5

We have 25 megs on the downstream and 2 megs on the upstream. When doing a speed test without anyone downloading (from Usenet, for example), we get the full download and upload.

However, if one person is downloading at max speed (let's say, 3 megs per second), it brings down the internet for the rest of the house, even though we have 25 megs on the download and only 3 are being utilized.

Why is this? Why does throttling it slightly fix this problem?

Thanks

3
  • 1
    Wait… isn’t Usenet discussion groups?
    – kinokijuf
    Commented Jan 27, 2013 at 20:39
  • 1
    @kinokijuf: Yeah you'd be surprised how much bandwidth large amounts of text takes up.
    – John Smith
    Commented Jan 27, 2013 at 20:51
  • 3
    ...especially if the text consists of base64 encoded binaries. Commented Jan 28, 2013 at 1:39

1 Answer 1

5

It's the sneaky difference between megabits and megabytes that's killing you. Internet connections are typically rated in megaBITS (Mb), whereas downloads are measured in megaBYTES (MB). To convert between the two, divide by 8:

25Mb downstream / 8 megabits per megabyte = ~3 MB per second
4
  • 3
    And throttling it slightly fixes the problem because when you are at the absolute maximum bandwidth consumption, queues tend to be full, destroying latency for everyone. If you're even slightly below the maximum, the queues tend to be nearly empty, preserving latency. Maximum-bandwidth applications (like downloading with no limit) don't play nicely with latency-sensitive applications (like browsers). Commented Jan 27, 2013 at 20:36
  • The exact formula: 25 * 1000 * 1000 / (8 * 1024 * 1024). The thing is that megabit is measured in powers of 10, but megabytes are usually measured in powers of 2.
    – Malcolm
    Commented Jan 28, 2013 at 1:46
  • @Malcolm: That's pretty close, but still not exact. The megabit measurements are line speeds, the megabyte measurements are useful data speeds. Some of the line speed has to carry address and control information. Commented Jan 29, 2013 at 2:31
  • 1
    @DavidSchwartz Yeah, but there are many layers in the network, and one protocol may require more traffic for headers than another, so it's quite hard to determine the exact bandwidth if we want take such subtle things into account.
    – Malcolm
    Commented Jan 29, 2013 at 3:02

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .