Topic Options
#3078 - 05/17/17 12:18 AM Weird packet loss in the evening, bandwidth not affected; Pingplotter vs Multiping
VincentZD Offline


Registered: 05/16/17
Posts: 2
I'm having a very weird issue with packet loss in the evening hours for around 5 weeks now. Pretty sure it's out of my control to fix but my ISP is refusing to assist because bandwidth is not affected and don't even want/are able to test for packet loss (haven't been able to convince them that it's a problem in their area of responsibility).

Problem in short:
- 7-10% packet loss for 1 to 2.5 hours in the timeframe of 7:00 pm til 11:15pm (specific time fluctuates/depends on week day)
- happens every single day
- no problems before 7:00pm or after 11:15pm ever

- bandwidth is not noticeably affected
- ms is slightly increased (50% increase, going from around 20ms to 30ms on average), jitter also slightly increased
- teamspeak/discord/skype, online games pretty much unusable in the affected time period
- videos, downloading, streaming, websites not noticeably affected
- affecting every site I test in Pingplotter (no common hops except my end and ISP's backbone)

- ADSL 2+ Annex J 16k (reaching 12,4k in reality)
- tested with both lan and wlan - no improvement
- tested with different router/modem - no improvement
- tested with different computer - no improvement

So I've kinda ruled out everything on my end except internal wiring but I don't suspect that to be the issue either cause then it would be constant or truly random wouldn't it?
I'm thinking congestion on the end of my ISP but that seems weird to me as well because the bandwidth is perfectly fine as is the ms and in typical cases I've seen these are heavily affected too?
Do you have any ideas what could possibly cause this issue of daytime related packet loss in combination with fine bandwidth? Because bandwidth seems to the only thing my ISP cares about.


I've tried troubleshooting with Pingplotter, Multiping, Pathping and WinMTR. I'm focusing on a EU Blizzard Server (185.60.112.157) here, yesterday's route via Pingplotter:

1. 84.61.224.1 (DSLAM/BRAS)
2. 84.61.224.1 (DSLAM/BRAS)
3. 88.79.30.162 (IP backbone of ISP)
4. 88.79.30.8 (IP backbone of ISP)
5. 92.79.212.109 (IP backbone of ISP)
6. 145.254.2.187 (IP backbone of ISP)
7. 145.254.2.187 (IP backbone of ISP)
8. 80.249.208.83 (Amsterdam Internet Exchange)
9. 37.244.10.35 (Blizzard)
10. 37.244.10.103 (Blizzard)
11. 185.60.114.153 (Blizzard)
12. 185.60.112.157 (Blizzard)

I'm starting with the WinMTR Screen (WinMTR Blizzard Server; 21:43, 17:43, 14:08, 8:49, upper left evening packet loss) here that indicates:
- latest possible point of origin for the packet loss hop 5
- can't judge hop 3 and 4 cause of unreliable responsiveness to ICMP echo requests
- hop 1 and 2 SEEM fine


Now I'm adding Pingplotter's time graph feature (Time graph Blizzard Server) to investigate this further:

- two caveats: from 14:20 to 14:35 I've watched a 1080p stream; around 21:35 I power cycled the router (I got a new ip; old hops 1/2 stopped, new hop 2 started)
- the 1080p stream is using a large chunk of my bandwidth, the ms increases, I even get 2 packet loss events, the ms of hop 3 and 4 reacts appropriately; the ms of hop 1 and 2 seems detached from that event)
- at 19:30 the evening packet loss slowly starts (at this point one computer via lan is the only connected device, my router software shows literally zero bandwidth usage, nothing but Pingplotter running), the ms increases (also for hop 3/4, not for 1/2)
- the packet loss continues and worsens, peaks at 21:15 til 21:45, lessens again and finally stops around 22:15

- from this I would deduce that the point of origin for the packet loss is somewhere between hop 2 and 3 but it's tough to "prove" cause of the ICMP echo request issue


But then there are also Multiping (Multiping zeit.de) and Pathping (Pathping Blizzard Server, google com; imgur.com/a/buL3g) which are heavily indicating that the problem actually originates between my router and hop 1. They show packet loss at the DSLAM/BRAS hop in the corresponding evening packet loss time frame. Pingplotter does so as well when I define the DSLAM/BRAS as target address (dslb-178-011...).


