dietro
join:2009-05-29
Jacksonville, FL
1 edit | Offtopic: World of warcraft latency
I realize this is not the correct forum for this, since it is not an issue with Comcast so I understand if it is moved. I'm only posting it here because I've come to value the opinion of several of the regulars of this forum and I am hoping, if they have the time, that they will read thru this and point out any incorrect assumptions that I've made before I make my case on the official game forums.
I've been suffering from unusually high and unusually fluctuating latency latey in World of warcraft. I've posted regarding this issue before and it was insight from that post in this forum that lead me to use a proxy server. While the proxy server has worked fairly well to lower and stabalize my connection to the world server I play on, when I enter any instanced content my latency starts fluctuating constantly. One input for example, like a spell cast, returns in what I would consider a normal amount of time but the very next moment the subsequent spell cast is delayed by up to one second.
I've come to suspect the issue is packet delay variation within the games network.
When I use the netstat -n command while inside of any instanced content I do not see a new IP listed, I still see the IP of the world server I play on. This makes me think that I'm connecting thru the world server to the instance server; in a sense using the world server as a proxy. If thats the case than it seems as though while playing on the world server I pass thru the jittery network or router twice, in and out. But then when I connect to an instance server I pass thru the network (or the jittery routers within the network) 4 times. (1) in to the world server (2) in to the instance server (3) out back to the world server and (4) then back out to my system. If that is right than it certainly would explain why the latency fluctuates more often and to higher numbers within instances than it does when connected to the world server.
These are some screenshots of a 3d tracerts taken that show some strange activity near the last hop. I dont know if this is a visible sign of the packet delay variation that I suspect is playing havok on my latency or if it is just a deprioritized response to pings.
»s22.photobucket.com/albums/b323/···pike.jpg
»s22.photobucket.com/albums/b323/···ike2.jpg
»s22.photobucket.com/albums/b323/···kes3.jpg
If you've taken the time to read this and have an understanding of the subject please point out any misconceptions I might have or any failures in my logic. I'm not a network admin so I dont want to start championing this issue until I've checked with some real network admins.
|
|
EG
The wings of love
Premium
join:2006-11-18
Union, NJ | Can you post traces to those servers using the Windows Tracert tool please ? |
|
dietro
join:2009-05-29
Jacksonville, FL
| Tracing route to 206.17.111.6 over a maximum of 30 hops
1 2 ms 1 ms 1 ms 192.168.1.1
2 * * * Request timed out.
3 8 ms 9 ms 9 ms te-9-2-sr02.southside.fl.jacksvil.comcast.net [68.86.171.197]
4 9 ms 8 ms 8 ms te-1-1-sr01.southside.fl.jacksvil.comcast.net [68.86.168.245]
5 8 ms 9 ms 11 ms te-7-4-ar01.southsiderdc.fl.jacksvil.comcast.net [68.86.168.113]
6 17 ms 15 ms 15 ms te-0-2-0-5-ar03.pompanobeach.fl.pompano.comcast.net [68.85.229.229]
7 18 ms 17 ms 16 ms pos-0-7-0-0-ar03.northdade.fl.pompano.comcast.net [68.86.164.5]
8 18 ms 18 ms 18 ms pos-0-5-0-0-cr01.miami.fl.ibone.comcast.net [68.86.91.81]
9 30 ms 32 ms 30 ms pos-2-2-0-0-cr01.atlanta.ga.ibone.comcast.net [68.86.86.66]
10 52 ms 52 ms 52 ms pos-1-10-0-0-cr01.dallas.tx.ibone.comcast.net [68.86.86.129]
11 54 ms 53 ms 53 ms Vlan549.icore2.DTX-Dallas.as6453.net [206.82.142.5]
12 53 ms 53 ms 54 ms Vlan6.icore1.DTX-Dallas.as6453.net [209.58.47.13]
13 53 ms 52 ms 73 ms 192.205.35.57
14 79 ms 80 ms 107 ms cr2.dlstx.ip.att.net [12.122.138.110]
15 80 ms 78 ms 79 ms cr1.attga.ip.att.net [12.122.28.173]
16 89 ms 80 ms 81 ms cr2.wswdc.ip.att.net [12.122.1.174]
17 81 ms 80 ms 79 ms cr2.n54ny.ip.att.net [12.122.3.37]
18 79 ms 81 ms 80 ms cr1.cb1ma.ip.att.net [12.122.31.126]
19 78 ms 83 ms 78 ms gar5.cb1ma.ip.att.net [12.122.145.29]
20 130 ms 199 ms 199 ms 12-122-254-14.attens.net [12.122.254.14]
21 80 ms 81 ms 80 ms mdf001c7613r0003-gig-10-1.bos1.attens.net [12.130.0.170]
22 * * * Request timed out.
