1

I am having a problem with my download speed (ADSL) at home with my home router (a Comtrend device) I have done a benchmark of my download speed on may occasions using wifi and ethernet connections.

I found out that at different times through the day I get different speeds ranging from 99%-100% of the bandwidth in the morning to 10%-20% at night. these numbers happen when I am downloading using only one connection (default browser download).

When switching to multithreading in downloads, using multiple chunks and connections (Internet Download Manager, JDownloader, etc…), I get 99%-100% of the speed all day long.

My benchmarks were done using the sam file form the same server (Mega.co).

My questions are:

  • How much is this difference in downstream speed due to the server I am downloading from?
  • Is there more benchmarks I need to perform to get a better understanding?
  • And is there a known way of pushing the speed to its max on single thread downloading?
  • From all you are saying, your ADSL connection stinks and the reason download speeds “increase” when you multithread is that when one connection in the pool dies, the other one picks up. So multithreading is basically catching the drops and compensating for them. To have consistent speed without using apps like that, you need to get that ADSL connection fixed. Simple as that. The benefit of download managers is dealing with occasional issues like this; like if you are traveling and have spotty Wi-Fi. But at home? You should get that fixed. – Giacomo1968 Jan 6 '20 at 17:11
  • Depending on the server, it's utilisation, latency and a bunch of other factors such as pathways to or from your particular part of the world or even in parts of the world you don't expect, there may not be anything you can do about this. Routes change over time and it also may be that at peak times a route to or from your ISP is saturated. Using multi-downloaders can make you effectively appear (from a traffic perspective) as multiple users and give you multiple shares of the bandwidth in any given link. – Mokubai Jan 6 '20 at 18:14
  • @Mokubai thanks, but couple of questions : could this idea of multiple connection be something that is configurable in my home router/modem ? second, is there anyway to practically fake multiple connections on a HLS stream ? and thus getting a full speed ? – Kaki Master Of Time Jan 7 '20 at 17:16
1

How much is this difference in downstream speed due to the server I am downloading from?

Mega.co probably implement's QoS and other bandwidth restriction based on several parameters... The two easier to find out are: 1. If you have a registered and paid account or not. 2. If you are alone on the server or if there's several people.

So i think your benchmark isn't relevant since you do not test your home connection but both your home connection and the server in front.

Is there more benchmarks I need to perform to get a better understanding?

There's services on the web that allows you to benchmark your network connection capacity on controlled server: speedtest.net is one of them.

If you do not want to use one of these services, then, for your benchmark to be totally relevant, i think you should better use at least two "neutral" services on the web. For example, a server you or a friend of your own and another server that is known to have strong bandwidth and hardware (windows update, linux FTP, whatever...)

And is there a known way of pushing the speed to its max on single thread downloading?

This is your operating system job. Linux, Mac OS X, Windows implements their policy regarding how a transfer is managed. It's hard to tell since we don't know your OS.

To finish, keep in mind that network performance is a complex subject since it depends on many factors:

  • Connection type: a microwave oven can interfere with WiFi ...
  • Router type: SOHO router or Operator router are often cheap ones that sometimes fail at managing too much connections from too much workstation. I also already saw routers that, the longer they run (in uptime), the poorer their performances were ...
  • Server type, as already mentionned.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.