Reinsert the "small send buffer" function?

Use this forum to flesh out your feature request before you enter it in <a href="http://dcpp.net/bugzilla/">Bugzilla</a>.

Moderator: Moderators

Locked
ratsalad
Posts: 4
Joined: 2006-01-29 10:27

Reinsert the "small send buffer" function?

Post by ratsalad » 2006-01-29 10:35

I'd like to see that function in future releases of DC++, downloads are effectively killing any possibilitys for me to download anything, true enough, sharing's the idea of DC++, but still it'd be nice to see that option again.

Anyone thinking differently?

TheParanoidOne
Forum Moderator
Posts: 1420
Joined: 2003-04-22 14:37

Post by TheParanoidOne » 2006-01-29 10:45

What version of DC++ are you using?

Note that in 0.6811 this item is now a configurable value. See Settings -> Advanced -> Experts Only -> Socket write buffer.
The world is coming to an end. Please log off.

DC++ Guide | Words

Ninjai
Posts: 3
Joined: 2006-01-23 03:36

Post by Ninjai » 2006-01-29 10:56

What would be (the) default values for the socket buffers? And what would be a good write buffer size???

ratsalad
Posts: 4
Joined: 2006-01-29 10:27

Post by ratsalad » 2006-01-29 10:58

Well, the version i'm using is 0.6811, wouldn't OP's take me adjusting this buffer as bandwith limiting? and how do i set it up so that it gives the same effect as the "small send buffer" effect?

ullner
Forum Moderator
Posts: 333
Joined: 2004-09-10 11:00
Contact:

Post by ullner » 2006-01-29 11:02

Ninjai wrote:What would be (the) default values for the socket buffers?
64*1024 bytes. It should say that aswell in the help file.

ratsalad
Posts: 4
Joined: 2006-01-29 10:27

Post by ratsalad » 2006-01-29 11:06

The reason i used tp praise DC++ was because of it's relative ease of use, this way of setting it up seems a bit awkward to me... I'll try, though ;) Thanks a lot for your time and your help :)

So I should just put the "64*1024" value, then?

TheParanoidOne
Forum Moderator
Posts: 1420
Joined: 2003-04-22 14:37

Post by TheParanoidOne » 2006-01-29 11:43

ratsalad wrote:So I should just put the "64*1024" value, then?
No, you should leave it alone and it will use the default value.

For that entire settings page, only put values in if you know what the value means and the ramifications of changing them.
The world is coming to an end. Please log off.

DC++ Guide | Words

ratsalad
Posts: 4
Joined: 2006-01-29 10:27

Post by ratsalad » 2006-01-29 17:41

So that would suggest that there's no way of getting the same results with the latest version of DC++, unless you understand the expert settings page, which i don't?
I mean, all i'd like is for me to be able to download a few files now and then, which i can't right now....
Will there be any help doc explaining the expert setting in the future?

Thanks again

ullner
Forum Moderator
Posts: 333
Joined: 2004-09-10 11:00
Contact:

Post by ullner » 2006-01-30 03:09

Since you know the standard value, try with a smaller value. Repeat until you notice that everything work fine.

Ninjai
Posts: 3
Joined: 2006-01-23 03:36

Post by Ninjai » 2006-01-30 03:59

Would you need to restart before DC++ applies these settings? Just wondering.

ullner
Forum Moderator
Posts: 333
Joined: 2004-09-10 11:00
Contact:

Post by ullner » 2006-01-30 04:29

No.

gemmell1
Posts: 26
Joined: 2003-11-24 09:52

Post by gemmell1 » 2006-02-15 04:10

TheParanoidOne wrote:
ratsalad wrote:So I should just put the "64*1024" value, then?
No, you should leave it alone and it will use the default value.

