In-Game latency vs WinMTR vs Game Lag?
|
Quick one that is really irritating me lately and hope that someone can provide help/suggestions.
I play on the European servers from South Africa on Standard leagues.. Hardcore is almost impossible for obvious reasons.. Normal in-game ping is around 215ms - not brilliant but perfectly ok to play. When everything is normal, WinMTR shows pings very similar to in-game (F1) results. My PC setup runs the game very well at max and never have performance problems other than network related. I play via Steam. Problem I experience is that in-game pings spike hugely sometimes... from 200 to 2000ms.. .The in game graph bubbles around 215 normally but there are some days when this jumps and flat lines over 1 sec and obviously no in-game interaction is possible.... i.e. you cant view your stash, chat timeouts, cant vendor things etc.. until the game catches up... sometimes 10 to 20 seconds later. When running WinMTR during this time to the 37.58.67.210 server in Europe (think its Amsterdam or thereabouts) it reports perfect 190 to 220ms pings... no matter how long I leave it running. Interestingly, if I get a gap I and can load a hideout or normal map, the lag normalises again... Not always, but sometimes... My only conclusion is that there are overloaded servers (or devices at the server end) causing this..... The network is not lagging and not losing packets all the way to the IP and if you can manage to load a map - which I understand can possibly be on another physical server - its possible that the issue goes away.... Now I am making assumptions here and its likely that I am testing to the wrong IP but whats going on. Suggestions.... Thanks |
|
|
The problem is neither: PoE's method of timing/sampling counter data is flawed and therefor the graphs (including FPS) displayed can be woefully inaccurate.
(This is based off a few things but mainly correlated data from VTune and Wireshark) "The problem is there ARE secure netcodes" -- Pewzor Last edited by Emjayen#1133 on Dec 10, 2014, 2:55:41 AM
|
|
|
Thanks for the reply....
So if I assume that your testing points to in-game reporting that is worthless or at the very least smoking the good stuff... What would be causing the in-game lag when the network level is not showing problems? To me this points to servers or localised network devices (on GGG side) that cannot handle load and are lagging... Or maybe someones doing a backup and killing the machine I am unlucky enough to be on. |
|
" There's a few possibilities: 1) Packet loss -- Tracert tools like the ones your using depend on a particular network protocol that is given special treatment in various ways. If for instance, your ISP was throttling traffic (which results in queuing delays and packet loss) then you wouldn't be able to determine it with a simple traceroute or ping. 2) Disk access -- PoE has a problem with page faults, particularly hard ones, which results in stalls in rendering while resources are being read from disk. 3) Client-side extrapolation -- PoE's prediction algorithm weaknesses (all games have them) could be amplified in the presence of theory#1 (I haven't looked at all into this aspect of the game, so it's just an educated guess) There's no real way to know without proper profiling and that should be done by the developers. "The problem is there ARE secure netcodes" -- Pewzor Last edited by Emjayen#1133 on Dec 10, 2014, 3:21:58 AM
|
|
|
Some ISP's will change routes based on your instance server. Try running another WinMTR to the instance server you are connected to while you are experiencing connectivity issues.
Instance server can be found at the bottom of the Path of Exile/logs/client.txt file. Also, smaller more concise WinMTR reports are more helpful than length 1-2 hour long reports. Taking a trace of 300 packets sent while you experience these issues will be most helpful for diagnosis. Feel free to post the report here in a spoiler tag. Need Support? support@grindinggear.com.
| |
|
Hey...
I have done what Yeran_GGG asked and compared tracerts to all the instance servers I could see in the client.txt... They are identical until the last hops at networklayer.com i.e. my ISP does not seem to change routes between 37.58.67.222, 159.253.137.134, 37.58.74.233, 159.253.137.133 and a few others... before its offloaded to networklayer.com there is no difference. Here is a copy of 300 odd packets via WinMTR to the current instance server: 37.58.67.222. I have experienced the odd spikes at least twice in this time for longer than 5 to 10 seconds each... the performance otherwise is pretty flaky while in town (Sarn) with a few high spikes - but these dont tend to flatline.