23 * * * Request timed out.
Here's a tracert and another question. Am I correct in thinking that from hop 14 to around 20 I'm going thru an mpls cloud? And then does that make hop 14 a label edge router? |
|
koshoka
join:2006-12-01
Pottsville, PA | reply to dietro
LOL little problem there with ATT? |
|
yt
Premium
join:2008-06-03
| said by koshoka  :LOL little problem there with ATT? Not necessarily. I think many WoW Game latency issues are server based. Unfortunately they call performance issues "latency" which make people think network.
Also, one traceroute is not really a good indicator of network issues. Try running multiple traceroutes and look for patterns. MTR is a good tool for this.
Things to watch out for when looking at traceroutes and pings
•The most important data is pinging the last visible hop. If that shows clean, all previous hops can be ignored and problems attributed to ICMP Rate limiting
•MPLS cores like AT&T will show many hops with similar data and the first hop a bit of a jump. This is (unfortunately) an artifact of how MPLS works.
Ping the last hop and run MTR. Also try different realms for better performance. |
|
dtv09
@nauticom.net
| Whoa... is that through your proxy service? You go through Atlanta to Dallas, back to Atlanta and up to Boston. Considering Comcast usually traces directly between Philly and Atlanta, is that side trip from Atlanta to Dallas back to Atlanta the proxy service or how Comcast is dumping you into ATT? |
|
NormanS
Premium,MVM
join:2001-02-14
San Jose, CA
 ·Pacific Bell - SBC
| said by dtv09 :
Whoa... is that through your proxy service? You go through Atlanta to Dallas, back to Atlanta and up to Boston. Considering Comcast usually traces directly between Philly and Atlanta, is that side trip from Atlanta to Dallas back to Atlanta the proxy service or how Comcast is dumping you into ATT?
Eh? Comcast ('ibone') is taking him through Atlanta, Georgia, to Dallas, Texas, then handing him off to Teleglobe, Inc. Teleglobe, in turn, hands off to "AT&T Services, Inc." in Dallas. And ATTS turns him over to "AT&T Worldnet Services". I believe that ATTW uses an MPLS network, so from this point, it is all dependent on ATTW, and how MPLS works, about where he is going. This is the WoW upstream, and beyond the control of Comcast. Routing from the WoW servers to the player is largely controlled by WoW's provider's BGP tables.
--
Norman
~Oh Lord, why have you come
~To Konnyu, with the Lion and the Drum |
|
EG
The wings of love
Premium
join:2006-11-18
Union, NJ
3 edits | reply to dietro
said by dietro : Am I correct in thinking that from hop 14 to around 20 I'm going thru an mpls cloud? And then does that make hop 14 a label edge router? Yes. AT&T uses MPLS. That is why the you get the spike in latency at that hop. We are actually looking at the other end of the interface there.
And an average of 80ms is a bit high considering that distance.
80 to 100ms RTT's are expected for coast to coast latency.
[Edit]
Apologies to yt for the echo. Didn't see your response.
[Edit#2]
And to NormanS as well. |
|
tshirt
Premium,MVM
join:2004-07-11
Snohomish, WA
 ·Comcast
| reply to dietro
said by dietro :I've come to suspect the issue is packet delay variation within the games network. From what you have posted so far, I agree.
