Testing build #133

[rejetto]

download @ www.dovedove.it/hfs/hfs133.exe

what's new
+ command line parameter -b <size> to control ICS buffer
+ command line parameter -B <size> to control socket buffer
* speed enhancer improved but now optional (menu -> other options -> Experimental high speed handling)
- possible fix for hanging disconnected connections introduced in build #131
- reduced flickering problem in connection header bar


a few words
With this version, those people who had problems with reduced performance on very high upload speed internet connections, can play with 2 important buffer sizes, and report their experience, so i can eventually integrate their findings in HFS.
The "experimental high speed handling" is the current way to automatically set those buffers. You can compare your results with this. The internal algo to set the buffers is the same for both, and it is very simple: every second set the buffer to a fourth of the data sent in the last second.
Being this setting just experimental, it is not saved in options, and it is disabled at every restart.

Moreover, those problems in progress updating should be solved, since i think they were related to the high speed handler, and it is disabled by default now. Since it was improved, you can try switching it ON, and see if the problem reappears. I hope it doesn't.

[Foggy]

Sounds interesting. Ill test my upload speed and see if I can get close to my maximum speed.


I got 2 people to download a 10mb file

person1
Default settings 80kBps +/- 10kB
Speed enhancer 30-100kBps

person2
Default settings 60kBps +/- 5kB
Speed enhancer 20-110kBps

All speeds were reported by HFS.
My modem is currently synced at 974000Bits, which should give me 100-110kBps speeds.

[Foggy]

Rejetto you haven't got any info about the 2 buffers that we can manually set, do you? I couldn't find anything in Google.

[rejetto]

ICS buffer is the internal buffer used by the socket lib you can find at www.overbyte.be

socket buffer is the buffer used by Windows through the call setsockopt(handle, SOL_SOCKET , SO_SNDBUF...

[GTOnitsuka]

Why isn't it avalaible to the update feature?

[rejetto]

sorry
done

[OGWATERMELON]

Tested the speed with the enhanced feature at 3 different locations.  I downloaded the same file which was 10.61 MB.
My results were pretty much the same.  My connection at home according to MySpeed internet connection speed test.  I have a download speed of 6.74 Mbps and a upload speed of 479 kbps. 
My normal average download speed with HFS is around 70kbps.

Location 1:
The download started out at 267 kbps and slowly dropped down to 58kbps.

Location 2:
The download started at 241kbps and slowly dropped down to 46kbps.

Location 3:
The download started at 265kbps and dropped down to 63kbps.

I will do some more testing and post any results I get.

[rejetto]

Is the "slowly dropping" only happening with the experimental thing?

You used to get 70kbps with previous HFS builds? That's about a 56k modem!

[Foggy]

ICS buffer is the internal buffer used by the socket lib you can find at www.overbyte.be

socket buffer is the buffer used by Windows through the call setsockopt(handle, SOL_SOCKET , SO_SNDBUF...

thanks, Ill have a look at that site a bit later and see if I can improve my upload speed.

Is the "slowly dropping" only happening with the experimental thing?

You used to get 70kbps with previous HFS builds? That's about a 56k modem!

I think they are just getting kb confused with kB?

Tested the speed with the enhanced feature at 3 different locations.  I downloaded the same file which was 10.61 MB.
My results were pretty much the same.  My connection at home according to MySpeed internet connection speed test.  I have a download speed of 6.74 Mbps and a upload speed of 479 kbps. 
My normal average download speed with HFS is around 70kbps.

Location 1:
The download started out at 267 kbps and slowly dropped down to 58kbps.

Location 2:
The download started at 241kbps and slowly dropped down to 46kbps.

Location 3:
The download started at 265kbps and dropped down to 63kbps.

I will do some more testing and post any results I get.

 With an upload speed of approximately 60kB(479kb) up why are you starting out with speeds of 240kB+ before dropping down to the maximum that your connection can handle?

[KalleB]

There might me some problems with download limits. One user was downloading two files from me, simultaneous downloads set to 20, from same address 7 and he got the "download limit reached" notification. The downloads were not in sections, they were normal downloads. I will try and reproduce when I have time... I think this might be related to that disconnection problem.