Spoiler
|------------------------------------------------------------------------------------------| | WinMTR statistics | | Host - % | Sent | Recv | Best | Avrg | Wrst | Last | |------------------------------------------------|------|------|------|------|------|------| | 192.168.1.1 - 0 | 313 | 313 | 0 | 0 | 3 | 1 | | 105-236-2-193-esr-lo.mtnbusiness.co.za - 0 | 313 | 313 | 6 | 7 | 90 | 7 | | ipc-recieve-rb-4a.za.mtnbusiness.net - 0 | 313 | 313 | 7 | 9 | 40 | 9 | | rb-dca-2.za--rb-cr-1.za-a.mtnns.net - 0 | 313 | 313 | 7 | 9 | 24 | 9 | | jh-cr-1.za--rb-cr-2.za.mtnns.net - 0 | 313 | 313 | 8 | 13 | 68 | 16 | | qux-jh-dca-2.za-b.za.mtnbusiness.net - 0 | 313 | 313 | 7 | 11 | 56 | 42 | | ls-cr-2.uk--rb-cr-1.za-g.mtnns.net - 0 | 313 | 313 | 204 | 209 | 279 | 206 | | ls-pr-1.uk--ls-cr-2.uk-a.mtn.net - 0 | 313 | 313 | 188 | 192 | 234 | 194 | | bbr01.lon01.networklayer.com - 0 | 313 | 313 | 188 | 191 | 224 | 191 | | ae7.bbr02.tg01.lon01.networklayer.com - 1 | 313 | 312 | 205 | 208 | 236 | 206 | | ae0.bbr02.eq01.ams02.networklayer.com - 1 | 313 | 312 | 211 | 215 | 336 | 213 | | ae6.dar01.sr01.ams01.networklayer.com - 0 | 312 | 312 | 196 | 199 | 311 | 201 | | po1.fcr01.sr01.ams01.networklayer.com - 57 | 312 | 136 | 196 | 205 | 324 | 198 | |37.58.67.222-static.reverse.softlayer.com - 0 | 312 | 312 | 196 | 202 | 222 | 203 | |________________________________________________|______|______|______|______|______|______| Here is the report for the previous instance server 159.253.137.134 - just for confirmation of the route. (I was not logging when the issue happened on this server)
Spoiler
|------------------------------------------------------------------------------------------| | WinMTR statistics | | Host - % | Sent | Recv | Best | Avrg | Wrst | Last | |------------------------------------------------|------|------|------|------|------|------| | 192.168.1.1 - 0 | 19 | 19 | 0 | 0 | 2 | 1 | | 105-236-2-193-esr-lo.mtnbusiness.co.za - 0 | 19 | 19 | 6 | 7 | 20 | 8 | | ipc-recieve-rb-1a.za.mtnbusiness.net - 0 | 19 | 19 | 7 | 18 | 89 | 9 | | rb-cr-1.za--rb-dca-1.za-a.mtnns.net - 0 | 19 | 19 | 8 | 11 | 47 | 8 | | jh-cr-2.za--rb-cr-1.za-a.mtnns.net - 0 | 19 | 19 | 9 | 13 | 23 | 20 | | qux-jh-dca-2.za-b.za.mtnbusiness.net - 0 | 19 | 19 | 7 | 13 | 89 | 9 | | ls-cr-2.uk--rb-cr-1.za.mtnns.net - 0 | 18 | 18 | 204 | 213 | 300 | 206 | | ls-pr-1.uk--ls-cr-2.uk-a.mtn.net - 0 | 18 | 18 | 189 | 198 | 239 | 207 | | bbr01.lon01.networklayer.com - 0 | 18 | 18 | 189 | 190 | 196 | 190 | | ae7.bbr02.tg01.lon01.networklayer.com - 0 | 18 | 18 | 207 | 213 | 227 | 209 | | ae0.bbr02.eq01.ams02.networklayer.com - 0 | 18 | 18 | 212 | 217 | 233 | 215 | | ae6.dar01.sr01.ams01.networklayer.com - 0 | 18 | 18 | 198 | 198 | 201 | 198 | | po1.fcr01.sr01.ams01.networklayer.com - 0 | 18 | 18 | 198 | 204 | 226 | 211 | |159.253.137.134-static.reverse.softlayer.com - 0 | 18 | 18 | 197 | 203 | 208 | 204 | |________________________________________________|______|______|______|______|______|______| Interestingly, in the first 300 packet report, you will see that there is packet loss at the second to last server po1.fcr01 but in the second report showing a route via the same ip does not show packet loss. Regarding the other possibilities suggested by Emjayen... Packet loss not showing on pings/tracerts - Ok.. I have sent my ISP a query to find out if they are shaping/prioritising or doing something that could be affecting POE traffic. Will see what they say. Disk Access. I saw this issue before but dont think its appropriate - I have a Samsung 840 EVO SSD so disk access is blisteringly fast and if this were the cause then shouldn't I experience it more frequently and in a predictable way? Also, apparently steam does a bitlevel (could be mistaken) type update so the pack files as I understand it would not suffer the same independent patch problem - could be talking crap here.. GGG would have to confirm. Can't comment on the prediction algorithm - that's above my pay grade... ;-) Anyway hope this helps. Cheers |
|
|
Those 5-10 second spikes you say you experienced in the first trace, were they the 'cannot interact with PoE' kind of spikes you specify in the OP?
