Found a way to increase BTV Link performance

[cmcquistion]

I think I found a way to increase BTV Link performance, in terms of general responsiveness. (Also Beyond Media.)

I need some other people to try this out and see if really helps or if it's just a placebo effect.

I have mid-high end hardware for my BTV Server and Link machine, as well as a gigabit network, but the general responsiveness of BTV Link and Beyond Media Music seemed to lack the speed that I expected. I started wondering if it was limitations imposed by the Windows OS that was slowing things down. I went googling around to find something to increase my network performance and I came across a website that explained that Windows XP SP2 put a cap on the number of concurrent network connections that can be open (10), in an attempt to slow down malicious software (worms) propagation over a network. People using Bittorrent were frustrated by their reduced P2P performance because of this imposed limit, so someone wrote a little program that will let you go in and change the limit to some other amount. By default, the tool changes the limit from 10 files to 50. (I tried the default setting of 50.) The Bittorrent folks claim that it increases their performance, since it allows more open connections, so I thought it might help me, too.

I've tried it on my BTV Server and BTV Link machine and it definitely "appears" to have helped responsiveness in BTV Link and Beyond Media. Scanning my music library, as an example, seems to be running much faster than before (I have about 30 GB of music.) The DVD Library seems to run quicker, also (all my DVD images and Music are on the BTV server, so the Link machine has to grab files over the network.) BTV Link seems to navigate around a bit quicker, also, with less "hangs".

I would really appreciate it if some other folks would try this out and let me know what you think. Maybe this doesn't really help, or maybe it does. If it does, I'm sure a lot of BTV/BM users would really appreciate a little free performance boost.

The tool is called "EvID4226Patch223d-en.zip". I don't have the original website I downloaded it from, but if you Google for it, it is all over the web.

Thanks.

[guncam_man]

If you are always using < 10 cuncurrent connections, increasing that number isn't going to help anything.

Also, my XP can have a lot more than 10 connections running at the same time, so that kind of sounds like BS.

[guncam_man]

Huh, I'm checking this out, and it seems Mircrosoft is still as farking clueless as ever. It seems they have limited the creation of sockets to 10/second.


Moronically thinking this will hinder the spread of worms.

Idiots.

But anyway, the limit is that it doesn't create more than 10 TCP connections per second. It is unlikely this limitation will hurt small network operations - but that is theoretical. How it is actually implemented might be a different story.

[cmcquistion]

Here's an article (three pages) that has a little explanation:

http://blog.davidkaspar.com/archives...nt-id-4226.php

[guncam_man]

BTW, thanks for this timely information. I had no idea XP had such a limitation, and I am currently working on a project that requires my XP machine to have the ability to create a lot of half open connections. This will be quite useful.

[JimF]

Here is another source:
http://www.speedguide.net/read_articles.php?id=1497

[guncam_man]

Originally Posted by JimF

Here is another source:
http://www.speedguide.net/read_articles.php?id=1497

From that site:

With the new implementation, if a P2P or some other network program attempts to connect to 100 sites at once, it would only be able to connect to 10 per second, so it would take it 10 seconds to reach all 100.

Me thinks MS isn't trying to hinder worms as much as they are trying to hinder P2P.

[karhill]

There's some confusion on this SP2 and 10 connection limit, so let me clarify based on what I know:

1) Microsoft has always, for a long time, had a limit on the number of windows-networking (i.e. SMB-based) connections that their workstation products can support. This is to prevent people from using the workstation product as a file server. Microsoft wants people to buy Windows server for their file servers. This limit was (and is) 10 in Windows XP Pro, 5 in Windows XP home. This limit has nothing to do with the number of TCP connections you can make, but people get the two confused sometimes.

2) SP2 for Windows XP puts a limit on the number of simultaneous outbound TCP connection ATTEMPTS that can be outstanding. This limit is a hard-wired value of 10. This is NOT a limit on the number of TCP connections that can be connected at one time (which is potentially unlimited, depending on a registry value). It is a limit on the number of connection attempts (i.e. outstanding requests to open a new connection that have not been ack'd). Likewise this is not a limit of 10 connections per second...it is an upper bound on the number of TCP connection attempts that can be outstanding without having progressed to the next step in TCP connection establishment prorcess. TCP connections that have been successfully established are in no way affected by this limit.

To quote the SP2 documentation from »www.microsoft.com/technet/prodte···1120120:

Why is this change important? What threats does it help mitigate?

