inconsistancy in DL/UL limit, task speed and bandwidth used

[chaoscommand]

Hello,

After reinstalling windows and bitcomet 1.28 i encounterd quite the problem. Namely, no matter to what vlue I limit UL/DL speed, no bandwidth is left. Even tasks alone breach the limit, but the main problem is that that it slows down all other internet stuff. Network meter clerly shows that no bandwidth is left, but returns back to 0 when i exit bitcomet. It's not a router problem, as it works just fine on other PC's and no settings have been changed.

the numbers:

DL/UL limit @ 60kB/s

actual network traffic: variable, from double the limit up to 350 kB/s. My connection allows up to 450, but has to share it with TV, so effectively with TV on i have 180 kB/s left.

See the difference between limit and actual value...

now, how do i fix this? it is possible as I stated some lines up.

thx

post if you need aditional info

[The UnUsual Suspect]

Your saying that you have your global upload and download both set to 60kB/s, and bitcomet exceeds these limits?

If so, by how much and which protocols are being used when it exceeds the limits? (hover over the stats in the summery tab and it will show how much is used in each protocol).

There is a known issue where these limits are exceeded, but all reports to data have shown this to be only by a small amount. This issue is related to the complex nature of combining protocols between bittorrent, ed2k, http/ftp and LTseed. Usually setting a limit slightly under your target limit will suffice until this can be addressed with a software update, however I'm not sure if you're reporting the same condition/issue or something more serious.

Another option, since you're not very concerned with how fast you download, you might want to try disabling all non-bittorrent sources. In most cases you should be able to maintain your very conservative limit of 60kB/s using only bt peers, and it will make bitcomet's job of managing your bandwidth much less complicated.

[kluelos]

I suspect that you need to research your connection more thoroughly. 60 KB/s may be far too high an upload limit for it.

Many people don't understand that their connections are asymmetric -- the often forgotten "A" in "ADSL". That means that your upload bandwidth is NOT the same as your download bandwidth. Many people hear ads or promoters talking about "450 KB per second" and assume that this is your upstream speed as well as downstream. It almost never is, except on very expensive connections (like ISDN or T1) and very slow ones like POTS modem. Your upstream speed is usually much slower than your downstream speed. The fine print in your documentation should tell you that.

You should check your actual, received bandwidth with speedtest.net or similar service. Make sure you have disabled everything else that might use the internet. Make sure you have your units straight, don't confuse bits with bytes. Test at various times of day and on the weekends to get a good average value of what your upstream speed actually is. Then set your upload limit to 80% of that value. This should allow a bittorrent client to coexist with other applications that use the internet too.

[chaoscommand]

Hello again,

thanx for answering so fast

the speed limit is always over the limit, Bitcomet shows 65~66 kB/s, it was much higher before (fixed after reinstalling comet). But actually it uses 160kB/s, which I see from net meter (so more than double that). Stats show nothing special, no breaches there. Also , emule plug-in isn't installed and it shows that all traffic is from peers.

as for the second comment. I know and understand the difference between bits and bytes (thats why i capitalized B's) and know my bandwidth in and out, believe me that 60kB/s upload is not too high... (there is no A in VDSL, is there?). No other internet applications were/are running while I'm solving this, meaning my traffic is 0-5kB/s after exiting bitcomet. I believe I wrote something in lines of this already. Nevertheless I thank you for pointing out possible errors in my understanding of the above.

with best regards,

C

ps:I apologize for misspelling the topic name..

[The UnUsual Suspect]

BitComet can get data from bittorrent or LTseed peers. See if you can tell which is causing the bandwidth to exceed. Please try to catch it in a screenshot. If sucessful, I'll send it to development.

[chaoscommand]

Hello,

Bitcomet itself shows no significant traffic over the limit, but netmeter says something different, as seen in attachment. I hope i'm hovering over the correct stat, but on other hand, every one of them shows the same.

btw, i noticed that net meter scale is missing, it spans 0-180 kB/s. No other programs were running when print screen was taken, for further reference... After reinstalling twice more (current and also print screen state), it's not so obvious anymore, before it was just a bit saw-like around 160 kB/s compared to shown in attachement.

Judging from the bitcomet installer, could it be that installed files are not actually checked? Meaning it believes windows saying all files OK and makes no hash cheks? It seems as if something was corrupt when extracting (with the instaler), and a single reinstall didn't help (had to do 4 for effect, but it's still not perfect)

[The UnUsual Suspect]

I suspect the difference between the two is the method of measuring. Your netmeter program shows spikes that are high, assuming they are accurate, they still don't account for bandwidth usage at that level. BitComet (correctly) attemps to average your bandwidth to the guidelines you set, but if you were to look at the raw data, you'd see it higher one instance, lower the next... this is unavoidable.

I think to get an accurate test to determine if bitcomet's readings were accurate you'd need a program that shows average speed, not peak speeds.