That packet loss in the second to last hop of the first trace isn't reach our servers, and as such is likely just data shaping. I'm assuming that you are comfortable with the average latency jumping from 10 - 200 halfway through your trace? Unfortunately both of your traces are not indicating these large latency spikes you say you experience in-game. You may need to instead run a PingPlotter trace (incase there is an issue with WinMTR here). I would try and enter more typical gameplay areas until you experience these connectivity spikes occuring again, then check your client.txt file for the instance server at the very bottom and run the PingPlotter to that instance server for a max of 10 mins while you're in that instance. It's unusual that you are experiencing connectivity spikes that go for so long, however you are not being booted to the log-in screen. Are you running any other programs while you play PoE? Just to confirm, are you running a wired connection to your modem/router? I would also check your Firewall and Antivirus are not unnecessarily filtering PoE aswell. Need Support? support@grindinggear.com.
| |
|
Yes... the game does not freeze (animations still working) but I can't interact with vendors, other players, chat etc.. I can move my char around but nothing else happens and no other players move around... sometimes there is a rubber-banding when the game 'works' again but not always. The F1 report shows the first two graphs all changing up and down as normal but the third graph 'latency' just flat lines when this happens... i.e. it jumps to a high value and then just sits for anywhere from 3 to 20 seconds. I usually exit to login screen manually if the delay gets more than 10 seconds but I have waited to see what would happen a few times if I just waited.
The jump from 10ms to 200ms you see in the trace is when the packets traverse the long-haul link from SA to Europe and is "normal" in my experience as a South African internet user so yes.. I am not concerned about that one. There are times when this 'flatline' problem happens and I am booted by an 'unexpected disconnection from the server' but this is usually when its really long. My setup: Hardware (less than 6months old) - Intel i5-4690k at 3.5Ghz with 8GB ram Radeon R9 280x 3gb ram Game files & Steam installed on Samsung 840 EVO 256gb no overclocking of anything and default settings for graphics card etc ASRock motherboard with on-board audio and LAN I am connecting via 1GB LAN to 6mb ADSL+ link with linksys modem (vanilla setup with no rules/Qos etc) Software - Windows 8.1 *yes I know.. shoosh. Usually have a browser window or two open and obviously steam but no other major apps I sometimes run realtemp to monitor my system but not usually I use Norton 360 as a firewall/antivirus and its logs do not show anything potentially indicating a problem with PoE Network - Problem occurs even if I am the only one using the Internet connection. when the problem happens my first thing to check is the local net and my ISP, then I ping things like google.com or run a trace to services I know are not local just to make sure my connection is not the problem. I do not allow apps/etc to download or do things in background without confirmation and I dont run messenger/chats/skype type services. My processor barely works up a sweat playing and temps in the system rarely exceed 55 deg. c and the GPU while hotter is not pushed at all when compared to running benchmarks or primes My event logs etc show no errors I will attempt the PingPlotter and see if I can notice anything specific that may point to one particular server/group. BTW. my ISP has confirmed that they will look into how they handle PoE traffic - they are pretty good at making sure that online gaming works... will provide feedback when I get it. |
|
" Oh wow, I wish ISPs over here in Portugal were like yours. Usually when they see a report about a game connectivity, they just scoff at the idea of attempting to solving it. |
|
|
Xarick - the ISP I use has been voted SA's best for 4 consecutive years so they tend to work for their business... ;-) and they started out as very consumer focussed, hosting Cstrike game servers for free etc... They are the kind of ISP that sends out an email saying they have decided to double everyones caps because they are getting bandwidth for cheaper... lol
Update to issue... Just had the problem occur again, this time I was logging with PingPlotter(freeware).  Basically what happened in game is that the latency graph (F1) hit about 1700ms and then flatlined... it did this three times while I was logging. Each time, I waited at least 5 to 10 sec and was rubberbanded and the scene changed to my chars death with the ask to respawn in town/checkpoint. Once I manually exited to login via esc. (I was playing in Broken Bridge and Crypt maps when these problems happened) The pingplotter graph shows packet loss @ networklayer.com that almost exactly matches when my problem occured. I also noted the following in the client.txt (timestamp in GMT+2) 2014/12/10 21:38:21 45807609 cc [DEBUG Client 2932] Object: 7915 died to degen from buff with owner (killer): 20449 and aura_owner: 0 Also, before I started logging, there was an unexpected disconnect with the following error in the client.txt (I started logging after this happened). 2014/12/10 21:31:58 45423640 2a6 [INFO Client 416] Abnormal disconnect: An unexpected disconnection occurred. 2014/12/10 21:32:24 45449656 502 [WARN Client 416] Steam: Not connected to steam Hope this helps... |
|