This change helps to limit the speed at which malicious programs, such as viruses and worms, spread to uninfected computers. Malicious programs often attempt to reach uninfected computers by opening simultaneous connections to random IP addresses. Most of these random addresses result in a failed connection, so a burst of such activity on a computer is a signal that it may have been infected by a malicious program.

In other words, this was a change that was made to increase security and not to limit a host's capabilities. It's designed to detect behavior that viruses have and limit the danger.

3) Some programs, like emule and some other P2P programs, have some behavior that is somewhat akin to a virus: they attempt to open connections to many many hosts and keep those connection ATTEMPTS going, even when the remote host doesn't respond (hoping to eventually be able to connect and pull some p2p content).

4) There are hacked versions of tcpip.sys that re-write the bytes that are the hard-wired connection attempts limit. These don't replace the entire tcpip stack, they just write-over the bytes where this value is coded in. See »www.lvllord.de/?lang=en&url=4226patch/faq, for example.

5) I suspect that in the future Microsoft will make this value set-able via the registry, but right now it is not. I suspect they feared that viruses would just change the registry value, negating the security of this feature. But if a virus is running as admin, it can just replace tcpip.sys, achieving the same effect.

All in all, this is a good thing and if the changes negatively effect you, rather than abandon SP2, you might consider a hacked tcpip.sys. The only real problem I have with this change is the failure to make it configurable via the registry. For the average user, having more than 10 outstanding connection attempts should not happen...it's a sign that something is wrong. The only way this hack would improve BTV Link/Server responsiveness is if the BTV Link/Server communications code was poorly designed or implemented.

[JimF]

Originally Posted by karhill

2) SP2 for Windows XP puts a limit on the number of simultaneous outbound TCP connection ATTEMPTS that can be outstanding. This limit is a hard-wired value of 10. This is NOT a limit on the number of TCP connections that can be connected at one time (which is potentially unlimited, depending on a registry value). It is a limit on the number of connection attempts (i.e. outstanding requests to open a new connection that have not been ack'd). Likewise this is not a limit of 10 connections per second...it is an upper bound on the number of TCP connection attempts that can be outstanding without having progressed to the next step in TCP connection establishment prorcess. TCP connections that have been successfully established are in no way affected by this limit.

Here is how Microsoft states the operation, in case that helps:

"The TCP/IP stack now limits the number of simultaneous incomplete outbound TCP connection attempts. After the limit has been reached, subsequent connection attempts are put in a queue and will be resolved at a fixed rate."

So you can fill the glass up from the top as the water leaks out from below, but you can't over-fill the glass.

[karhill]

Originally Posted by JimF

Here is how Microsoft states the operation, in case that helps:

"The TCP/IP stack now limits the number of simultaneous incomplete outbound TCP connection attempts. After the limit has been reached, subsequent connection attempts are put in a queue and will be resolved at a fixed rate."

So you can fill the glass up from the top as the water leaks out from below, but you can't over-fill the glass.

Yes, that's a good analogy. And it's important to re-stress that this is NOT a limit on the number of simultaneous connections that can exist, it is a limit on the number of connection establishment attempts that can simultaneously be outstanding.

And seriously, this should NOT be an issue in BTV Link/Server communications performance unless the BTV Link/Server communications is poorly designed or implemented.

I would be interested to know if this limit really does impact BTV Link/Server communications, but not interested enough to bother running an experiment myself.

[cmcquistion]

Originally Posted by karhill

And seriously, this should NOT be an issue in BTV Link/Server communications performance unless the BTV Link/Server communications is poorly designed or implemented.

I understand that this SHOULDN'T make a difference if BTV and LINK are well designed.

I tried it, however, and it DOES appear to make a difference, which is why I started this thread. I'm hoping that some other BTV users will actually TRY IT and leave some comments about their EXPERIENCE with using it.

[sainswor99]

after reading the articles, I thought "why bother?" I only saw a couple of the errors in my error logs over a period of weeks, so I didn't really think that it would make that big of a difference. I applied the patch, however, and I do see a noticeable difference in my Beyond Link. I also use wizd for my linkplayer, and I see some speed improvements there.

I don't know why, but it does seem to make things faster.

Just my .02

Stu

[cmcquistion]

Thank you!

I was really hoping some other people would just try it and give an assessment of their experience.