Seeing strange fluctuations in throughout

[slovokia]

Hi,

Starting with Vuze 5.4.0 I am seeing strange throughput fluctuations on torrents that I download. The torrents typically have 100's to 1000's of peers as reported by the trackers and each one is connected to 80 peers at a time. What I see is a pattern of behavior where the download speed becomes very erratic swinging between saturating my downstream bandwidth capacity and then going to practically 0 bytes per second. The upstream throughput is more stable but occasionally swings down as well. When I look at the statistics download speed graph I see a lot of drop outs in the graph producing a very ragged appearance. In older version of Vuze the download speed would be stable. My global max download speed is set to 0. My upload speed is set to around 80% of the capacity of my internet connection when leeching. I have autospeed disabled.

Using the SAME torrent file with uTorrent does NOT produce such behavior - I achieve a nice stable download speed that is relatively stable after the client hooks up to it's peers. I tested the performance of utorrent with the same torrent within a few minutes of trying it in Vuze so I doubt that changes in my internet connection can account for this disparity in performance.

In my Vuze client I have a large number of completed torrents in a queued state - they are not actively being seeded but they are resident in my Vuze client.


 

[GaryE]

With that many peers on the torrent coupled with the fact that it works fine with another torrent client.  I would think it has something to do with the maximum number of connections or peers or similar.

Go to your preferences/settings/options whatever it is called on the OS you are using and check out the settings for:

tracker->Client
Override options->Limit the number of peers the tracker returns

It is possible that is set too low.

What are your encryption settings?

Again under preferences/settings/options whatever

Connection->Transport Encryption

is "Require transport encryption" checked?
If Yes
Which (if any) of the three fall-back options are checked?

It sounds like you are probably using a public tracker . . . I really do very very very little public torrenting but I would imagine that there might be some weired DHT setting which might be able to cause this problem.

Is there any kind of [time] period to the fluctuations you are seeing?  Meaning does your download speed drop every 5 minutes or something like that?  Or does it seem completely random?  If the fluctuations have a periodic nature it is possible that the fluctuations are being caused by something like the update interval set by the tracker -- if the tracker is a private tracker and or you only accepts peers from the tracker (because DHT and peer exchange are disabled ).

If you only were having problems with the one client I would say it sounds like you are being traffic shaped.  However, the fact that one client works well and the other does not means that a traffic shaping problem is pretty unlikely.  The only way traffic shaping could explain the problem with only one client is that if your uTorrent is set up with the proper settings to avoid your ISP's traffic shaping efforts and Vuze is not.  However, I would think that if you set up uTorrent to avoid traffic shaping you would know that you had to do the same to Vuze . . . so I find that pretty unlikely.  However, it is still worth me asking the question on the off chance it is possible.

So are you using the same port(s) in uTorrent and Vuze?  Are the encryption settings the same in both clients?  It could be as simple as the default port in uTorrent is not shaped by your ISP and the default port in Vuze is being shaped by your ISP.  I find it pretty unlikely that is the case but I guess it is possible.  Same thing with the encryption settings . . . it is possible that the default encryption settings on uTorrent are enough to spoof your ISP traffic shaping efforts while the default encryption settings on Vuze do not.  Again unlikely but possible.

Well I think that is all of the ideas I have LOL  Maybe someone else can offer more definitive advice but take a look at all of that and report back your findings/settings and we can go from there.

Good Luck!

[slovokia]

tracker->Client Override options->Limit the number of peers the tracker returns is set to 100

Encryption is set to Require encrypted transport. Neither of the allow non-encrypted options is checked. Use the cryptoport option is checked. (On my uTorrent client I also force encryption for all connections).

Both Vuze and Utorrent are using ports that are connectable from the external internet.

I am using public trackers for these torrents.

The kind of time frame these fluctuations occur at are on the order of 5 to 10 seconds. I see radical changes in throughput over such intervals.

I check my connection with Glasnost and it reports no traffic shaping. I also checked with some other techniques - the most that could be occuring is some light congestion within the DSL network that services my connection. But again Utorrent does not display the same wild oscillations in throughput on the SAME internet connection at the SAME times.

Finally this is a CHANGE in behavior relative to the way Vuze used to work on my computer over many years under the same conditions.

 

 

[GaryE]

Well the fact that you require encryption does explain why you can only connect to a small percentage of peers especially on a public site.  However, that is not really your complaint LOL

On a public torrent 100 peers might be too little.  I am not sure but change that to 500 or so and see what happens.  I doubt it will fix the problem but it can not hurt to try (you might have to increase memory so I would go up in increments of about 75 or so and make sure you do not get an out of memory warning after about 10 minutes and then increase again etc etc etc) .  It might be that the default on the latest version of vuze (depending upon how you installed it) might be significantly lower than it was in the past.  I really do not know.  I more or less exclusively use private sites and getting more than 100 peers on an individual torrent is pretty rare so it does not effect me; but it is a possibility.  Also since you are limited to 100 peers in Vuze and it is a public site . . . it is possible that you are just getting connected to slow peers with bad connections or whatever.  It is pretty unlikely that you would connect to exactly the same peers on both clients so this is a possibility.  However, I doubt this is the problem.  I am sure if this only happened with a small number of torrents you would not posting here and the odds of you getting 100 slow peers on each of say 5 torrents are pretty astronomical.  So I doubt it is the problem but if it is only with one torrent at a time or something or only a few torrents it is a possibility .  .  . granted a remote one . . . but still a possibility.

The fact that glasnost says you are not getting traffic shaped is very definitive.  So that is clearly not the problem.

Depending upon which version of uTorrent you are using there might be a problem.  There have been versions (and I do not remember which versions) that would vastly favor uploading to/downloading from other uTorrent users.  If you are using a fairly recent version of uTorrent I am pretty sure this would not explain your problem . . . but if you are using an older version . . . it is a possibility.  There have been several versions that would preferentially upload to/download from other uTorrent users so it is worth checking out.

The ports you forwarded.  Are they both in the approximate range of 50,000 - 60,000 ?  Since your glasnost test says you are not getting shaped it is unlikely it will make a difference . . . but I am grasping at straws trying to explain your problem so if say your uTorrent port is in that range and your Vuze port is not there is a very small chance that chaning your port to that range might solve the problem -- but I doubt it.  LOL  But this will only take a few seconds to check .  .  . so take a look.

Hmm since you are using public torrents . . . are your DHT and peer exchange settings the same in both clients?  If they are different it could explain the problem.  But since you know what a glasnost test is I suspect you know enough to have the correct settings . . . but it will only take a minute or two to check . . . so why not verify that?!

That is all I have . . . if none of this helps . . . I think I am out of ideas.