Beta5 Traffic limits based on % of speed not working correct 614j16

[rwills2]

Windows XP, FF 7.01, to easynews.com. My maximum throughput is 1787KB/s. When I select heavy, it jumps to about 2300KB/s for a couple of seconds then drops to 1787. When I select medium it goes to 1787KB/s and stays even though I have the percentage set to 60% which should equate to about 1072KB/s.

[Usher]

Thanks for your report.

Yes, this option for many s is not usable yet. For example:
- it ignores bittorrent settings;
- it doesn't count and speed separately;
- it doesn't allow to quickly switch from automatic to manual settings, because manual settings are reset to 1;
- it shows speed in bytes per second with binary prefixes while bandwidth is usually defined in bits per second with decimal prefixes (for example Ethernet may be 10 Mbps=10 000 000 bits/s), so you must recalculate speed on your own.

Now you understand why this version is called beta, I hope.

[Alex]

When I select medium it goes to 1787KB/s and stays even though I have the percentage set to 60% which should equate to about 1072KB/s.

It can do so for some small period of time (about 15-25 seconds) in order to recalculate the network speed. In some rare cases it can do so during more than 30 seconds (we'll fix it).

[Alex]

- it ignores bittorrent settings;

Bittorrent settings were separate always.

- it doesn't count and speed separately;

It does.

[Usher]

- it ignores bittorrent settings;

Bittorrent settings were separate always.

- it doesn't count and speed separately;

It does.

Well, I mean using torrent settings for automatic speed measurement, not for ed/ed bytes count.
In my case on the start I've got 12 torrents that are not seeded. With automatic mode on FDM tries to measure speed and counting only communication with tracker it goes to very low speed (some 100-200 bytes/s). With that speed FDM allows to start only the first torrent, 11 others must wait.
In the next test I've got the first torrent only seeded (ed), not ed, and automatic speed measurement was also only for . Results are the same as above - only the first torrent starts.
Now FDM starts with manual settings and 12 torrents that are not seeded again. This time it starts only 3 first torrents and 9 others are waiting a few minutes with an hourglass.
So:
- Where and how does FDM use traffic for speed measurement?
- Why some torrents are waiting if FDM doesn't use network settings for torrents and vice versa?
And there is no possibility to set bandwidth for network in manual mode.

[Alex]

In my case on the start I've got 12 torrents that are not seeded. With automatic mode on FDM tries to measure speed and counting only communication with tracker it goes to very low speed (some 100-200 bytes/s). With that speed FDM allows to start only the first torrent, 11 others must wait.

Currently FDM will start 1 only for network speeds less than 6Mbps.

In the next test I've got the first torrent only seeded (ed), not ed, and automatic speed measurement was also only for . Results are the same as above - only the first torrent starts.

I'm not sure if I understood you, but as I said - it will run 1 only for speeds < 6Mbps.

Now FDM starts with manual settings and 12 torrents that are not seeded again. This time it starts only 3 first torrents and 9 others are waiting a few minutes with an hourglass.

This can be due to torrent files rechecking. Libtorrent prefers to not start other torrents in such cases.

- Where and how does FDM use traffic for speed measurement?
- Why some torrents are waiting if FDM doesn't use network settings for torrents and vice versa?

It does not at all for now. But (by design) it calculates /s speed separately. We just disabled (temporarily) new speeds calculation and management algorithm (due to some internal reasons).

And there is no possibility to set bandwidth for network in manual mode.

This is not implemented in interface currently. Limits for traffic are specified using the old way.

[Guest]

i'm using this the way i always have and it's not working at all. these are not torrent s. regular http from a services that allow parallel s.

firstly i thought it was because it had measured my speed incorrectly, it was something crazy. orders of magnitude away from the correct value. something like 67MB/s in fact it was displayed in in scientific notation it was so big. like 6.7E07

but then i entered a manual value and it still doesn't work properly. happen the same faulty logic is being used to regulate as was to measure.

[Alex]

i'm using this the way i always have and it's not working at all. these are not torrent s. regular http from a services that allow parallel s.

firstly i thought it was because it had measured my speed incorrectly, it was something crazy. orders of magnitude away from the correct value. something like 67MB/s in fact it was displayed in in scientific notation it was so big. like 6.7E07

but then i entered a manual value and it still doesn't work properly. happen the same faulty logic is being used to regulate as was to measure.

Please describe in more details what you expected to get and what you got.

[Usher]

Currently FDM will start 1 only for network speeds less than 6Mbps.

