Where is the bottleneck in this networking situation?

[Epsil0n00]

At the school I work at we are going to be doing some online testing tomorrow. All of the students will be using laptops, connecting wirelessly to the testing site on the internet. I am concerned that their wireless connections will bottleneck their access to the testing site... Here are all the links in the chain from the classroom they will be using, out to the internet... if you wise folks could take a look at this and let me know where the bottleneck will be (should be) I would appreciate it! I am hoping that the only bottleneck will be our DSL speeds from the ISP--I want to make sure everything we can control on our side is not slowing things down.

1. Student's laptop > 802.11b wireless card
2. 802.11b wireless card > Cisco Aironet 350 WiFi
3. Cisco WiFi > 100Mb connection to Cisco switch
4. Cisco switch > Fiber connection to Cisco 4000 router
5. Cisco router > 6Mb DSL connection to internet

The way I figure it, the 6megabit DSL connection should be the slowest connection in the chain and therefore the bottleneck (assuming that every student gets a full 11Mb/s from their wireless connection. We have two wireless access points in the room and a maximum of 16 students attached to the AP at anytime.

What do y'all think?
Thanks in advance!
Epsil0n

 

[spidey07]

That's close, the wireless "could" be the bottle neck due to half-duplex and media sensing - ie only one person can talk at a time.

Some actual testing yield real results.

one or two people at 11 mbs wireless? DSL is bottle neck. Bunch of people on wireless at same time? hard to tell.

 

[ScottMac]

More than likely, the wireless link will be the botleneck.

First, the "11 Meg" throughput is really ~5 meg typical.

Second, the AP is acting as a hub: the bandwidth is shared equally (depending on teh NICs) amongst the users. So, depending on the type of traffic and the timing, the AP is likely to have a little congestion.

There are some parameters you can tweak to optimize the system, but regardless, the wireless link will probably be your choke point.

Good Luck

Scott

 

[Epsil0n00]

Darn... well, I guess we'll give it a go and see how we do. I will post back and let you know how great or badly it all went.

Thanks for the feedback.

 

[spidey07]

Scott,
That was my line of thinking, but testing with the application in question or traffic generators would be the only way to truly tell.

 

[Kelemvor]

What exactly does this "testing" entail? If it's just going to a website and loading up info on the page every so often it would probably be fine. However if it involves constant sending and receiving of data (not just loading a web page) then you could run into problems.

You could always start them up one at a time and wait a minute or so between and see when the first people start complaining that it starts running slow. Or get some G cards and use those if the APs support it. Or better yet, get a bunch of Cat5 cables and let everyone plug in. heh.

 

[ScottMac]

Spidey: of course ... best guess based on the info provided.

Franky: 350s are 802.11b only.

We'll see .......

Scott

 

[Epsil0n00]

I personally predict problems but the boss says that it is not worth the investment to purchase 2 additional 100megabit ethernet hubs to ensure that everything goes smoothly... so we shall see what happens...

 

[melthemoose]

if the primary purpose of the wireless connection is to allow internet access, mainly browsing, then both the wireless and DSL connection will compete to be your slowest links. wireless has the greatest set of variables (distance from AP, overall signal strength, power settings on the laptop etc).

 

[Epsil0n00]

RESULTS:

The first group of testers went fairly smoothly. Each question took about 10 seconds to load as it was pulling audio and video over the internet. Things went fairly well...

However, around 11:30 our internet traffic always spikes as more people jump on during lunch time. It became so slow that connections were timing out and pages wouldn't load. This rendered the testing useless as they could no loger bogart the majority of the bandwidth... and the testing load made the rest of our campus' internet connection non-functional. I sent out an email asking everyone to refrain from using the internet until the testing was complete and I think that helped the testing proceed sucessfully.

I think this shows that the DSL connection was the bottleneck... some of the slowness in loading the questions (DLing the audio files) could be attributed to the wireless connection... but the internet saturation was certainly campus wide and a bottleneck of our DLS line. I guess that would be the difference...

If the wireless were the bottleneck people would still be able to get the testing information, but at a slower speed than optimal. However, when the DSL gets saturated it refuses connections and totally stops functioning.

Eps.

 

[spidey07]

well hell!!!!!!!!!!!

you didn't include the traffic patterns of your internet connection in your question.

time to get a packet shaper if you want your apps to perform well.

 

[randal]

Or buy more bandwidth ... even though there's no such thing as "enough bandwidth." ;-)

 

[thriemus]

Setup a proxy server running squid and snort. These days its easy to setup all sorts of authentication from NT Auth or radius, to basic acl's (Access Control Lists) based on ip range. The proxy will maximise your 6 mb line in that it will not be downloading the same page or files over and over. When you consider that most students all visit the similar web pages, this will also let you firewall off pornography and other restricted sites that you would obviously not want the students to be accessing on the schools network.