Now I'm unsure what could cause this packet loss without affecting the bandwidth and whether the problem starts:

- between my router and the DSLAM/BRAS (hop 1/2 84.61.224.1)
- or between the BRAS (hop 1/2 84.61.224.1) and the start of the national IP backbone of my ISP (hop 3 88.79.30.162)



Would really appreciate any input and help from you guys. Thanks in advance.

Uhm, didn't intent for this to become such a wall of text, sorry.









Attachments
WinMTR Blizzard Server.png (485 downloads)
185.60.112.157 23.pp2 (466 downloads)
Time graph Blizzard Server.png (588 downloads)
Multiping zeit.de.png (585 downloads)
dslb-178-011-192-001.178.011.pools.vodafone-ip.de.pp2 (409 downloads)



Edited by VincentZD (05/17/17 12:55 AM)

Top
#3080 - 05/17/17 03:57 PM Re: Weird packet loss in the evening, bandwidth not affected; Pingplotter vs Multiping [Re: VincentZD]
Gary Offline
PingPlotter Staff


Registered: 10/30/13
Posts: 185
Hi Vincent,

Thanks for writing in - and thanks for sending over such detailed info on your issue!

From what we can see in the results you provided, and your description, this really *appears* to be a possible bandwidth saturation issue (https://www.pingplotter.com/fix-your-network/common-network-problems/#bandwidthsat). While your router may not show any bandwidth usage during problem periods - it’s possible for bandwidth saturation to happen *outside* your network (at one of the ISP owned hops), which your router wouldn’t be able to see.

In the first pp2 file you’ve provided here, there’s a pretty clear pattern happening at the final destination (during the hours you mention) where the latency increases (and gets a bit “jagged”), and there’s an increase in packet loss. It’s a bit tough to see, but this pattern *does* seem to start at hop #3 (if you zoom out to a 12 hour view, you can see a pretty clear pattern of latency through the packet loss - which continues through through the entire route). This would imply that the issue is originating somewhere between hop #2 and #3.

Bandwidth saturation can have a number of symptoms, including packet loss. If a device is receiving more requests than it can service, it may start dropping packets, which could result in patterns like the one you’re seeing here. The fact that the issue is happening consistently between 7p and 11p - which would likely be the busiest times for the network - would be another strong indication of bandwidth saturation.

We’ve got an article that goes over some tips for dealing with bandwidth saturation (including some guidance on what to do if it’s possibly on your ISP’s end), which may be of some help to you here:

https://www.pingplotter.com/wisdom/article/bandwidth-saturation.html

Hopefully this helps get you headed in the right direction. If you have any other questions, just let us know!

Best wishes,

-Gary

Top
#3082 - 05/17/17 05:53 PM Re: Weird packet loss in the evening, bandwidth not affected; Pingplotter vs Multiping [Re: Gary]
VincentZD Offline


Registered: 05/16/17
Posts: 2
Thanks for your response, Gary.


So should I trust Pingplotter's data in regards to the origin of the packet loss (hop 1/2 fine, and it starting from 2 to 3)? What could cause Multiping and Pathping to show packet loss already happening between my router and hop 1/at hop 1?

Have you seen isolated packet loss caused by bandwidth saturation on the ISP's end? I believe that to be the cause as well but can't get my head around the fact that my actual bandwidth is not significantly reduced (ms also only slightly increased in absolute numbers), yet I'm experiencing the packet loss.


Thanks again for your assistance!



Attachments
Multiping zeit.de.png (341 downloads)



Edited by VincentZD (05/17/17 05:57 PM)

Top

Search

Who's Online
0 registered (), 18 Guests and 1 Spider online.
Key: Admin, Global Mod, Mod