So hows that server latency?

Technical Support
21

Thanks, I’ll post a full WinMTR when I get a chance here, but note that much of the time, when this becomes a noticeable issue and is visibly affecting gameplay, the netgraph is difficult to capture due to the circumstances and even when I have gotten a clean shot of a problem spike, they don’t always seem to reflect what was happening in-game. Much of the time when I check immediately after a blink that rubber bands, it looks stable.

All the same, I will check to see if I have any recent captures that might work and even try to get you some fresh ones with a fresh install. I decided the 200 gb BO4 install for 3 maps of zombies is a complete and total waste of space, so I need to do some clean up anyways.

However, focusing on analyzing the current route will only tell us so much and ultimately get us as far as it has every other time this game became unplayable. The point I’ve been making is that forcing your playerbase to compete on an unfair regional distribution of host servers is kinda retarded, when your competitors are doing a much better job. you only need 12 players in a game plus what, 4 spectators? Splitting the player base shouldn’t really be that much of an issue, we all connect to NA but get split from there, doesn’t stop us with playing with players on the other side of the country. I understand we may not have the numbers your state side players present, but we still pay to play as well, and we’re not getting to even do that half the time.

22

True there’s not many posts about the issue but that’s because its an inconsistent bout of input lag/lag where shots are inconsistent and the double animation part is only noticeable on characters that have longer than normal shot animations like Genji’s right click. The average player wouldn’t notice or care to notice. Not to mention your player base has declined in the years so I doubt there’s as many serious players out here (just an observation not an insult). From what I’ve noticed this issue is Province specific to Alberta at the very least as I haven’t had any reports from other provinces or locations. I’ve had about 3 people in Alberta contact me of their own accord after finding my Twitch video of the double animation/input lag glitch.

So really the conditions for people to notice are either:

  • The player is above average and plays competitively (most likely grandmaster or above) and notices input lag
  • The player is a Genji main that frequently flick shots

The shot delay aspect is more difficult to prove which is why I fixate on the double animations.

Like I mentioned to EpicGG I have a friend down the street that plays and streams Overwatch on occasion and he had no earthly idea he was affected by the issue because he plays exclusively support roles that don’t require split second reactions (mercy or Moira for example). I had him test Genji and told him to right click while flicking and sure enough the issue came up. But if you look at the video I posted in that thread you can see clear double animations with no change in ping, or SIM whatsoever. I’m not going to reiterate what I wrote in that thread but I’ve tested this thoroughly at different locations with different PCs using both Shaw and Telus. It can’t be considered an isolated incident if I can test anyone else’s equipment in Alberta and mimic that result. I’d even be more than willing to film myself using my own computer and then replicating the issue downtown at a gaming Cafe and trying the 50+ computers they own if that’s what it takes.

I don’t want to have to move to the USA just to play a game at a professional level.

23

Hello again,

I can’t seem to record these events effectively enough to demonstrate the problem and the netgraph isn’t showing unusual spikes despite still feeling occasional conflicts in the movements and fire input. I haven’t been able to reproduce a large enough rubber band from a dash or blink that it would show up clearly, however I did have two incidents while testing this morning.

#1, clearly threw a flashbang into sigma and an ulting sym on Lijiang, had no affect despite being well within in range of both characters, it seemed to be blocked by the sym ult before it appeared visually, while also causing a small rubber band and input freeze, not unlike being stunned for under half a second.

2# Crit headshots on a soldier in Nepal caused a brief spike that fired particle for the hit, with no allies in LOS to mistake the hit fx from, but with no damage applied, at the same time as killing me with a helix that triggered before soldier even started his animation for firing the rocket.

I then captured this MTR, where I noticed after a short movement lag just walking around that the Worst value under 137.221.65.6 jumped up to the thousands from only about 200 previously. A quick lookup confirmed that this is directly from Blizzards internal servers as you can see in the following capture of the window…

https://drive.google.com/open?id=1ZLpl5QqTmKNyj7Y_qJZVShruMCTJMqNs

I’m not sure this will tell us anything more than we already know, but I should also mention that I have replaced my old Hitron modem with a newer one and have ensured that the hardline connection is not being hindered in any way, (crossed wire interference, faulty slotting, etc…). I replaced my CAT7 with CAT6 cables just in case, and have ensured that the modem wifi was off for the test and this was the sole connection, to at least rule out any other local sources like Telus’ network TV systems.

So I’m curious, assuming this spike is correlated with the gameplay bugs, what is it that causes these response issues from your servers and how are they affecting the client simulation so negatively?