Update: I was not able to reproduce this.. At least not when downloading from 127.0.0.1

[OGWATERMELON]

First I did put the wrong terminology, it should of been KBps not kbps.  But as for the readings in the 200KBps range, maybe that is a fault reading by the download manager in IE 7.  But all of the downloads do start out that way.  I have tested it a few more time and get the same results.  Part of this could be due to my cable ISP connection is not always the same speed. It will vary one or two megs  for download speed.

[psyxakias]

Rejetto,

The "Experimental high speed handling" doesn't seem to work to me. Since I upgraded to newer build than #131, I feel the performance downgraded again.

Do you remember which ICS/socket buffer settings #131 build was using? So I can test values and compare with them. Even if you don't, I will do some tests soon.

Special Thanks )

[rejetto]

it's not a problem, i keep a backup of every build.
but let's discuss this speed thing in a specific topic.
nice to see you again psyxakias, i need to fix this thing before i publish the new stable.

[psyxakias]

Heya rejetto,

Here are my results:

[Without experimental]
1 connection: 46 KB/sec
2 connections: 90 KB/sec
5 connections: 220 KB/sec
10 connections: 420 KB/sec

[With experimental]
1 connection: 45-400 KB/sec (quite unstable, average: 153 KB/sec)
2 connections: 88-520 KB/sec (quite unstable, average: 287 KB/sec)
5 connections: 200-850 KB/sec (quite unstable, average: 516 KB/sec)
10 connections: 350-950 KB/sec (quite unstable, average: 680 KB/sec)

------------------

[-b 1024]
1 connection (without experimental): 45-46 KB/sec
1 connection (with experimental): 50-400 KB/sec at first, then goes down to 4-5 KB/sec

[-b 2048]
1 connection (without experimental): 45-46 KB/sec
1 connection (with experimental): 45-920 KB/sec at first, then goes down to 20-25 KB/sec

[-b 65536]
1 connection (without experimental): 45-47 KB/sec at first, then 170-260 KB/sec (unstable #1)
1 connection (with experimental): 50-1000 KB/sec (unstable #2)

[-b 131072]
1 connection (without experimental): 45-47 KB/sec at first, then 10-300 KB/sec (unstable #1)
1 connection (with experimental): 60-300 KB/sec (unstable #2)

[-b 1048576]
1 connection (without experimental): 45-46 KB/sec at first, then 170-380 KB/sec (unstable #1)
1 connection (with experimental): 40-150 KB/sec (unstable #2)

------------------

[-B 1024]
1 connection (without experimental): 8-9 KB/sec
1 connection (with experimental): 8-9 KB/sec

[-B 2048]
1 connection (without experimental): 15-17 KB/sec
1 connection (with experimental): 15-17 KB/sec

[-B 65536]
1 connection (without experimental): 50-330 KB/sec (unstable #2)
1 connection (with experimental): 50-350 KB/sec (unstable #2)

[-B 131072]
1 connection (without experimental): 60-700 KB/sec (unstable #2)
1 connection (with experimental): 60-690 KB/sec (unstable #2)
2 connections (without experimental): 100-1050 KB/sec (unstable #2)
5 connections (without experimental): 92-1810 KB/sec (unstable #2 -- it max'ed my ADSL's downstream at times)

Unstable #1: The speed is quite unstable, up/down all the time
Unstable #2: The download starts from slow speed, speed increases with an additional 20-25 KB/sec every second, then suddenly drops. Then starts from slow speed again, speed increases the same way as before and then it suddenly drops.

Overall, whenever I use "experimental high speed handling" (with no command-line parameters) OR use the -B parameter, Unstable #2 occurs. While when I use -B parameter, experimental option doesnt make any difference whether it's on or off. About the increasing and then drop issue, although I have no technical knowledge about how exactly it works, it makes me feel that it fills some buffer and once it's full, speed drops.

I have to admit that I'm a bit confused with the -b and -B, like which settings are better and which ones are not. Especially on "Unstable #2", it may take up to 1-2 minutes to complete the increase/drop, so taking a 60-secs average won't help me choose.

If you need any further tests or give me some specific requests to make tests (like specific -b/-B or something), feel free to ask.

Thanks.

PS: I've done all these tests with #137
PS2: I'm thinking of doing some combination of -b and -B - I just realized how it works, I should be setting both -b and -B and dont use experimental so it won't auto-change their values.

[Foggy]

Very interesting I look forward to looking at any other tests you do.