faces

When I said I couldn't think of another reason I probably should've said I couldn't think of another good reason for it to be happening.

It could well be unintentional but I can't think of a situation where having a high ping could lead to that much of an increase. For me its more than doubling my ping, from 250ms outside the game up to ~600ms in-game, and it varies even between servers on the same battlegroup (like I said before, I get better pings to Mal'ganis than Duskwood even though Mal'ganis is a lot more popular).

 

Cyn

If their was an issue with just their server code not handling larger time differences between communication between the client, then that would be fair enough, however, running through a tunnel does not change this, the same physical communication delay is there, you aren't remotely running the WoW client on the other end of the tunnel.

There are three parts to the communication (latency's are half of what you see in ping, considering we're looking at each step separately):

A) Australia to the server we can see: ~90ms
B) Server processing: X
C) Server to Australia: ~90ms

X is the real mystery here

Now running a tunnel in a pure networking sense, only effects step A and C, and if you're current routing is already efficient, it has a minimal difference.
So for an Australian, A and C add up to 180ms, and for an American, maybe somewhere in ~50ms range.

So this brings us to the mysterious X, given an American may have in game latencies of ~150ms, this implies that there should be a universal value of X close to 100, putting Australian ping at ~280ms. So why is it that X is closer to 300-400 for an Australian, even when you have Australians and Americans standing right next to each other in game, and thus should in theory, have the same X value.

So this leads me to the conclusion that A and C have some effect on X, but why? If the actual physical delay caused additional server processing, and thus was a direct relationship between A, C, and X, then a tunnel should not effect it. So finally, how else is X influenced purely by using a tunnel? Programatically based on source location is the only conclusion I can come too.


Alternatively: The client shows ping between the US end of the tunnel and US servers, and everyone here is running on a revolutionary drug called placebo.

 

Drunkmunky

Another thing I just thought of is that ping packets don't really use a whole lot of bandwidth, if the links are being choked at prime time (I'm pretty sure they are which is why new links are slowly being built) and the packets that wow sends are significantly larger than your normal ICMP echo packet then that could add to the slowdown. Of course that all fails because blizzards own in game ping goes up, unless of course they measure in game ping from game packets instead of seperate ping packets?

 

faces

I'm fairly sure this would affect proxied traffic as well, which it doesn't seem to since finding a low-latency US proxy definitely speeds things up significantly.

 

Cyn

A tunnel achieves two things.

A) Makes the source IP look like the other end of the tunnel
B) Give you a pre-determined routing path, rather than being at your ISPs whim.

Now B is a non-issue if your ISP already has a decent unsaturated routing path.
So why does A matter so much here?

 

ter

There is a slight flaw in what you're saying here. With or without the tunnel, your traffic (the actually packets) are never pursuing a predetermined path. The only thing you're guaranteeing is that the lowerping server will be one of your hops. Telcos use weird and wonderful methods of determining routes. Complex algorithms take into account many factors such as saturation, hop count and cost (not talking $$$ here) to determine what is the best path for each of you packets.

One other thing that should be considered is that the route to the servers may be fine, but the return route may be inefficient (no, the to and from routes don't necessarily have to be the same). Not sure if this has any real merit however.

The only logical conclusions that I can see are:

1) Blizzard discriminates.
2) QoS occurs somewhere between your modem and the blizzard servers. Higher priority to ssh than to WoW traffic. (Unlikely)
3) Network issue exists close to blizzard, such that routing through the tunnel to your proxy routes around the bad hops. This could also explain why some people using a private proxy sees no benefit (they're still routing through the trouble area)
4) Australia's sins have have incurred the wrath of God.

I used lowerping over the weekend. I went from 450-500 pings to ~180. This was shown in the quartz casting bar, and the general feel of play. I had no disconnections what so ever, service was flawless. That said, I'm a raider first and foremost, and I see no benefit for raiding. I may end up paying for it for purely the PvP side, but I'm not doing this just yet.

Ter

 

Soul

Well, if packet volume is high, and your router is overloaded, what metric is used to determine what packets get transmitted and what packets don't in the absence of QoS (or, at least for packets with equal priorities under QoS rules)? In my limited experience, a lot of routers use hop count as a metric, so a 15-hop packet has a higher chance of getting dropped than a 10-hop packet which has a higher chance of getting dropped than a 5-hop packet. To a router outside the tunnel, the tunnel looks like a single hop...

Of course, this is a zero sum game, so some poor schmoe out there is probably experiencing an increase in dropped packets as more people start using tunneling services.

 

ter

This could be true. My understanding, however, was that hop count isn't widely used as a metric on the internet, only in small(relatively) scale networks. The problem being that protocols such as RIP and IGRP (these use hop count as the primary metric) don't scale anywhere near well enough to cater for a network as large as the internet.

Truth be told, we probably won't get answers too this problem unless a knowledgeable network engineer who actually deals with the internet drops a line. In the mean time, some of us have found a handy way around the problem, which, in my opinion, comes at the right price.

 

Adis

I have a trial of lowerping going atm. It feels better, but quantified testing is harder to nail down.