[chaoscommand]

this net meter also shows average speeds, not shown on screen capture above. I'll repeat this, right now the difference between peak/low is quite large, but when i started this thread, it limited around 160, and average showed that. Also, I cant reproduce the problem in such magnitude anymore, now it averages at 120, sometimes 90, meaning it's not a priority problem anymore.

[chaoscommand]

there, finnaly managed to catch it in the act

yellow graph is download, and it's constant this time (scale is 0~217 [on auto, so it has to be equal or higher than 187.0] , also time scale is 5s for ech line meaning the whole graph spans one minute). Don't know why upload only hits 25, though.

Summary tab hovering again shows only bit transfer, so no eMule or LTseed and is same as seen in window name

Edited August 5, 2011 by chaoscommand (see edit history)

[The UnUsual Suspect]

If I was convinced that this netmeter results are accurate (which I'm not), then the first thing I'd look for is something else that could be using your bandwidth besides bitcomet.

Don't know why upload only hits 25, though.

If you shut down everything that uses the net, then go to speedtest.net and run some tests, you'll probably find your upload speed is much slower then download. Most connections are.

Also, there have been countless examples of users running bitcomet on limited connections where it consumes all available bandwidth, and BitComet's display has never been shown to be inaccurate. If it used twice (or more) then it displays, then this forum would be filled with user reports from the 2.5 million users daily, but so far, your's is the only one I've ever seen reporting this, so I have to question your data.

Perhaps you could look for another program that monitors bandwidth and see if it supports this netmeter's results, but to be accurate, you'd need a program that could identify only bitcomet's usage, you could have any number of processes that are using bandwidth.

[chaoscommand]

net meter was alway accurate (till yesterday it seems). I noticed that my line is filled in first place because firefox took forever to load a page - did it in an instant after shutting down bitcomet. Net meter just confiremd that.

what network meter do you recommend as ˝accurate˝

Once more, no other programs were runnig while I was taking the print screen and everytime before I posted, for that matter. Also, after shutting down bitcomet, net meter always goes down to zero, so obviously no other net based apps are running.

Maybe not 2.5M users, but googling for this problem(first thing i did) returns quite some unfinished threads, where answers range from set global limit to you have a slow line and bit/byte problem, so i'm not the only one. I admit that it is possible that this is a user specific problem, but it worked fine before.

bandwidth isn't a problem

[The UnUsual Suspect]

According to your test, you have a maximum download speed of about 490kB/s. Let's try an experiment, download a torrent from openoffice.org here. Set bitcomet's download limit to unlimited and run the torrent. With nothing else using your bandwidth, you should see bitcomet using most of your 490kB/s (probably about 450), and if netmeter reports higher, then you know it's faulty. On the other hand, if netmeter reports 490 and bitcomet reports much less, then this may require further research.

Save screenshots of the process in case we need to alert development of any potential problems.

[chaoscommand]

did as told

with bitcomet running

I'm intrigued... how does this show anything? speedtest got slower results as bitcomet was using bandwidth, right? also, bitcomet shows higher speeds than net meter.

(3.93 mbps=503.04kilobyte per second) so bitcomet shows wrong values here (you claim it's accurate?).

now, just for my info, i started speedtest without bitcomet running (to get actual download speed) and watched the net meter. netmeter was off by 5 kBps wich i deem within the tolerance area.

also, i think this has nothing to do with my issue that is, bitcomet uses up 3 times more bandwidth than what it shows

Edited August 5, 2011 by chaoscommand (see edit history)

[The UnUsual Suspect]

Your tests show that bitcomet is using as much of your bandwidth as possible on the openoffice torrent, slightly higher then the speedtest indicated, clearly showing the speedtest was conservative, but this isn't unexpected.

What it does show is that the amount of bandwidth bitcomet reports is accurate. If it was reporting less then it uses, then it wouldn't be reporting 517kB/s, since we know that this is approximately the max your isp will allow.

bitcomet uses up 3 times more bandwidth than what it shows

This is not supported by the facts. BitComet is showing that it is using 517kB/s at the moment of the screenshot. This is about a normal peak usage for a 4mbit broadband connection. Your speedtest proves bitcomet cannot be using more.

[chaoscommand]

err, i can't argue on this one

BUT

this test right now (thought it was to determine if either network meter or bitcomet is wrong..) was made with no global limit, so naturally the whole line got hoged.

If we return to the problem i reported: I set a global limit-> bitcomet task shows speeds equal or lower than the limit, bitcomet floating window, window name show limit+10%(roughly). no problems here. Now, my understanding is, that quite a lot of broadband is still free (bitcomet does show a reasonable value). But my network meter and my intuition say something else. Namely, internet browsing is painfully slow -a fact-, supported by netmeter showing that nothing of the line is left. Shut down bitcomet-> fast browsing, free bandwidth to spare according to netmeter, just as it should be.

517 kB/s equals 4.04, but since 4 is what ISP offers it can't be reached (not with this ISP)and sure as h*** i wasn't waiting for a spike to take the screenshot. but this is not an issue...

