Extricom - Sudden Poor Throughput.

[educontractor]

Hello all,

 

We have a mix of Meru (x2), Extricom (x2) and Ruckus (x12) managed wireless networks across our schools.

 

All have their pros and cons (this isn't the purpose of my thread) but we've experienced problems with one of our Extricom schools recently, related to poor connection speed and throughput.

 

The school (a single form entry Primary) have An Extricom EXSW-800G controller with 8 x EXRP30n access points. They have a 100Mbit Virgin Media leased line paired with a Watchguard XTM330 for firewall/filtering. Wired LAN is using a Cisco SG-200 series 52 port Gigabit Ethernet Switch.

 

The wireless system has been extremely reliable, and performance sufficient that nobody has complained about it since installation in 2012. This is no longer the case, however!

 

Despite there being no physical reconfiguration of the network, performance has become markedly slower and connection speeds to the WLAN have dropped too. Whereas previously it was common to see 100Mbit+ TX connection rates throughout the building, it's far more common now to see 36Mbit and below. Internet speed (as measured - imperfectly I admit - with online speed tests) registers ~10Mbit up/down via the WLAN, but wiring in at the same time sees ~85Mbit down and ~95Mbit up.

 

I've reconfigured the WLAN from scratch to the same specification and the exact same results can be seen. Placing a laptop close to an access point sees *connection* speeds reach 130Mbit (albeit after a minute or two) but throughput is still disappointing compared to a wired connection, never achieving more than 45 Mbit/s in either direction. This certainly didn't wasn't the case previously.

 

Unfortunately, Extricom are pretty much a closed book with regards to support in this country (one of the principle reasons we moved to installing Ruckus) so I can't obtain updated firmware for the controller / AP's and nobody responds to my emails for technical support from Extricom HQ in Israel.

 

Has anybody here got any advice?

 

The following are the software versions the controller is running:

 

[TABLE=width: 100%]

[TR]

[TD=class: GeneralText, width: 60]OctopusFS:[/TD]

[TD=class: GeneralText, width: 614]v4.5.09.10~fr_2011-Dec-28-1758[/TD]

[/TR]

[TR]

[TD=class: GeneralTextBold]Serial Number:[/TD]

[TD=class: GeneralText]xxxx[/TD]

[TD=class: GeneralText]AppsFS:[/TD]

[TD=class: GeneralText]v4.5.09.10~fr_2011-Dec-28-1758

[/TD]

[/TR]

[TR]

[TD=class: GeneralTextBold]Domain:[/TD]

[TD=class: GeneralText]xxxx[/TD]

[TD=class: GeneralText]Kernel: [/TD]

[TD=class: GeneralText]#1 Wed Oct 26 18:14:47 IST 2011[/TD]

[/TR]

[/TABLE]

 

Forum cliché alert: Thanks in Advance!

[Michael]

From the information you've given, I'd say the problem points to the controller itself.

 

The LAN sounds OK, as does testing the VM connection via Ethernet.

 

I presume you've checked the link speed between the Cisco switch and the controller? If there's a problem here, it would slow everything up. I presume also you've rebooted both the Cisco and the controller too?

[psydii]

I would guess there maybe some broadcast traffic (Apple TV?) flooding air-side. If you've got a MacBook Pro to hand, fire up wireshark in it and have a look. Otherwise attach wireshark to your core and watch all the VLANS filtering down to the ones presented by the extericom.

[educontractor]

Thanks for replying Michael.

 

Both switch and controller have been rebooted, and the switch is reporting the controller connecting at 1000Mbit/s FDX. The wireless controller's web interface loads snappily and doesn't exhibit any of the sluggishness I'd associate with a problem at the ethernet layer.

 

It's very subjective I realise, but activity on the WLAN (gauged by looking at activity LED's) looks very "broadcasty". It's impossible to sniff traffic AP-side with the extricom controller, however, and running Wireshark on the wired LAN doesn't show excessive broadcast traffic.

 

All very strange.

[educontractor]

Actually, embarrassed to say I hadn't though of doing it this way (Wireshark whilst attached to the WLAN) as I usually connect remotely and run Wireshark from a virtual machine on the wired network.

 

Trying that now.

[Michael]

Just a thought, I presume you're using WPA2-AES encryption? Can you monitor via the controller what devices are connected?

