Harlquinth (18th January 2011)
Your only going to get 300Mb of throughput if both your clients and AP fully supports it.
In addition your clients still share this 300mb trunk (if it were attainable) and will be queuing up for their turn to talk to the AP.
If the rouge AP is better than -80db in the vacinity of your laptops there is every chance they will hear it and this will compound the issue further with clients having to discard and retry all the time.
The Ruckus, well smart it may be but, its only going to be able to reject the co-channel interference on Channel 1 thus making the channel unusable anyway so your theoretical bandwidth is going to be significantly lower than you expect anyway.
In all honesty you might get better results running points with dual G radios and load balancing rather trying to get everyone bashing the same point with N...
The spectrums a mess and its getting worse everyday, 2.4ghz is like the M25 on Monday,Friday and Bank holiday rolled into one.
Whilst I was at BETT I did a sweep with the etherscope as I was desperate to try and do a bit of remote support to a client.
The spectrum was swamped and for anything more than a quick email check or web page it was next to useless.
I could see all makes and models of wifi fighting for control of their bit of the airwaves, by far the best performing Access Points were ARUBA!
I have never owned one or sold one, heck I have never even seen one but at BETT they were outperforming everything else and the scope never lies....
Harlquinth (18th January 2011)
We did originally have the laptops set to use a wireless-G network on the Ruckus point, but we changed it to use N last time we were trying to solve it.
Would having a second Ruckus point in the room make any difference? May try borrowing one from another part of the school and see if it does.
Anything can be a source of interference, usually Microwaves always get the blame but thats not always the rule.
A strong digital PCM is pretty robust but there are loads of things that mess with it.
Freeview digital can be wrecked by low frequency stuff, my son lives with a Pylon and a 40Kv power line directly between him and Crystal Palace the only solution was to upgrade the antenna and point it to Bluebell Hill in the other direction!
Cellular stuff runs between 850-950 Mhz so there is no direct harmonic to interfere with the 2.4Ghz band but that doesnt stop it from mixing with another source and generating another cloud of interference.
So whilst I wouldn't think the cells themselves to cause an issue there is always the 5Ghz backhaul links that are on the same mast!
I recently found in St Albans a whole mesh of Wifi Bridges that have been connected to remote Pay and Display machines linking hem back to the Council Offices along with CCTV.
As for N, it's certainly not going to be the solution to everything. - Taken from Wikipedia
To achieve maximum output a pure 802.11n 5 GHz network is recommended. The 5 GHz band has substantial capacity due to many non-overlapping radio channels and less radio interference as compared to the 2.4 GHz band.
An 802.11n-only network may be impractical for many users because they need to support legacy equipment that still is 802.11b/g only. Consequently, it may be more practical in the short term to operate a mixed 802.11b/g/n network until 802.11n hardware becomes more prevalent.
In a mixed-mode system, an optimal solution would be to use a dual-radio access point and place the 802.11b/g traffic on the 2.4 GHz radio and the 802.11n traffic on the 5 GHz radio.
This setup assumes that all the 802.11n clients are 5 GHz capable, which isn't a requirement of the standard.
40 MHz in 2.4 GHz
The 2.4 GHz ISM band is fairly congested. With 802.11n, there is the option to double the bandwidth per channel to 40 MHz which results in slightly more than double the data rate. However, when in 2.4 GHz enabling this option takes up to 82% of the unlicensed band, which in many areas may prove to be unfeasible.
The specification calls for requiring one primary 20 MHz channel as well as a secondary adjacent channel spaced ±20 MHz away. The primary channel is used for communications with clients incapable of 40 MHz mode. When in 40 MHz mode the center frequency is actually the mean of the primary and secondary channels.
Therefore, are your clients a/b/g/n or just b/g/n?
I say this as I use a single Dual Band Dual Radio point to service 40 Journalists every other Saturday without a problem and I use that setup.
2.4Ghz b/g and 5Ghz a/n
Needless to say all the MACs jump on to the 5Ghz and all the legacy hops on the b/g.
Last edited by m25man; 18th January 2011 at 02:05 PM.
Harlquinth (19th January 2011)
Quick update: We found that the wireless cards on the laptops had a couple of power saving settings, so we set those settings to high power, which increased the logon time by about 50%. We then swapped the Ruckus point (which was an older model that didn't have 5Ghz) with a newer one from another part of the school that did have 5Ghz - that seemed to increase the logon speed by about 50% again when we tested it.
This is where it gets interesting - on one of our tests the laptops refused to login at all, then we noticed on inssider that the 5Ghz ssid's had dropped and vanished completely (which was causing the laptops to not connect). When checking the logs on the Ruckus controller we noticed that at exactly the same time the 5Ghz cut-out the was a strange log entry saying "AP[68:92:34:1e:97:c0] detects radar burst on radio [11a/n] and channel  goes into non-occupancy period".
If it is from the mobile mast is that normal, or could it mean that the mast is faulty?
Code:Date/Time Severity User Activities 2011/01/19 13:47:04 Low AP[68:92:34:1e:97:c0] detects radar burst on radio [11a/n] and channel  goes into non-occupancy period. 2011/01/19 13:47:04 Low AP[68:92:34:1e:97:c0] detects radar burst on radio [11a/n] and channel  goes into non-occupancy period. 2011/01/19 13:45:53 Low AP[68:92:34:1e:97:c0] detects radar burst on radio [11a/n] and channel  goes into non-occupancy period. 2011/01/19 13:45:53 Low AP[68:92:34:1e:97:c0] detects radar burst on radio [11a/n] and channel  goes into non-occupancy period. 2011/01/19 13:44:19 Low AP[68:92:34:1e:97:c0] detects radar burst on radio [11a/n] and channel  goes into non-occupancy period. 2011/01/19 13:44:19 Low AP[68:92:34:1e:97:c0] detects radar burst on radio [11a/n] and channel  goes into non-occupancy period. 2011/01/19 13:43:10 Low AP[68:92:34:1e:97:c0] detects radar burst on radio [11a/n] and channel  goes into non-occupancy period. 2011/01/19 13:43:10 Low AP[68:92:34:1e:97:c0] detects radar burst on radio [11a/n] and channel  goes into non-occupancy period. 2011/01/19 13:39:50 Low AP[68:92:34:1e:97:c0] detects radar burst on radio [11a/n] and channel  goes into non-occupancy period. 2011/01/19 13:39:50 Low AP[68:92:34:1e:97:c0] detects radar burst on radio [11a/n] and channel  goes into non-occupancy period. 2011/01/19 13:31:54 Low AP[68:92:34:1e:97:c0] detects radar burst on radio [11a/n] and channel  goes into non-occupancy period. 2011/01/19 13:31:54 Low AP[68:92:34:1e:97:c0] detects radar burst on radio [11a/n] and channel  goes into non-occupancy period. 2011/01/19 13:30:21 Low AP[68:92:34:1e:97:c0] detects radar burst on radio [11a/n] and channel  goes into non-occupancy period. 2011/01/19 13:30:21 Low AP[68:92:34:1e:97:c0] detects radar burst on radio [11a/n] and channel  goes into non-occupancy period.
That I wouldnt like to say without an RF survey but I have a 5ghz bridge on the rear of my house that connects to my workshop about 180 ft away.If it is from the mobile mast is that normal, or could it mean that the mast is faulty?
In order to keep the 150Mb link I run it on full legal but I can see my bridge SSID quite easily within beam width 3km away! I haven't measured it at a fill 1watt yet but I'm told it can reach 6km.
So yes Im sure that it's quite possible for you to get a Microwave hit.
BTW where are the school kitchens?
I do LOL when people talk about microwave ovens affecting their wifi networks!
Have you seen the size of those Hobart Commercial units!
Some of them could blackout GCHQ when in use...
The kitchens are quite nearby actually as well, just along the corridor. Don't know what sort of microwaves they use though.
Is there any software that can do some sort of RF survey, or would it have to be some sort of specialised hardware?
To be honest all of the free stuff uses normal hardware and focuses on Signal Strength when you really need to be measuring SNr and looking at the wider field of view with a spectral analyser.
There are a few "cheap" USB devices and even some of the manufacturers drivers and management software that can do a lot of stuff, but to make any real sense of the spectrum you need to spend some money.
These guys have 3 offerings
You would need to spend at least £1500 for the cheapest Fluke device to look at the problem or
AirMedic for about £900
I personally own the Etherscope II with all of the options which I have had for 3 years and couldnt work without it. It was the best £6k I ever spent.
Hi guys I used to install CCTV and Intruder systems and you may not be aware but the alarm detectors that have passive infra red and microwaves known as a dual tec can cause the exact problem that you are experiencing. We had a site that proved it to us using a spectrum analiser.
From observation if the detectors have more than one led i.e. green red and amber it most likely has a microware. You can cover the detector with silver foil to prove this but I wouldnt recommend it as it may have anti masking make sure you don't leave the foil over the detectors at night.
I too have a Ruckas wireless LAN across my school with 30 AP's installed by Nexgen Solutions. Worked great in year 1, had issues when I introduced Vista with pre-authentication problems. Basically the kids logged in and did not get their home drive (H. They had to logout and in again and it worked. Since last summer we upgraded everything to Windows 7 and also installed a Server Farm running RD Services to 50 10Zig thin clients. This uses multicasting to talk between the clients and servers. Since then we can have no more than 50 clients and our AP's drop off like flies. We have one default VLAN on all HP switches, no subnets anywhere. The server farm runs great but we have had two experts in, one a Cisco engineer who says if we turn the farm off the probelm goes. Not a lot of use that, because we have to use it. Yesterday we had a HP switch expert who said the core switch chassis was faulty but HP don't agree. My point is this: I am sure that the Ruckas Zone Director and AP's use multicasting and the only thing that really fails is the wireless. Nexgen say there is nothing wrong with it and I have done 5 firmware upgrades and am currently on 188.8.131.52 build 7.
The only way forward seems to be to either put my server farm and clients on a VLAN or use subnetting to isolate traffic. I have bought a Fluke Inline Series II Tester and noticed that when plugged into any switch when the multicast reading goes to 95% or more the wireless falls over. At 99% the wired network slows down.
I am desperate for any ideas at all because I cannot, dare not add more thin clients and on a system where with 30 AP's that should safely handle 50 clients we have never exceeded 55 clients on the entire network.
Your biggest enemy in any collision domain is broadcasts or unsolicited multicast traffic.
You control broadcasts by subnetting and routing and you control Multicasting by IGMP Snooping.
If you have a L3 core switch make sure this has IGMP snooping enabled and enable it on any peers that support it and then see if your Fluke shows any improvement in multicast traffic levels.
This is as good as any explaination I can give here.
I have seen school networks with a single collision domain and a 1000+ hosts with a bad unc path in a policy file that didnt exist!
It was taking 12 minutes to login until the non existent unc path was identified, once removed logins returned to less than 30 seconds.. (Traced using the same tool you have )
Just remember without vlans subnets and routing, a broadcast from any one of your wireless clients is going to pop up on every port on every switch in your entire lan.
30+ laptops all broadcasting their heads off because they havent been out of the trolley for a few days, is enough to bring the best kept single subnet to it's knees for several minutes whilst everyone finds out where everything is!
Then as soon as you do get associated, there will be the good old Windows and Adobe updates flooding your uplinks and gateways along with the roaming profiles and folder redirections with menus and shortcuts all trying to squeeze down a highly contended wifi slot that struggles to deliver 60mbps in near perfect lab conditions....
If you have your DNS setup correctly and all of your static devices correctly set and all of your shared resources, folders, printers etc online and available, broadcast traffic should easily be kept below 5% of total traffic.
As for multicasting unless your ghosting an entire room of PC's and your Ghost Server is at the end of several uplinks IGMP Snooping should keep things manageable.
Thank you so very much for taking the trouble to reply so comprehensively.
I am pleased you recommend subnetting and routing as the way to control the broadcast traffic. I was coming to the conclusion that this is my only hope to move forward. The engineer who said I need not bother was in conflict with the EMBC guidelines which recommended this. I will have to dig our my training course notes from 10 years ago now to plan a managed approach to this.
Interestingly my server farm with 50 10 ZIg clients adds 36% multicasts to the measured multicast % displayed on my Fluke In-line tester. The first Cisco engineer who came blamed this for causing the wireless to crash and yes the wireless did work when we disconnected the farm from the main network and connected it to a dummy one joined to a cisco switch which was not connected to the school network. However we did not try adding more wireless clients to see if the measured multicasts goes up as more clients are added. It should do in theory.
Thanks for the advice about paths, resouces printers etc because I do have an issue with my Buffalo NAS boxes which randomly loose connection so this is worth checking. This problem also got worse when we upgraded the two domain controllers to Windows Server 2008 r2 x64 last summer and I unfortunately lost my entire DHCP entries because the MS guidance did not warn me that this would be lost when I wiped the main DC. I had to restore the old DC from a backup, manually copy all the DHCP entries (I even took photo's of the screen!) because I discovered I could not restore the backup of the DHCP from Server 2003 and restore it to Server 2008 after rebuilding the DC.
I then had to manually recreate the entire DHCP from scratch - no mean feat. There may still be machine wrongly set etc. I will scrutinise my DNS setup again.
I had already turned IGMP on on all my HP Switches and I have noticed a drop in the average multicast broadcasts but it is still up to 50% before school and got up to 90% at one point in the day.
I have now started to log every MAC address on ervey port on every switch so I know what I am seeing when I go on the switches. One school in Lincolnshire even assigns a group of 5 MACs to each port so that should someone unplug and move a PC to another area they will not work! That way he stopped kids plugging their laptops in and playing Quake over the network.
We do use Ghost but soon learnt not to ghost a room without disconnecting the entire room and putting it on a switch and plugging a machine with the ghost cast server into that switch. The accounts department were not happy with the slow network otherwise.
Again many thanks and I will post back when I have made changes that improve things. Any other feedback would be welcome from anyone. Hope this helps others
Last edited by mtdant; 16th February 2011 at 05:22 PM. Reason: Bad spelling
In which case I would be seeking someone on your lan streaming something... or do you have cctv devices on this VLan?I had already turned IGMP on on all my HP Switches and I have noticed a drop in the average multicast broadcasts but it is still up to 50% before school and got up to 90% at one point in the day.
mtdant (17th February 2011)
We do not have any CCTV on the network, but do have an old networked webcam. I will investigate this. Interestingly I just measured the multicasts at 8:45 this morning when the thin clients were all off and it still reads 15% multicast.
mtdant (19th February 2011)
There are currently 1 users browsing this thread. (0 members and 1 guests)