For that entire settings page, only put values in if you know what the value means and the ramifications of changing them.
Maybe anybody on ADSL should not bother upgrading to a client that the authors won't/can't tell us how to get to work correctly....

ullner
Forum Moderator
Posts: 333
Joined: 2004-09-10 11:00
Contact:

Post by ullner » 2006-02-15 04:22

ullner wrote:Since you know the standard value, try with a smaller value. Repeat until you notice that everything work fine.
There is no magical value.

gemmell1
Posts: 26
Joined: 2003-11-24 09:52

Post by gemmell1 » 2006-02-15 04:32

So what was did the checkbox do before?
what value did checking that box insert?

joakim_tosteberg
Forum Moderator
Posts: 587
Joined: 2003-05-07 02:38
Location: Sweden, Linkoping

Post by joakim_tosteberg » 2006-02-15 07:41

gemmell1 wrote:So what was did the checkbox do before?
what value did checking that box insert?
1024 according to the source.

gemmell1
Posts: 26
Joined: 2003-11-24 09:52

Post by gemmell1 » 2006-02-16 04:05

thank you :)

PseudonympH
Forum Moderator
Posts: 366
Joined: 2004-03-06 02:46

Post by PseudonympH » 2006-02-16 13:44

The behavior of the old "small send buffer" and the new "socket write buffer" are entirely different. The old one only limited it to writing a small amount at a time to the socket, the new one actually sets the TCP buffer. This means that the maximum throughput is the buffer size divided by the latency: set it to 1024 and it'll slow everything to a crawl.

Carraya
Posts: 112
Joined: 2004-09-21 11:43

Post by Carraya » 2006-02-17 02:17

I was just going to say that... If you have a high speed connecting 1024 is a very very small buffer and will probably do more harm than good...
<random funny comment>

dcjoheb
Posts: 1
Joined: 2006-02-11 10:39

Post by dcjoheb » 2006-02-18 15:25

I just hope that developers will Reinsert the "small send buffer" function...

Big Muscle
Posts: 72
Joined: 2004-01-23 14:45

Post by Big Muscle » 2006-02-18 16:31

dcjoheb wrote:I just hope that developers will Reinsert the "small send buffer" function...
but they won't. Read the second post in this topic.

gemmell1
Posts: 26
Joined: 2003-11-24 09:52

Post by gemmell1 » 2006-02-21 04:06

ok,

so we stick at 0.674 as we know we can use this program.

I just can't understand why we have to play with the settings ourselves, when it worked fine before.

Quattro
Posts: 166
Joined: 2006-01-11 09:23

Post by Quattro » 2006-02-21 04:45

you don't HAVE to play with them...
^^
just leaving them alone will have the same effect as not enabling the option in < 0.674
You can send a message around the world in 1/7 of a second; yet it may take several years to move a simple idea through a 1/4 inch of human skull.

gemmell1
Posts: 26
Joined: 2003-11-24 09:52

Post by gemmell1 » 2006-02-22 04:26

Quattro wrote:you don't HAVE to play with them...
^^
just leaving them alone will have the same effect as not enabling the option in < 0.674
FFS :evil: - have you read the thread? I want the option enabled so i can download properly !!!!!!!

Pothead
Posts: 223
Joined: 2005-01-15 06:55

Post by Pothead » 2006-02-22 05:56

gemmell1 wrote:FFS :evil: - have you read the thread? I want the option enabled so i can download properly !!!!!!!
FFS :evil: - have you read the thread? :roll:
It is enabled. 1024 is the default value, which is also the value used in older clients, when this option is enabled.

Code: Select all

#define SMALL_BUFFER_SIZE 1024
Or put even simpler.
The Default setting of newer clients, is EXACTLY the same as having that box ticked in older clients.

Carraya
Posts: 112
Joined: 2004-09-21 11:43

Post by Carraya » 2006-02-22 09:14

Actually pot it's not, because the buffer size is on recieve buffer as well as I remember, at least it's not exactly the same, but close...
<random funny comment>

GargoyleMT
DC++ Contributor
Posts: 3212
Joined: 2003-01-07 21:46
Location: .pa.us

Post by GargoyleMT » 2006-02-22 11:47

This deserves to be said again, since he's right:
PseudonympH wrote:The behavior of the old "small send buffer" and the new "socket write buffer" are entirely different. The old one only limited it to writing a small amount at a time to the socket, the new one actually sets the TCP buffer. This means that the maximum throughput is the buffer size divided by the latency: set it to 1024 and it'll slow everything to a crawl.

