are the hard ones to track down. Perhaps you have some users Bit-Torrenting with lots of small packets, but customers on all AP's that the same time is unlikely. I'd suppect Hidden Note if it's mostly only on the Tx side. Good luck though.

  7. #7
    Join Date
    Apr 2005
    Location
    Conway, AR
    Posts
    548
    Rep Power
    10

    Default

    So, one of the last things you did was upgrade to 1.1.4? Have you tried reverting back to ver 1.1.1 and then run some tests?

  8. #8
    Join Date
    Nov 2004
    Location
    Cork, Ireland
    Posts
    242
    Rep Power
    10

    Default

    going back to v1.1.1 seems to give some improvement, there doesn't seem to be a v1.1.1 for the wrap? so I will do testing on v1.1.2 on our wrap boards.


    We will give an update after the testing.

    Thanks
    Keith

  9. #9
    Join Date
    Nov 2004
    Location
    Cork, Ireland
    Posts
    242
    Rep Power
    10

    Default

    All our tests so far seems to point to a problem with the OSBridge connecting to the newer versions of StarOS.

    Can I ask has something changed in the new firmware that can be identified to be causing issues with OSBridge CPE's connecting and is this something that can be fixed? - Or do we just have to live with using v1.1.1 of V3 if we want to use OSBridge CPE's

    I don't want to get into you should use the StarOS CPE, and this isn't our problems conversation, so I've just had a look to see could I find the price of the 5GHz WarTenna to see if there is a possibility of changing to Valemount provided CPE's - And I can't find any price at the store etc.

    Pricing the 2 port WAR with 1 radio, Pigtail, POE & Power Adapter @ $175.19 (for 10 + and shipping to the US) + $60.95 for the 5GHz RooTenna from Pacwireless give a total of $236.14 for a CPE.

    OSBridge price the 5G XLi CPE @ Euro 149 = about $187 (for 10+)


    We would usually purchase between 20 & 50 CPE's each month


    Regards,
    Keith Nolan
    Last edited by knolan; 10-16-2006 at 11:48 PM.

  10. #10
    Join Date
    Oct 2002
    Location
    Pattaya, Thailand
    Posts
    11,879
    Rep Power
    10

    Default

    Did you look at http://www.star-os.com/store and in the WarTenna section? Too bad you did not see that and the YSYL sale we just had. The price on 50 units is now $219.27 each with shipping to be calculated if you go wish to do this. The FedEx cost for high volume is not good so we use a freight forwarder for higher volumes and they go on a pallet. Contact me privately if this is of interest.
    Don't take life so seriously. It's not like you're going to get out alive.

    http://www.star-os.com/

    http://www.star-os.com/store

Similar Threads

  1. LAN Packet Loss
    By Premier in forum Support
    Replies: 20
    Last Post: 11-25-2008, 09:36 PM
  2. Packet Loss
    By Premier in forum Support
    Replies: 59
    Last Post: 06-07-2008, 12:16 AM
  3. 2% packet loss
    By Beebe in forum Support
    Replies: 4
    Last Post: 01-22-2008, 03:52 PM
  4. packet loss
    By rickg in forum StarOS™
    Replies: 5
    Last Post: 12-15-2007, 10:31 AM
  5. High packet loss on eth port
    By h3ll in forum Support
    Replies: 10
    Last Post: 03-22-2004, 04:06 PM

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts