Help: 300 mbit N network capped at 11 mpbs until streams reset?

[Morty_UA]

I have a fairly standard 2-channel, 2-radio setup:

• Netgear WNDR3700 release 2 access point
• On the 2.4 GHz bands, a 54 mbit wireless network with SSID "Home"
• On the 5 GHz bands, a 300 mbit wireless network with SSID "Home5Ghz"

My mobile phone, Wii etc use the "Home" WLAN. My desktop and laptop PCs use the "Home5GHz" WLAN.

Now, I frequently download large files from a Internet server to my desktop PC, using multiple parallel FTP streams. When I do this, I very often get capped at 11 mbit/s total download speed for all streams combined.

If I then reset the TCP connections a couple of times (by restarting the FTP program), then it suddenly jumps to ~48 mbit/s (which is fair given the ~50 mbit/s max speed of my Internet connection). I do not reboot the PC. Usually it's enough to wait for many minutes, or re-start the FTP program several times.

When the connection is capped at ~11 mbit/s it's an almost perfect flat line -- the speed is very even.

This is very annoying. Does anyone have a suggestion for what might be causing this? How do I troubleshoot?

The 300 mbit wireless is connected via a USB 2 port, and the WiFi link is within the same room, maybe 8 meters from AP to USB NIC. The Windows systray applet says the WiFi is connected, with 4 of 5 or 5 of 5 signal strenght, and 270 mbps line rate.

Heeelp please? This really bugs me!

 

[thiggins]

How are you measuring the 11 Mbps download speed?

 

[Morty_UA]

The 11 mbps are measured using the FTP program. It has a built-in bandwith monitor. When I'm running these downloads they are the only source of traffic on this network & internet connection.

 

[sfx2000]

Welcome to the world of deep packet inspection and bandwidth management from the operator space.

http://www.sandvine.com/

 

[tipstir]

USB NIC tend to overhead also tend to be slower than the Mini-PCI-E cards inside your system desktop or laptop WiFi. Speed will vary. I've stopped using these USB 2.0 NIC a fews years now. Most of them go DUFF (dead)

Use LAN SPEED TEST on that 300mb WIFI to another WiFI system. Set the mb to 150 to 270mb. Measure the reads and writes and then report back.

 

[thiggins]

Welcome to the world of deep packet inspection and bandwidth management from the operator space.

http://www.sandvine.com/

Can you explain this comment, please?

 

[Morty_UA]

I ran an informal speed test, using LAN Speed Test 1 to a Windows SMB share, and noted average speed over 3 runs. I got around 72 mbit/s.

The quick speed test isn't entirely valid (for example I measured from the '300 mbit' 5 GHz WLAN to wired LAN). But still, 72 mbit/s is well above the limit at 11 mbit/s that is my problem, and also well above my Internet connection which is advertised as 50 mbit downstream.

 

[Morty_UA]

Bump. Does anyone have some ideas; I'm really desperate to solve this.

 

[jdabbs]

Are you downloading from the same FTP server? They could cap clients that are active for more than X minutes.

 

[teknojnky]

Are you downloading from the same FTP server? They could cap clients that are active for more than X minutes.

or that have too many simultanious connections

 

[Morty_UA]

@jdabbs, @teknojnky: You are both right that the FTP server could be capping bandwidth. However, the IT department managing the server has responded to me, and promised that the FTP server doesn't cap transfer rates and that no traffic shaping is used on their side. ( If it was the server, It would also seem odd that re-setting the TCP streams several times (by closing & restarting the FTP software) eventually fixes the problem. )

It seems the problem is on my LAN, or with my home ISP. My ISP also claims to not use traffic shaping, but the only answer I got on that comes from a less-than-knowledgeable first line supporter.

 

[teknojnky]

well, have you tried from an ethernet connection? this would take the wifi out of the equation.

if you don't have any problems via ethernet, then I would suspect some kind of shaping or QOS on the wifi streams of the router.

 

[sfx2000]

Can you explain this comment, please?

Going back to the Deep Packet Inspection comment - it's the dark side of QoS - boxes like what Sandvine (and others) allow policy controls to be put into place to regulate and allocated bandwidth based on source/destination/port/traffic type - on both a static and dynamic basis.

A good example would be a Cable Service Operator that provides video on demand over IP - using a box like the Sandvine mentioned above, I can put a policy in place to prioritize this traffic over let's say Hulu or Netflix. This is just one example.

Another would be to say http traffic is value "x", FTP traffic is "y", and SIP/RTP is "z" - once you quantify that traffic, you can set the knobs accordingly.


Another would be to subtly transcode content to reduce bandwidth impact - with streams like video, it's fairly easy to transcode it down 10-15 percent with little to no impact on the video quality - or turn the screws down during peak time to ensure adequate bandwidth for all users downstream.

What's interesting about technology like this, it also provides for policy enforcement - in other words, bandwidth caps and throttling - and the ability to do session based services - for example, you can set an unlimited CAP at 50GB, after that it could be throttled down to let's say 1MB/1MB. This can all be done inside the core network, not at the edge/endpoint.

fun stuff...