Home Forums Networks & Cross-Platform Networking Internet d/l speed v u/l speed?

I recently upgraded my DSL service from 3Mb/sec to 7 Mb/sec and I'm mystified by two aspects of the change...

1. While my d/l speed more than doubled, from 2.7Mb/sec to 6.4Mb/sec, my u/l speed increased by less than 10%, from .67Mb/sec to .72Mb/sec. (All more or less, of course.)

2. When I run multiple tests, i.e. Verizon, Speakeasy, and Speedtest, my results differ slightly, but they are consistent in that as d/l speed goes up, u/l speed goes down.

Can anybody explain?

Thanks.

Check to see what level of service you are buying from your internet service provider. I guarantee you are buying ADSL. The download/upload speed differential is why it is called ADSL or Asynchronous (or Asymettric) Digital Subscriber Line. For a LOT more money per month you may be able to get SDSL (Synchromous or Symetrical Digital Subscriber Line) service. ADSL simply means you have different upload and download speeds. The providers reason that the normal user has much more download traffic than they do upload traffic so they can save themselves a ton of money in equipment costs by going with the slower upload speed. This is generally acceptable for the vast majority of users and few ever notice the difference. If you are hosting a web site or for some reason have a heavy upload demand most providers can provide DSSL service but they are going to charge the user for the additional equipment and bandwidth that is required.

For example, my son and I have the same 18Gbps download speed with the same provider. However, because of his requirements for uploading large volumes of data he has SDSL and I have ADSL. My monthly bill runs $35 and his bill is over $100 a month more.

> I guarantee you are buying ADSL.

Thanks for that; it is, indeed, the answer my question.

My 2nd question, "When I run multiple tests, i.e. Verizon, Speakeasy, and Speedtest, my results differ slightly, but they are consistent in that as d/l speed goes up, u/l speed goes down," still has me scratching my head, though.

Does it ring any bells?

Don't get wrapped around the axle over speed test results unless they are continually significantly less than you are paying for. You need to take a look at the speed tests themselves and what they are reporting.

• The test results you see are not averaged over the length of the test rather they are the peak upload or download speed that was achieved during the test. I think that is because speed tests are often used for "bragging material" so just as people will say their car gets 55mpg they neglect to mention the car acheived this for a ten minute period rolling downhill from the top of Pikes Peak. IMHO it would be much more informative to report the average speed over the entire test. Of course that would make a lot of people very unhappy with their high speed internet service.
• You are not testing the local line loop segment that you or your ISP have some control of. You are testing the entire internet network from your location to the testing server. There is a lot of hardware in that loop, and traffic can and does have a major impact. All of which neither you nor your ISP can control.
• The "speed" you get reported on a test may have been acheived for only a fraction of a second and could in some sense be considered an aberation.
• Because of all this small variations are essentially meaningless and tell you little about actual performance.
• Upload and download tests are run serially not in parallel and as a result there should be little if any relationship between the two other than the average network traffic load at that point in time.
• You will almost never get 100% of what you are buying because of network conditions beyond the control of the ISP.

I just ran a series of tests and over six trials to a test server less than 50 as the crow flies miles away (cable distance is probably significantly greater than that) and my speeds were:

17.11 — 2.09
14.69 — 2.10
17.22 — 2.09
13.93 — 2.10
21.40 — 2.09

I then ran a trial to a server in London England and the results were

7/73 — 1.95

Noticed the upload speed to London was not nearly as impacted by the distance as the download speed. My interpretation is there is no significant difference, statistical or otherwise, in an upload speed of 2.09 to 2.10. Basically I am getting the upload speed I am paying Charter to provide and in at least one trial I got significantly more download speed than I am paying for. FYI my ISP is Charter Communications.

Thanks for another exemplary post!

> Upload and download tests are run serially not in parallel and as a result there should be little if any relationship between the two other than the average network traffic load at that point in time.

My results, of which these are the most recent, seem to contradict that (even if only marginally)...

Verizon:     Down = 6.40Mbps / Up = 0.77Mbps

Speedtest:  Down = 6.46         / Up = 0.72

Speakeasy: Down = 6.47         / Up = 0.72

That scenario is 100% reproducible... The Verizon test always shows the lowest d/l and highest u/l speeds, while the other two tests always yield similarly differing results, and u/l speed always decreases as d/l speed increases.

That seems to point to some sort of connection between the two despite your comment to the contrary.

But please don't get me wrong; I'm in no way displeased with my Verizon service...merely mystified by that apparent connection where no connection is supposed to exist.

Thanks for bearing with me on this.

The very slight inverse relationship between upload and download speeds is curious, but I am unconvinced that it is remarkable. My results, as you can see, are similar to yours. My contention is the variations we are seeing are too small to be statistically or practically significant. You mentioned that some of your tests were through WhatIsMyIP and I was testing through the same service. As small as the variations are, it is entirely possible the diffferences are an artifact of the testing methodology and software or it could be an artifact of momentary server load. And before I forget it, there is also the effect other tasks running in my system can and will have on the testing results.

Notice that where there was a major difference in network speed, in my tests ie. testing to the server in London England the upload speed fell off proportional to the difference in downlaod speed which is what you would expect.

If I were interested in accurate statistically valid results, with absolutely no other tasks running in my computer, I would run several sets of say 100 trails each, scattered throughout the day. Even that would not escape the influence of the particular testing methodology and software used by the test site. But in this case the more variables you remove, the less the testing will reflect your actual experience in day to day use. As a practical matter, the differences we are seeing are only detectable in testing and will never be noticed in real world operation unless you are moving VERY large volumes of data one way or the other.

Many thanks for sticking with me!

My curiosity will now yield to your expertise, and I'll file the subject matter of this thread away for future reference should a "trigger" ever arise.

Thanks again.