TCP/IP Limit (Dreaded EventID 4226)

  • Thread starter Thread starter James
  • Start date Start date
J

James

Back when XP SP2 came out, this was a significant problem for a number of
people. I need to either dramatically increase or remove this artificial
cap on the number of concurrent TCP/IP connections I am "allowed". The
patch file that fixed this under XP will run under Vista, but encounters
problems and fail. I am extremely tentative of using a Hex Editor to mess
with this as there doesn't seem to be a lot of documentation re: Vista's
version of this fle. Anyone else hit this problem again, and/or fixed it?

Thanks
 
Yeah, I hit the artificial wall last night. This is insane. Their foolish
method of "slowing down worms" actually cost me money last night. An
inconsequential amount of it, granted...but it did. I can't stand this
crap. If you're going to put in a moronic setting, at least offer a
straightforward opt-out.
 
Sounds like you are trying to run a desktop system as a server.
If you need a server buy a server OS.
 
James said:
Yeah, I hit the artificial wall last night. This is insane. Their
foolish method of "slowing down worms" actually cost me money last
night. An inconsequential amount of it, granted...but it did. I
can't stand this crap. If you're going to put in a moronic setting,
at least offer a straightforward opt-out.

Just curious - how did it cost you money?
 
I'm not in the slightest. My machine is isolated and not even on a local
network. I'm an application developer who uses peer to peer softwore
occasionally for obscure files and can have a lot of browsers or other
applications that use TCP/IP ports, open.
 
Just playing low limit poker in my spare time last evening, online.
Typically play at tables that don't offer disconnection protection...for
other reasons. I know that my internet connection was stable and I would go
to open a browser window that would display a video (i.e. YouTube or
something) and it would start queueing my traffic and disconnecting me
mid-hand. Every time this happened, I'd check the event log and...sure
enough. While I cannot blame any "losses" incurred after the first incident
on anyone other than myself, it's an intentional design flaw...not the
remnants of a beta product. The point is...this limitation can completely
screw people that actually use their computer to any semblance of its
potential.
 
Gary,

Sounds like the OP had a legit question for desktop o.s.'s
That limitation is no longer an excuse to make it the providence of
server's only;
in fact, that "imposed" limitation has little or no effect on MSFT's
"supposed" reason of denial of worm spreading; to begin with.
It's a PITA, out and out.
An excuse, is more like it; from MSFT, to force a person to server o.s.'s
and in fact, many people now use their home pc for more than a few connects.
Home office people; business people, your avrg pc user, all can now hit that
"wall" with little difficulty.
It's an outdated excuse for limitations-and a effort for a money grab.

Try it yourself, multitask a few network intensive apps at once.
You'll see what it does.

Jeff
 
James said:
Just playing low limit poker in my spare time last evening, online.
Typically play at tables that don't offer disconnection
protection...for other reasons. I know that my internet connection
was stable and I would go to open a browser window that would display
a video (i.e. YouTube or something) and it would start queueing my
traffic and disconnecting me mid-hand. Every time this happened, I'd
check the event log and...sure enough. While I cannot blame any
"losses" incurred after the first incident on anyone other than
myself, it's an intentional design flaw...not the remnants of a beta
product. The point is...this limitation can completely screw people
that actually use their computer to any semblance of its potential.

I'm not sure what to say about this - it "shouldn't" happen in the scenario
you describe and I've made far 'higher' use of the Internet than what you
describe without seeing this issue.

I only see the problem with connection limiting when running peer to peer
software, because of a peculiar situation this causes, and I would not
expect to see this error in 'normal use'. I'm not sure if you understand
when these events occur exactly (forgive me if that isn't the case) so I'm
going to outline a few things here.

Firstly, this error is not related to how many network connections you have
open on your LAN or to the Internet. It isn't an error you would expect to
be caused by opening too many web browser windows, or by having too many
'ports' connected on your machine.

Instead, this error refers to ports held in a particular state (usually
shown as "SYN_Waiting" if you open a command line and type 'netstat -a')
which is generally referred to as a 'half-open' port. This is part of the
normal connection 'dialogue' of any TCP/IP connection (e.g. any connection
over the Internet, for a start) but due to the way many peer to peer
networking systems (including network worms!) operate they open a larger
number of these connections at any one time than you would normally expect.
TCP/IP Limiting is designed to put a break on the amount of connections held
in this state at any one time, and the number allowed is set a little low
but should actually be ample for the majority of workstation use.

So if you're not running a peer to peer application, it is rather odd to see
this message appear. I'd look very carefully at the applications you're
running and the ports they are opening if you are not running anything you
might expect to cause this problem.

As for it being a 'fault' in the operating system, while I feel that
Microsoft's reasoning for adopting this approach is an example of
astonishing stupidity, it is difficult to really call it a "fault" in the
OS - it's designed to work in a certain way and achieves that design. Of
course, lots of aspects of Vista's design have been causing me to say things
like this lately, but then this is Microsoft's product and they've made
their choices. We can vote with our wallets if we dislike those choices.

Even on a Mac OSX laptop which does NOT use this rate-limiting approach I
have seen a problem with Internet performance with peer to peer networking
programs due to the large amount of connections they attempt to open at
once. I find a good approach is to limit the peer to peer program itself -
perhaps this is something you might consider.
 
LOL, I run several applications, which I've written, locally that are
listening for inbound traffic on varying ports. I'd file that under fair
use. Running a P2P application just makes it all the more obvious. There
is no question as to what's happening, my apologies if I wasn't clear. I'm
hitting the limit that Vista is imposing without a straightforward,
clear-cut method of getting around it. Period.
 
James said:
LOL, I run several applications, which I've written, locally that are
listening for inbound traffic on varying ports. I'd file that under
fair use.

Oh I never doubted that you were using the system 'fairly'.
Running a P2P application just makes it all the more
obvious. There is no question as to what's happening, my apologies
if I wasn't clear. I'm hitting the limit that Vista is imposing
without a straightforward, clear-cut method of getting around it. Period.

Yep I understand that. I use Mac OSX for anything like that now because it
is so much more reliable at that kind of work.
 
Actually, you can have any number of connections you want to have as long as
they all respond.

How does this affect P2P?

Lets say your client sends SYN to 50 other machines "at once (or in a short
period of time)". machines 1, 23, 40, and 45 respond, while everything else
doesn't. You now have a near immediate 4 connections.

With the TCP/IP half open limitation, it would send SYN to the first 10
hosts. Host 1 responds, so it sends now to host 11. The other 9 time out, it
now sends to hosts 12-21. 11 times out, it now sends to 23, which responds
,and then begins on 24.

So what happens is a slight delay in the connection process as it goes
through the lists of hosts. This has no effect on the transfer speed between
Hosts 1, 23, 40, and 45. Once that connection is established that's pretty
much it for the TCP/IP limiter. If all 50 responded, you could connect to
all 50 hosts without any limitation at all.

So the real issue on P2P networks with regards to this "problem" are the
people who don't understand how to configure their internet connections, or
the people who configure them with the intent of not allowing incoming
connections.

John
 
Back
Top