Carraya
Posts: 112
Joined: 2004-09-21 11:43

Post by Carraya » 2006-02-23 03:03

Weeeeeee I was right... :)
<random funny comment>

charly123
Posts: 7
Joined: 2006-06-12 18:20

Post by charly123 » 2006-06-12 20:01

Hi, I'm also having trouble downloading when uploading using the latest release (v0.691), too.
Using v0.674's 'Use Small Send Buffer' worked a dream, and my connection (512/256 ADSL) could max out either/or/and any way together, as needed.

However, with v0.691 the download speed decreases massively when uploading.
So reading from this thread, I gather I don't enter 1024 as the Socket Write Buffer, and that the Socket Write Buffer isn't the same as the previous Use Small Send Buffer feature anyway.

So what do I do?

Think I'm going to have to go back to v0.674. I don't really want to use a netlimiter as it would slow other people downloading from you, and you can get booted from certain hubs if you use one anyway.

Thanks

joakim_tosteberg
Forum Moderator
Posts: 587
Joined: 2003-05-07 02:38
Location: Sweden, Linkoping

Post by joakim_tosteberg » 2006-06-13 00:29

charly123 wrote: I don't really want to use a netlimiter as it would slow other people downloading from you
Which is more or less the same thing you do by using a small send buffer, the computer will just not be able to send that much data on the same time as it can with a larger one.
charly123 wrote: , and you can get booted from certain hubs if you use one anyway.
The hubs has no way of detecting that you are using netlimiter (if you do have a reasonable limit, usually somehere at 80% of you max upload speed).

charly123
Posts: 7
Joined: 2006-06-12 18:20

Post by charly123 » 2006-06-13 11:38

joakim_tosteberg wrote:
charly123 wrote: I don't really want to use a netlimiter as it would slow other people downloading from you
Which is more or less the same thing you do by using a small send buffer, the computer will just not be able to send that much data on the same time as it can with a larger one.
Thanks for the reply. Although, that's not what I experienced in physical, real-world use. Enabling the UseSmallSendBuffer option caused absolutely no drop in speed for outgoing (uploading) data. It could max out and upload at 30kbytes/s without any effort. The only discernable difference was a substantially increased download speed for me. To re-iterate, upload speed was unaffected, regardless of download traffic.

For now, I have reverted back to DC++ v0.674.

twilightking
Posts: 2
Joined: 2006-06-26 01:00

Post by twilightking » 2006-06-26 01:14

It's been a while since I've used DC and I just downloaded the latest version 0.691. Immediately I could tell there was a difference in download speed. I looked for the small send buffer option but couldn't find it so it so I did a google search and landed on this forum. My experience has been the same as charly123. Upload speed (~45k) was never affected with the small send buffer opton enabled. Now without it, my download speed has taken a sharp decline. Even surfing simple web pages takes a long time to load. Before I didn't mind sitting and idling in hubs to let other users download from me because there was no noticeable difference while surfing the net. I hope the programmers bring that option back or let us know how to optimally set the new options. Until then I'm also reverting back to v0.674

Quattro
Posts: 166
Joined: 2006-01-11 09:23

Post by Quattro » 2006-06-26 03:20

twilightking wrote:It's been a while since I've used DC and I just downloaded the latest version 0.691. Immediately I could tell there was a difference in download speed. I looked for the small send buffer option but couldn't find it so it so I did a google search and landed on this forum. My experience has been the same as charly123. Upload speed (~45k) was never affected with the small send buffer opton enabled. Now without it, my download speed has taken a sharp decline. Even surfing simple web pages takes a long time to load. Before I didn't mind sitting and idling in hubs to let other users download from me because there was no noticeable difference while surfing the net. I hope the programmers bring that option back or let us know how to optimally set the new options. Until then I'm also reverting back to v0.674
have you actually read this thread?
You can send a message around the world in 1/7 of a second; yet it may take several years to move a simple idea through a 1/4 inch of human skull.