WOW is extreme popular, single hosted (via AT&T, I believe) and frequently reported as overloaded (most often the ISP is blamed, but if troubleshoot'd (I know that's not a real word  ) the problem APPEARS to be at the game server end. (or very close to it) well beyond comcast's responsibility. |
|
dietro
join:2009-05-29
Jacksonville, FL
| reply to dtv09
That isnt the thru the proxy. I suppose the proxy works because by juking the game into thinking I live somewhere else it allows me to hop on the cloud thru a different and more directly located label edge router, in NY. So it seems like I tunnel to NY or thereabouts and then hop on the ATT cloud there as well instead of traveling all over the country to get on the cloud in Dallas. |
|
dietro
join:2009-05-29
Jacksonville, FL
| reply to yt
Yt
I see you think it might be a server load issue. Some of the games instance servers have been locking players out lately because they have reached the designated capacity. Is their any way to differentiate between network latency and server response time?
Also, I've tried getting matts tracert to work but for some reason, probably the version I installed, it wont. I picked up the winMTR link from the wiki link you posted.
And regarding the tracert I posted. That tracert is an exemplar of the many tracerts that I've taken as well as others who have posted regarding the issue on the official forums only to be ignored or directed to their isp. |
|
dietro
join:2009-05-29
Jacksonville, FL
1 edit | reply to NormanS
Norman
To the best of your knowledge would the mpls be a feature that ATT presses on WoW, or is it a special service purchased/rented by the game to assist with traffic management?
Also from what you've said it I guess it sounds as though the first ATT hop on the tracert is not the begining of the cloud, the entry to the cloud is much later in the tracert?
And hello again, thanks for stopping in. |
|
NormanS
Premium,MVM
join:2001-02-14
San Jose, CA
·Pacific Bell - SBC
1 edit | I thought I understood "the cloud", but your question is causing me to reconsider what I know.
MPLS is a routing method used in, I believe, the AT&T Worldnet Services core. AT&T is one monstrous company, with many parts. Your trace shows two of those parts; AT&T Services, Inc. ([192.205.35.57]), and AT&T Worldnet Services (beginning with [12.122.138.110]). MPLS would be applied at least from this point, and throughout their network to the end.
As for "the cloud", I never knew there was a distinct point where you could say it began. From the standpoint of my LAN, I assumed that "the cloud" began on the far side of my residential gateway router. From the standpoint of my IP connection to the Internet, it would begin on the far side of my IP gateway router (which is an ATTIS device). From the standpoint of peering, I assumed "the cloud" began at the peering connection between ATTIS and "The Internet". So where does "the cloud" really begin? And where does it end? I guess it is really all just a murky cloud?
P.S. Comcast subscribers aren't the only ones having trouble with WoW and latency:
»Horrible latency on WOW
--
Norman
~Oh Lord, why have you come
~To Konnyu, with the Lion and the Drum |
|
EG
The wings of love
Premium
join:2006-11-18
Union, NJ
1 edit | reply to dietro
said by dietro :Norman To the best of your knowledge would the mpls be a feature that ATT presses on WoW, or is it a special service purchased/rented by the game to assist with traffic management? It's neitherAlso from what you've said it I guess it sounds as though the first ATT hop on the tracert is not the begining of the cloud, the entry to the cloud is much later in the tracert? Not so |
|
yt
Premium
join:2008-06-03
1 edit | reply to dietro
said by dietro :I see you think it might be a server load issue. Some of the games instance servers have been locking players out lately because they have reached the designated capacity. Is their any way to differentiate between network latency and server response time? Yes. When the application says 600ms latency (i.e Game Latency), but an MTR shows 80ms on the last hop, and you don't seem to be having other Internet problems, then the game server or your PC's processor load is suspect. NOTE: Given many people are pinging that last hop, even that data is suspect. The reality is people from many ISPs are complaining of disconnects and game latency. Blizzard has some work to do.
said by dietro :And regarding the tracert I posted. That tracert is an exemplar of the many tracerts that I've taken as well as others who have posted regarding the issue on the official forums only to be ignored or directed to their isp. That is irresponsible and pushes the cost of supporting their paying customer to another company who can't solve the issue at hand. I've read some of these posts and it is pretty obvious that the Blues don't no how to read traceroutes properly, don't understand ICMP rate limiting and selectively ignore posts which point to issues on their ISP AT&T or have no obvious network symptoms.
Blizzard is a mult-billion dollar online company. They should properly build their environment (vs outsourcing it all to 1 ISP) and properly support their customers.
Perhaps you should be directed to a new source of online game entertainment. |
|