The very sluggish nature of the WoW ingame ping readout (it updates its average at quite large intervals) makes it difficult to assess meaningful correlation between what it says and current ingame performance. I don't use quartz and don't know enough about it to comment on how useful it is for quantifying real performance. People are saying that it is reporting improvements, but is that comparing pre-2.3 numbers with post? Because I have a suspicion that the fastcasting changes in 2.3 could have effected the way the castbar is transacted compared to what it was before the patch.

My test scenario was to run two clients on my machine and set one character to follow the other. The distance that the follower trails the leader by from the leader's perspective gives an indication of real latency. You need to find a nice flat area with no obstructions or annoyances (I chose shimmering flats) then zoom out to max, go to a perfect top down view, then compare how close the follower is to the edge of your screen etc. Every time you start moving you are subject to spikes and imperfections in the follow logic in setting up the actual server side follow distance - for this reason I would start moving observe the stabilised following distance, stop, and repeat over and over to get an average. The test isn't foolproof of course, but its the best I could come up with for quantifying something so difficult to measure as relative position according to clients and servers when you can only observe the client(s).


I use Telstra cable in Brisbane. I tried running both WoW clients without lower ping, then with, then without etc and so on for sessions of about 5 minutes each, running back and forth over the same bit of flat ground... and compared the follow distance. After an hour or so of total sampling I determined that there was negligable difference on average and if anything follow distance was more consistent without connecting through lowerping. The ping reported by the game averaged ~400ms when running as normal, and ~225 when running using lowerping.

My conclusion, lowerping's real impact on gameplay (if any) does not match its apparent impact on the ping readout.

 

• Falk

Just so it's clear, the 2.3 changes to fastcasting did absolutely no change to the delays between keypresses and the appearance of a castbar. What it did was make the client act more intelligently in predicting what should happen next. (This wasn't limited to just the castbar and GCD - take cancelform for example) The actual delay remains unchanged.

 

Excession

Orcon are now telling me that they had run out of international bandwidth, and are getting more around the 27th of November. A pity that it took them so long to notice.

 

Joy

Be wary that it takes "up to 6 weeks" for Telecom to provision new Bandwidth to ISPs.

 

Drunkmunky

This of course assumes that all the ISP's links are using RIP and IGRP, we don't know that and can't possibly know for sure. There are many different routing protocols that use different factors to choose routes and I agree, unless we have a network engineer who works on some of these links come and tell us how it all works we can only speculate.

 

Avengement

All I can say is that I went from ~500ms down too ~230ms. The game feels "smoother" my casts "start" right away when i request them to be started and my insta casts "go off" right away as well. For my $7 a month I'm more than happy with this service.


Veng

 

Joy

Seriously people give this a go if you can afford it.

Brilliant.

 

moowalk

"Hey there's this new legislation so we have to open up our infrastructure to competition" "Don't worry, we can still screw over companies who try to compete with our monopoly"

I've switched to telstra-clear and it's great. No more 900+ ping at peak times. No port filtering for skype calls. It's probably the best one can do atm.

 

Xavias

Just tried lowerping.com 's service and it is nothing short of amazing.

My quartz is reporting 195MS consistantly on every cast. I used to get spiked quartz readings ranging from 350 to 800.

Absolutely amazing.

 

◊ forostie

It can take up to 6 weeks for Telstra if they hate you (from the ISP's perspective), although I've requested additional bandwidth from both Telstra and Optus on behalf of my ISP (Asianetcom and Southern Cross Cables respectively) and have had both actioned within 4 working days.

I've heard reports that people are being suspended for using Lowerping.com 's service due to TPA violations; I've tried it myself and even if there is the risk of getting suspended, it makes such an awesome difference.

 

Doriangray

Oh yes, this topic strikes nerve with me. I actually play from the other side of the ocean than the OP, i.e. from Europe. Ping times are equally bad and it seems to quite a few of us that they've gotten significantly worse with the infamous 2.2 neverending-story-patch. I'm in the lucky position to rarely have pings above 550-600 and the occasional 300, but i know we have people that never play with less than 800, and at that point it can get really annoying.

What really strikes me as odd though is that there still is no (even paid) possibility to transfer between regions. I asked several times, and I always got a polite - in the help desk way - but resolute answer of "no way, now please leave us alone". *shrug* We don't really all want to reroll here for obvious reasons, so our guild is stuck where we are.

 

Vaagren

What about setting servers up in Hawaii for Aussies and Oceanic players would that make much of a difference or no?

 

• Beliandra

I'm almost positive that any connections from Australasia to Hawaii would go via California. Anyone know a site physically located in Hawaii that i can traceroute to?

 

Dolohov

[Beliandra]
>I'm almost positive that any connections from Australasia to Hawaii would go via California.

Not necessarily: Southern Cross Cable - Wikipedia, the free encyclopedia .

Of course, there's no guarantee that your packets will actually be routed over the Southern Cross Cable.

 

drizzy

A few years ago I was looking at routing to Hawaii for another game (QWTF), and it's pretty much useless. About the only network in Australia that'll directly go to Hawaii is AARNet to uhawaii (Traceroute). Anything commercial pops over to the west coast and back, I think.

 

Jordanish

Well our new prime minister has promised world class broadband!
At least that is something...
Now we just need local servers =((

 

Drunkmunky