spacer search

After Hours Information Technology -
The small IT business for your small business.

Main Menu
Web Hosting
Domain Lookup
Speed Test
Contact Us
Privacy Policy
News Feeds
Microsoft ISA Server
Home arrow News arrow Microsoft ISA Server arrow I'm not getting my maximum bandwidth!

I'm not getting my maximum bandwidth! PDF Print E-mail
Frequently Asked Questions - Microsoft ISA Server

Time and time again we're asked to check a clients systems to see why they're not getting their maximum performance out of their Internet connections.

Chances are everything is running fine and the theory behind why they believe theres a bottleneck is flawed

(Not really an ISA specific "problem" but I though this was the best section to put this article)

Lots of people run monitoring software on their links to tell them if the link is down, and/or how much traffic is passing through. One of the most common is the ever useful MRTG.

There is however an inherent flaw in using MRTG graphs to think that your not getting full bandwidth. MRTG only ever takes a "point in time" snapshot of how utilised a link is, more often than not its once every 5 minutes. This means it's checking the link something like once a second in every 300 seconds (5 minutes x 60 seconds). This means its showing at best about .3% (yes point 3 of a percent) of the overall usage. Now, the theory is that MRTG shows an "average" of "common" utilisation which is fine, but because of this you cannot say that you not getting full utilisation just because your graph isn't maxed out.

To do this you really need a sustained transfer over at least a couple of reporting measurements. So if you're checking every 5 minutes, it should be a transfer of AT LEAST 10 minutes, if not longer since you don't know exactly when those 5 minute intervals start.

The other reason people may think they're not getting max speed (and in actual fact are quite possbily right) is because of the "weakest link in the chain" theory. This theory states that "a chain is only ever as strong as its weakest link". How does this apply to bandwidth? Lets try a case study.

  1. You have a 512/512K SDSL link to your ISP
  2. The ISP has a 100Mb network backbone
  3. The ISP has a 10Mb upstream link to "core Internet" or upstream bandwidth provider
  4. The remote ISP has a 2Mb link to the internet
  5. The website your trying to access is hosted on a machines hanging off the end of a 64K ISDN connection.

Now, what's the maximum speed you can download this particular webpage? Most would immediately assume its 512Kb. Let's look at the numbers to see what the FASTEST speed is. In fact its the 100Mb network backbone of our ISP. Woohoo! We can get the webpage at 100MB.. actually no we cant, because out link is only 512Kb.

Let's look then at the MINIMUM speed in that chain. It's the 64K ISDN link at the remote end so therefore the maximum theoretical speed we can view/download this particular webpage is 64K.

Why did I say "theoretical" there? Simply because we live in "the real world" where theory rarely represents what actually happens.

Connectivity, bandwidth and infrastructure cost significant sums of money. Because of that there is not some endless supply of bandwidth that means that ISP's, core transit providers etc allow for every one of their users to their individual "full" bandwidth at any one time. In fact, they provision their networks only for what's "normal" and not the maximum. Is every road we drive on a 6 lane freeway in case there are thousands of cars that want to drive on it? No. Let's try an exmaple based on what we had above. (These are theoretical numbers only!)

If an ISP has say 1000 customers who all had 512Kbps ADSL links, then the theoretical maximum would be 512,000Kbps. That would mean to satisfy the theoretical maximum demand of every customer simultaneously they would need a 512Megabit uplink, basically a gigabit ethernet link!

In actual fact, there would be an assumption that say at any given time, all customers might be using an average of 10% of their max link speed, so if we divide our theoretical maximum from above by 10 we now need a link speed of 51.2Megabit, or a normal 100Mbps Ethernet connection. In truth, this assumption about customer activity may be even lower at something like an average of 1% usage meaning a 5.1Mb link - something thats FAR easy to get provisioned and I would expect significantly cheaper than a Gigabit link that we started with.

Internet traffic is "bursty" stuff, meaning it comes and goes in short bursts, hence they principle of not provisioning for full maximum theoretical bandwidth usage is completely sound. The actual ratio of formula that ISP's is undoubtedly something they hold very close to their chest and is called a "contention ratio".

So, back to the statement right up the top there. How do we check to see what speed you're getting on YOUR link. The best bet is try try and transfer a large-ish file directly from your ISP. Most ISP's provide your own private webspace so if need be transfer some files in there to do some test downloads. Many ISP's even maintain their own "mirrored" copies of files (things like web-browsers and other internet apps.) Some even have movies trailers and the like.  What this will mean for you is that the content will be as close to you "on the network" as possible and the likelyhood is that of all the links between you and the actual content, your link is probably the slowest.

We've provided a download speed test that will be of some help to Austalian users. The actual content is held on a an Optusnet webserver so they should have quite decent peering/bandwidth arrangements to most locations in Australia. Outside of Australia, this test is probably pretty pointless given the couple of hundred millisecond time delay and congested cables under the pacific ocean (or Indian depending on which side of Australia the cable is coming from!)


< Prev   Next >
Locations of visitors to this page


Copyright 2004 2005 After Hours Information Technology