How did you get 6 Mbps? Is it a bandwidth for any standard connection type?
This decision is arbitrary and completely baseless. With such conservative settings FDM possibly won't run at a full connection speed, so it cannot measure that speed properly. We have a vicious circle here - FDM won't run fast because it doesn't want to run fast.
In the old times I have succesfully ed more than 1 file at a time even using standard phone modem. Nowadays many home servers are connected via some xDSL connection with bandwidth 256 kbps, and their owners usually don't allow to grant all the bandwidth for server use, so starting with 6 Mbps as a base is a huge mistake.
In my opinion - s should have possibility to set that limit as they wish, and one file for 56 kbps is good for the start, I think.

I'm not sure if I understood you, but as I said - it will run 1 only for speeds < 6Mbps.

With my 100 Mbps Ethernet card I can use:
- 100 Mbps connection with my WiFi router and a LAN ftp server;
- 54 Mbps WiFi connection with my notebook;
- 10 Mbps Internet connection.
How does FDM know which connection uses which speed?

Staying only with my 10 Mbps Internet connection - speed measurement in FDM is wrong, because FDM counts less than 6 Mbps. When I set manually speed to 1 MiByte/s ~= 8 Mbps, FDM starts 3 torrents, not 1 and 1/3.

Testing LAN ftp server, FDM speed measurement is also wrong. FDM can safely use 100 - 10 = 90 Mbps bandwidth, but the real count in FDM is lower: 60-70 Mbps speed for a single huge file in 4 pieces and 70-80 Mbps for 2 huge files in 8 pieces.
FDM starts every as a single piece and it takes more than a minute to count speed and go to 4 pieces per file. That way for small files in automatic mode FDM just has no time to accelerate to the full speed and total time is always longer then for manual settings.

Now FDM starts with manual settings and 12 torrents that are not seeded again. This time it starts only 3 first torrents and 9 others are waiting a few minutes with an hourglass.

This can be due to torrent files rechecking. Libtorrent prefers to not start other torrents in such cases.

1. I reordered torrents on the list and nothing has changed, so it's definitely not file rechecking.
2. File rechecking goes with a yellow arrow (and big U load), not with an hourglass.
3. Why in the beginning FDM starts only 3 medium torrents and then allows to run 9 huge torrents at once? Why not next 3 ones?

[flagpole]

Please describe in more details what you expected to get and what you got.

well for a start this is how it measures by bandwidth:
http://nigelcoldwell.co.uk/test/fdm.png
86MB is a little high for a 7Mb connection.

but setting the limits now has no effect, even if i manually tell it my network speed. it doesn't slow the at all. previously i have had
-heavy mode as unrestricted for when i am away from my computer
-medium around 90% (worked out in my head) to allow reasonable browsing whilst ing
-light around 40% for doing things like looking at youtube and keeping my s running.

none of this works now. the traffic usage mode is useless. everything s at full speed meaning browsing is extremely unresponsive and if i want to do something like look at streaming videos i have pause my s.

this seems to be one of the few fundamental changes you have introduced during this beta cycle, it surprises me you would introduce it so close to the end.

I would like to understand the logic used to determine the maximum speed. it's a bit frustrating to see FDM sat there ing and maxing out my connection, the exact value of my connection speed there in the status bar, yet the reported number being crazy.

[Alex]

...

Do you use FDM to files from network disks?
E.g. using something like "\\server\c\myfile"?

For manual settings - make sure you specified it correctly.
7Mbps = 0.87MB/s.

[Alex]

If FDM shows 86MB/s speed of your connection - this means it s with this speed sometimes for some unknown reason. Do you use some caching software or something like this?

[Guest]

no. there is nothing unusual about what i'm doing.

they are just http, well https s from as it happens Rapidshare.

i've never used it for local or LAN s. and i have no funny monitoring or odd AV software that would confuse it save the network meter gadget.

i understand the difference between b and B. i put them in bold to show the distinction.

the only time i have seen a network meter confused like this is when the u is tasked to something with a high priority and then when it cycles back it miscalculates the speed as all the delta hitting it at once.

i guess the clocks did go back on saturday. which oddly can mess with these things. but i've tested it more than once. anything that is unusual is FDM itself. i do have FDM set to cache the writing of s do disk. but i thought this was a default option. if it is measuring this then i guess 86MB/s is a reasonable speed for that.

win 7/64 btw.

[flagpole]

no. there is nothing unusual.....

that was me btw. always forget to to this forum.

[flagpole]

so the limits seem to be sort of working. it's just the speed estimate that is wrong.