24

Hey again EpicGG,

Thanks for the details and the screenshot of the WinMTR. I understand that it can be difficult to record the events given that they are unpredictable. The most useful thing would be a video recording of you playing while having the network graph on screen (although I realize that it’s tough to play effectively while having a big graph in the middle of the screen). This would give us the most accurate idea of what the connection is doing when these issues occur.

The two scenarios that you describe sound like they could have happened due to a combination of a blip in the connection and “favor the shooter” mechanics. Although it is difficult to say for sure without seeing what actually happened.

In regards to the WinMTR results, the large latency spike you are seeing on the as57976.net node is due to whats called “ICMP de-prioritization”. It’s a filtering technique that can be applied to high-traffic routers in a datacenter, and it basically treats the test packets as lower priority than other data packets. WinMTR, traceroute, and pathping all use ICMP packets to measure latency, while the game itself will use UDP packets to transmit your movements/shots/etc. Not all routers will have this enabled, but it is a common feature used by some datacenters and ISPs to protect against DDOS attacks via ICMP.

If we look lower down the WinMTR test at the next 2 hops we can see that the latency returns back to expected levels. If this high latency actually indicated an issue we would see the high latency persist down those last 2 hops as well. You would also see a noticeable latency spike in-game, but it doesn’t sound like you did in this case.

It could be that in your case, like VeiledFlame’s case, the connection issues do not actually show up in a WinMTR. Think about this. WinMTR only sends 1 packet per second to the destination IP, while Overwatch sends 64 packets per second (on a good stable connection). So if a couple of packets are being delayed and it is happening very very briefly, you could see issues in-game but nothing would show up in the WinMTR.

Here’s where things can get interesting. What happens if you try connecting through a VPN before loading up and playing Overwatch? A VPN will change the route your connection takes through the internet to get to the destination, and it will also encapsulate your packets (encrypting them). Your latency would also change, sometimes (depending on where the VPN server is hosted) it can be pretty similar to your normal latency. But the real test would be to see if it stabilizes your connection and removes the little “blips” that you feel while playing. If things feel better while using a VPN it would help narrow down whether or not the issue is specific to your “normal” route that your ISP gives you, to get to the Overwatch servers.

1 Like
25

What I mean is that the nvidia captures cause so much input lag that it’s hard to differentiate between latency and input lag, I’ve been monitoring the netgraph while playing basically all morning. As stated, it doesn’t seem to display anything alarming during the minor occurrences as I’ve been having regularly, and the more major issue that pop up like rubber banding hasn’t happened while testing yet, but I’ll keep trying to get a capture until it does.

The blips you mention are very regular and don’t require a death to occur, leading to some very messed up scenarios. Favor the shooter is great, so long as the shooter has the better latency and gets there first, I guess.

I understand what and how packets are prioritized and even gone as far as to set custom QOS rules in the past, on previous modems and routers running Tomato.

I see that in this example, the latency does return to normal after and usually that would be satisfactory, however the fact that it popped up almost lock-step with the movement glitching, just walking around is what made me think it indicated a deeper issue.

You make a valid point on how that MTR was recorded at a single packet/s so I’ve increased the rate to 0.01 (capped to 0.0?) and continued to record a few extra tests. During QP, I see a 4% packet loss on average at the first hop. That might indicate something to look into, given the current “default and isolated” state of my network I’m not sure what could be causing it, but I do also see more examples of 1000+ ms latency continuing down the chain past the cutoff. I’ve posted more examples of the 4% packet loss here:

https://drive.google.com/open?id=1oTJuLTw9MTC1Kn5JNSUIVtHJOGEQcOOS
https://drive.google.com/open?id=1PBTRRM0MexSiXfg05nJt91FBGHZDrl1e

These are simultaneous recording of both the posted ip for US West, as well as the ip of the game server as shown in the netgraph. Given the actual number of packets and the duration of the test, being that of a full qp round, ~3-5min, I’d say this is still less than the packet traffic you’d see under a UDP connection.

I hesitate to hook up any VPN’s due them never providing any benefits in the past, causing hundreds of ms in additional latency and despite claiming to protect your traffic, being one of the least secure methods of doing so while also one of the highest potential for abuse. Do you recommend a VPN specifically, I can’t find any with a reputable profile, just lots of sleezy and overstated advertising claims that prove untrue according to any review not curated directly by the authoring company…

Given my age and experience with game networks on UDP connections both on local networks and hosted, it really feels like the game is constantly fighting me in every action I try to make, sometimes much worse than others which leads to what is classically described as rubber banding. The client simulation does a decent enough job hiding it, but when you feel it, the game might as well end right there.