twilightking
Posts: 2
Joined: 2006-06-26 01:00

Post by twilightking » 2006-07-05 02:24

Quattro wrote:
have you actually read this thread?
If you wanna be a smartass about it, yes I have read the thread. And no, your claim to leaving the settings alone did not have the same effect as (I'm assuming you meant enabling) enabling the old option. There was a substantial decrease in download bandwidth for me, so much that I couldn't even simply surf the internet. That is why I posted, hoping the programmers would hear more voices and make changes on future versions.

Asc3nti0n
Account Disabled Due to Policy
Posts: 15
Joined: 2003-01-24 21:42
Location: AUST
Contact:

Wow, so much arguement between the people who have no idea

Post by Asc3nti0n » 2006-07-05 03:02

Not saying I do know anything, but taking this from a completely third party stance on the issue (I'm on a LAN only internet networked hub of 200 odd users)

Qn to Moderators/Admin: Is there still a function inside the Settings of DC++ that replicates the effectiveness of "Small Send Buffer" or was it really a placebo function in previous clients?
coz MacsRBest

winthorpejones
Posts: 1
Joined: 2006-08-31 15:28

Re: Wow, so much arguement between the people who have no id

Post by winthorpejones » 2006-08-31 15:42

Asc3nti0n wrote: Qn to Moderators/Admin: Is there still a function inside the Settings of DC++ that replicates the effectiveness of "Small Send Buffer" or was it really a placebo function in previous clients?

This still seems to be an eminently reasonable question unanswered in the thread.

Using .674 with "use small send buffers" selected, everything works great. Using .691 lacking that option, my Web surfing slows to an absolute crawl. It's almost unusable. I have tested with .691 running and not running many times. It's definitely DC++ .691, and .674 exhibits the same problem if I don't have "use small send buffers" checked.

So the question remains: are there any settings to use in .691 that duplicate "use small send buffers" from .674?

Many thanks in advance for feedback. Seems it could be helpful to many.


GargoyleMT
DC++ Contributor
Posts: 3212
Joined: 2003-01-07 21:46
Location: .pa.us

Re: Wow, so much arguement between the people who have no id

Post by GargoyleMT » 2006-08-31 18:57

winthorpejones wrote:So the question remains
Some people believed that the socket read and write buffers would duplicate the option, but they do not (see earlier in the thread). There are no other features that would be even remotely similar.

toejam
Posts: 1
Joined: 2006-08-19 16:45

Post by toejam » 2006-09-15 16:47

We really need "use small send buffer" function back! After I installed 0.691 my DC++ uploads goto absolute max and my Internet surfing suffers really bad. I can download pretty much normally in DC++ though.
I tried to set "Socket write buffer" to 4096 and some other values but the uploads just went really low or died completely.
Since no-one wants to inform us 0.691 users about what to insert in "Socket write buffer" and only suggest leaving that untouched there is only one option left, go back to 0.674.

ADSL 1024/512

Xan1977
Forum Moderator
Posts: 627
Joined: 2003-06-05 20:15

Post by Xan1977 » 2006-09-15 17:33

toejam wrote:go back to 0.674.
The other choice is to manage your bandwidth with some third party application. Uploads saturating your internet connection is not a problem that is unique to DC++.

dj christian
Posts: 1
Joined: 2006-07-26 22:00
Location: Sweden

Post by dj christian » 2007-01-02 08:15

TheParanoidOne wrote: Note that in 0.6811 this item is now a configurable value. See Settings -> Advanced -> Experts Only -> Socket write buffer.
What value should i put in it? Im using a 2mbit/500kb line.

My downloads tends to slow down a lot in newer versions so the only option is to go back to 0.674 with "send small buffer" option enabled.

The newer versions are very bad :(.

Ive read through the whole thread atleast 10 times.

Locked