- some PL -

[DW_Wraith]

Is there a way to tell your computer Not to use a certain hop?

[zeus]

Is there a way to tell your computer Not to use a certain hop?

For us mere mortals.. no.

While there are multiple pathways on the Internet, routing protocols determine the most "efficient" way to get from point A to point B. As an end user, you really have no say in that. All you can control what is on your network.

It is just like any other utility. You can't really control how electricity or water gets into your house. You can just control what to do with it when it is there.

[DW_Wraith]

Thanks for the info Zeus.

Someone fix the internet! It's broken!!!

lol

:cheers:

[DW_Snaggle]

Take this as you wish, but don't take the intermediate node P/L numbers too seriously. The most important numbers for P/L are the ones for the destination host. The intermediate nodes are not *required* to respond to ICMP Echo/Ping requests. While their numbers can be good to determine where delays may be, any P/L *at* that node may just be them ignoring the requests.

So if you see that the destination is only seeing 1% P/L while the intermediate node(s) have 10% it may just be that they are ignoring you. While there may be a strong correlation between the numbers you see, it could also be that some non-intelligent piece of equipment like a dumb switch (which will not show up in a traceroute) may be having an issue.

Just my $0.02

Snaggle.

[DW_Snaggle]

Here's a sample of my trace. First two hops removed for privacy reasons.

Code: Select all

                                           Loss%   Snt   Last   Avg  Best  Wrst StDev
  3. 3.ge-2-1-0.gw.sr.sonic.net        0.0%   100    7.2   7.8   6.7  15.1   1.1
  4. 0.ge-0-1-0.gw4.200p-sf.sonic      0.0%   100    8.5   9.2   8.2  18.4   1.3
  5. ge-6-2-0.408.ar4.sfo1.gblx.n      0.0%   100    9.0  13.1   8.5 113.5  15.7
  6. so6-0-0-2488M.ar1.SJC2.gblx.      0.0%   100   13.7  18.6  10.8 155.3  21.9
  7. so-7-2-0.mpr2.sjc7.us.above.      0.0%   100   10.7  17.0  10.7  65.5  11.4
  8. so-5-0-0.mpr3.sjc2.us.above.      0.0%   100   15.1  17.7  10.8  61.5  10.8
  9. so-4-1-0.cr1.ord2.us.above.n      0.0%   100   64.7  69.9  60.1 131.5   9.1
 10. so-3-0-0.mpr2.ord7.us.above.      0.0%   100   65.6  67.6  60.1  91.1   6.2
 11. so-0-0-0.mpr1.ord7.us.above.      0.0%   100   74.3  68.8  59.7 102.9   6.7
 12. so-1-0-0.cr2.ord2.us.above.n      0.0%   100   66.3  70.6  60.2 146.5  11.8
 13. ge0.0013.core2.ord.servernap      0.0%   100   72.6  68.3  59.5  81.7   5.9
 14. gc0.aggr1.chi.servernap.com       0.0%   100   65.9  68.1  60.6  88.9   5.6
 15. 69.65.1.18                        1.0%   100   62.9  67.9  59.5  98.8   6.1

As you can see, my big problem is a 50ms delay in the middle of above.net. And it only gets worse as the traffic gets heavier. :sad5:

Also, notice that the only P/L for this particular trace was at the server itself. I'll have to check my P/L numbers more closely 'in game', I guess.

BTW, thanks Toxic for all your hard work. And keep up the good fight. :cheers:

Snaggle.[/code]

[Namu]

Is there a way to tell your computer Not to use a certain hop?

For us mere mortals.. no.

While there are multiple pathways on the Internet, routing protocols determine the most "efficient" way to get from point A to point B. As an end user, you really have no say in that. All you can control what is on your network.

It is just like any other utility. You can't really control how electricity or water gets into your house. You can just control what to do with it when it is there.

I have one way of doing it, depends on the routers routing protocol. Find that hop that is ugly and just spam it to death ( i have set this up with 50+ pings going to the same router) it becomes overwhelmed and will advise its neighbor routers to not send any more stuff. Though this depends on the type of protocol that it is using :). My firewall is being a bitch with this WinMTR< but I will post my tracert soon.

[DW_ToxicWaste]

Take this as you wish, but don't take the intermediate node P/L numbers too seriously. The most important numbers for P/L are the ones for the destination host. The intermediate nodes are not *required* to respond to ICMP Echo/Ping requests. While their numbers can be good to determine where delays may be, any P/L *at* that node may just be them ignoring the requests.

So if you see that the destination is only seeing 1% P/L while the intermediate node(s) have 10% it may just be that they are ignoring you. While there may be a strong correlation between the numbers you see, it could also be that some non-intelligent piece of equipment like a dumb switch (which will not show up in a traceroute) may be having an issue.

Just my $0.02

Snaggle.

True, but only to a certain extent.

If you were being ignored, you would have 100% packet loss. HOWEVER, in most cases routers give ICMP traffic the lowest possible priority. Packet loss at these routers, while maybe not indicitive of true packet loss, imply that these routers are a little too busy doing other things, and maybe have too much of a workload, or have a hardware issue.

The issues appear to be on level3 routers. I can request that they move my outbound traffic to savvis. That won't help you on the incoming however, as you are stuck with the route you're given.

I'm hoping the Dallas server will be up this week. I'm putting the server together this weekend. If you want to get an idea of how good tour routing will be there you can tracert http://www.colo4dallas.com.

Rich (TW)

[DW_Snaggle]

Heh! I tried that. Guess who I have to transit through to get from GlobalCrossing to http://www.colo4dallas.com ..... ???

Yep! You guessed it: level3.net :cussing: I can't win. And I also have a 50ms hop in there between two of the level3 nodes!!!!! Conspiracy? I think not! Just dumb f*cking luck. :sad3:

But I know that many others will benefit so it's still a Good Thing(TM).

Good luck with the new server build. BTW, are you going with AMD or Core2 ??

Snaggle.

[DW_e_aLpHa]

Well .. The isue is that we can't do anything about it..(about the best path for our conection to the server) and Everything is just reflected on the game Server at Playing time.. and at least for me thats enough to know something is not good.. neverminding any tracerouts results or if the pl is 100% real or what pass is it comming from . .. The thing is I join the server I get a high ping a little pl .. and client side my fps get way low and then thats just not fun .. for some reazon my fps drop when the ping goes up .. I may really ignore if one thing has to do with another.. :dontknow:

made my tracert to
http://www.colo4dallas.com

and i get this pl on this pass ... Know that Im in San Antonio

ae-5-5.ebr1.Dallas1.Level3.net - 57

and 0 % at the server

[DW_ToxicWaste]

Heh! I tried that. Guess who I have to transit through to get from GlobalCrossing to http://www.colo4dallas.com ..... ???

Yep! You guessed it: level3.net :cussing: I can't win. And I also have a 50ms hop in there between two of the level3 nodes!!!!! Conspiracy? I think not! Just dumb f*cking luck. :sad3:

But I know that many others will benefit so it's still a Good Thing(TM).

Good luck with the new server build. BTW, are you going with AMD or Core2 ??

Snaggle.

Sticking with AMD for the time being. I procured some Opteron 256s for a dual CPU box.

Woodcrests look damn impressive so far, but I think I'll wait a little while until I get more real world feedback.

Rich (TW)