So do you have any idea what could be the cause of a ~4% packet loss from my local modem on high speed traffic? If I get a chance to record the netgraph doing anything weird I’ll post it, until then I’m just waiting to reproduce and isolate an incident. As you see in the captures, the latency does end on the hop when monitoring the public ip for US West, however the actual game hops show signs that the latency could be building and extend passed the masked hops, of which there are 21 more of, that are unreported.

I’m not entirely certain the latency isn’t hidden on the public IP MTR…

Finally, getting back to the main point, this felt bad enough when I was playing from Vancouver and now that I’m in Alberta, it’s the most frustrating and least enjoyable experience in my current list of games to play, which is quickly making it fall off that list. If you want to keep your player base in this region, something has to be done to improve it, cause at least I have expired every option I am aware of, or that anyone has given me.

26

We can’t officially recommend any one VPN, since these are third-parties we’re talking about and Blizzard isn’t officially affiliated with any VPN company. There are some reputable ones out there that I have personally found, but perhaps you can search around more and get ideas from other players who have had experiences with them.

The 4% packet loss on the home router could be DDOS protection, just like we are seeing on the later networks. If it was actually 4% loss this would show up in-game in the Network Graph and you would definitely feel it. This is likely just the router dropping some of those test packets due to the increased rate at which you are sending them. I’ve messed with increasing the rate of packets in WinMTR before and things start to get weird the faster you go, so that adjustment is just skewing the results.

Again though, we can’t really discuss expanding the server pool here on the Tech Support forums. Anything that goes into the territory of how the game engine (or netcode) works, or where/how servers are hosted, is way beyond what we do here. If you would like to suggest that a Canadian Overwatch server be opened up that would be something you’d want to discuss on the General Discussion forums, which is where our Community team can read it. Game ideas, suggestions, and feedback are collected from the General forums and passed along to the Overwatch designers.

As far as what we can do here, we are just here to troubleshoot and attempt to resolve issues. We are more than happy to continue troubleshooting, so long as we actually have something tangible and verifiable to work with.

27

Well I’ve exhausted my time already trying to find some way to improve it. I’ll try to get more captures later.

The only thing I could do to tangibly show the issues reliably, is try to record examples of how far off the client is compared to the server replays, at any given point, but the input and recording latency have made it difficult to get clean comparisons due to the way the videos export.

All the same, streamers have lots of footage of similar issues and it shouldn’t be hard to find many examples of. I expect once the rubber banding starts acting up again I’ll get some better data to work with for my case at least.

Do you know if the Devs have actively polled the community in any way to try and get information regarding gameplay experience from the community as a whole? I’ve been trying to find threads discussing this but googling returns literally hundreds of forum posts under this topic that don’t ever seem to get a response or a deeper discussion. Seems like almost every post in the general forums are either hugely popular topics or dead weight and abandoned shortly after. How does anything get seen through all the sym and sigma complaints right now?

It’d be nice to just see some activity, some improvement over the lifecycle of this game, instead of a degrading experience with more and more cosmetics.

28

Having glanced at the many posts since I last saw the thread the VPN suggestion I know for a fact does nothing to mitigate this issue. I do find it a little funny that you guys aren’t allowed to suggest VPNs, but do anyways on a technicality. All a VPN does is reduce the ping slightly but the symptoms remain no matter where you route I’ve tried every American and Canadian server through VPNs and it makes absolutely no difference in the way the game feels; double animations still occur.

I too have tested faster packet sending in WinMTR, even matched it to overwatch’s rate. I always get the false packet loss on the router’s IP that doesn’t propagate to the other hops as well between 5-10 percent. Same Equinix hops show high latency and often show 1 percent loss between the 3 that I usually see, which as you said is most likely ICMP de-prioritization as well. However that being said you guys block pings on live servers to the destination so I can’t properly see what the connection is doing beyond those Equinix routes.

But yeah, VPNs do absolutely nothing to mitigate the problem which can only mean 2 things:

  • Our connections are bad from our ISPs; which is unlikely as I’ve tested 3 ISPs with clean connections
  • Something on blizzard’s side is causing the issue; which is more than likely as I seem to have no problems with any other game. Which I might add is a common story i see among the threads, where people are saying their other games are fine. To properly test this point I might need to find another game that’s coded much of the same connection-wise, if you guys have suggestions let me know.