 

Seeing as you only have 8 APs, have you tried disconnecting each one in turn to see if the problem goes away?

[educontractor]

Nothing exciting to see there, unfortunately. I captured c. 50,000 packets and nothing stood out. A few malformed HTTP sessions, normal amounts of ARP broadcasts, few unanswered ICMP requests. No multicast traffic. My laptop was the biggest talker on the segment... all very "meh".

[educontractor]

Just a thought, I presume you're using WPA2-AES encryption? Can you monitor via the controller what devices are connected?

 

Seeing as you only have 8 APs, have you tried disconnecting each one in turn to see if the problem goes away?

 

Haven't done that I must confess. I can monitor connected devices, for encryption it's set to WPA/WPA2 with AES or TKIP. ASCII PSK for the password.

[psydii]

Extricom is SCA isn't it? Perhaps something has changed in your near environment and the channel it s on is no longer clean enough? What other wireless networks can you see? Has anyone installed a wireless CCTV camera recently?

[Michael]

Alarm systems can also create problems. Is 5GHz an option?

[psydii]

50000 packets... How long was the capture running, and what else was going on nearby? That's about 100 Mbytes of data by my reckoning once you factor in wifi overheads. That is quite a lot for just a short capture in a quiet segment.

[educontractor]

Apologies for letting this thread "fade away" - especially with so many helpful replies.

 

Still having issues with this, and it's very difficult to pinpoint why.

 

This evening, as semi-related exercise I stripped out all of the patch leads from the comms cabinet and recalled it from scratch. As part of this process, I reduced the network to its bares bones:

 

My laptop, the backbone switch (nothing too exciting, a 50 port Gigabit Cisco Small Business SG200), the wireless controller and one access point (the one nearest to me, about 30cm away), and the firewall and everything beyond it.

 

Wired into the switch, running speedtest.net I got 94Mbps down and 92.5Mbps up.

 

Running the same test wirelessly I got 17Mbps down and 26Mbps up.

 

Reset the controller to factory defaults, recreated the channel blankets (radio 1: n/g channel 6, radio 2: n/a channel 36, radio 3: rogue detection), ran the same test wirelessly and got 15Mbps down and 25 Mbps up (take this to be the same as previous accounting for vagaries of internet speed tests).

 

Recreated the channel blankets from scratch on channel 1 and channel 11 (my laptop just didn't seem to fancy connecting on 5Ghz for some reason) with the exact same results bar a couple of Mbps either way.

 

Each time my laptop (Retina MBP) reported itself connecting at 130Mbps.

 

Really got me stumped why the performance would be so poor. Why also would the upload throughput be almost 50% greater than the download?

 

Thank you very much for all of the input thus far, and apologies again for the long period without reporting back.

[Michael]

Looking at your original/starting question, connecting to my Virgin Media router wirelessly it generally connects at 130Mbps. Performing a speedtest it tops out at 36Mbps download, which is very comparable with your results.

 

Wireless AC 'can' achieve very comparable results to 100Mbps Ethernet, but again there are a lot of factors involved.

[educontractor]

Comparing apples with oranges, but we have several Ruckus systems in schools with otherwise identical WAN provision (same firewall, same 100Mbit leased lines) and I see 50+ Mbps internet performance when connected at a reported 78Mbps wireless transmit speed.

 

When the wireless connects at 130Mbps+ I see the full bandwidth of the connection.

 

It might well be that this is the best throughput this Extricom kit can manage, but if that's the case it's very disappointing.

 

Also, why would the upstream speed be significantly greater than the downstream bandwidth?

[Michael]

It could simply be that the Ruckus kit is MIMO and the Extricom kit isn't. The actual wireless adapters in devices makes a big difference too, but you already know this

[educontractor]

Thanks Michael.

 

You may be right and this is the best we can expect. I wish that I had some hard stats from when the system was installed back in 2012 but subjectively it "felt" faster. I certainly never received any complaints until earlier this year anyway.

 

The thing I keep returning to (like a broken record!) is the disparity between reported download and upload bandwidth. As each test is carried out sequentially, with nothing else using the network, surely each should be the same even if we accept that it will be a long way short of the expected "wire speed"?

 

It's almost as though there's some sort of bandwidth management, or ersatz QoS taking place on the wireless controller. There is a whole section on WMM/DiffServ but nothing has been changed there and when I did a factory reset the settings all remained the same, too.