Now I'm not saying somethings wrong with the UL/DL display, but solely that something in bitcomet uses up my connection even with global limit set. I guess it's not actually torrent traffic, since the actual time needed for downloads is the same as computed for limited speed.

for the not supported by facts: of course it's not supported by any fact. I just tried setting limits to different values. Responses were similar, first a spike to top speed possible, then gets steady at about three times limit value (60-> 180, 40->120 etc.)

All this has been written in earlier posts

Edited August 5, 2011 by chaoscommand (see edit history)

[The UnUsual Suspect]

All I'm saying is that we tested bitcomet while it was using the max your broadband offers and it displayed accurately. It's also possible to exceed your isp imposed limit. They aren't 100% accurate in imposing speed limits and I often exceed mine. Normally they begin to throttle your connection only after it's exceeded the limit, and some will allow a margin above your limit to account for overhead,so it's normal to see any bittorrent client (I use others also) go above the isp limit.

So we have ruled out that bitcomet exceeds the limit in this one test. I can't think of a conclusive test to tell if bitcomet uses to much when it's own settings are controlling the speed. Your test is both subjective and circumstantial, no "smoking gun", and not even enough to present to development as something they need to look into. In order to confirm such an event, you'd need to have bitcomet running in a controlled environment where all network activity was monitored, only then could you tell exactly what bandwidth was used by your computer. At this point, we haven't even proved that your actually using as much bandwidth as the netmeter indicates, let alone that it's bitcomet that's doing it. There may be something with the large number of connections and packets that bittorrent protocol requires that is causing netmeter to report inaccurately... that's just one possibility.

If you want, try using another client and see if you can run the same task with the same limits. If netmeter still reports the same bandwidth usage, then you'll know it's faulty (somehow).

Also, to make the test fair, you should disable all non-bittorrent features in bitcomet, so it will behave like any other bittorrent client. This includes LTseed, emule, cometID, and all the "services". In fact, I'd do this first and repeat your tests, then try it again with another client.

ps. also your report that running bitcomet effects other net activity could be related to a weakness in your router, and not be bandwidth related.

[The UnUsual Suspect]

@support staff: can you please review this topic and see if there is anything here that I may have missed?

[chaoscommand]

installed and tested uTorrent with openoffice torent. Works normal, am able to surf with no delay, netmeter follows values stated by uTorrent, average netmeter results is same as uTorrents. hate the look of it, but I guess I'll change my bitclients

still, bitcomet was working just fine a week ago.

thank you for helping me in this thread, although it was just about me proving what i reported, not a single tip, solution or anything (yeah, i know, you never encountered something like this before. But google would have found a soltion if it wasn't something unprecedented and no post would been made here)...

g'night for now

[The UnUsual Suspect]

It's hard to offer a solution for something until the problem is pinpointed, and you're more then welcome to switch clients. However, did you try running bitcomet as a bittorrent client only? if so, I suspect it would behave exactly like utorrent (I use both by the way).

[chaoscommand]

By using it as as bittorent only, does that mean long time seeding and emule disabled? That's how I have been using it. Logging off passport didn't help either, same with unchecking everything in options/services. Am I missing something else that could cause this?

[The UnUsual Suspect]

Yes, bitcomet has the ability to download from sources outside the normal bittorrent swarm. None of these options have ever cause the problem you report, but if you disabled LTseed, emule and cometID, then it should limit you to only the bittorrent swarm.

It is strange that your not getting the same behavior with uTorrent. I'm really not sure what to make of this, because from what I understand, you're saying that it isn't downloading faster then your max setting, but for some reason it's using more bandwidth, like it was downloading redundant data (rubbish data perhaps). Can you please look at your "summary" tab on this task and hover over the "downloaded data" and see what is listed for "dropped data". If this is unusually high, then it might explain where the bandwidth is going, but even then, that's only half the picture.

If I could make some sense of this, then I could send the data off to development for them to do some testing, but I have yet to come up with a theory.

Can you please tell me who publishes this netmeter, if I can find the time, I'll see if I can do some testing with it.

ps. I'll leave this topic open so the rest of our support staff can review it.

[chaoscommand]

I'm really not sure what to make of this, because from what I understand, you're saying that it isn't downloading faster then your max setting, but for some reason it's using more bandwidth, like it was downloading redundant data (rubbish data perhaps).

Correct

dropped data 0 B

network meter: http://addgadget.com/network_meter/ and net meter: http://www.hootech.com/NetMeter/

[greywizard]

Just out of curiosity, if you disable DHT can you:

1. navigate normally on the Internet while BC is running?

2. see the same abnormally high speeds reported by Netmeter while BC is running?

[chaoscommand]

Disabling DHT really did make things manageable.

Does it have something to do with the huge amount of all peers?

[The UnUsual Suspect]

GreyWizard can explain better, but I think the reason is some routers have trouble with a lot of udp connection attempts. There is a way to limit them in BitComet, but you should probably see if there is a firmware update for your router.

Wiz may have other thoughts to add too.