O Canada! > Canadian > TekSavvy > Packet loss rates up to 2% on TekSavvy DSL

Packet loss rates up to 2% on TekSavvy DSL

I believe Teksavvy has ICMP packets set to low priority, so when you ping/trace, and the router is busy it will show as timed out. This doesn't mean it is down, if you never get dropped downloads/uploads then your line is fine.

If teksavvy had any issues on their internal network they would have fixed it by now or at least told us about it.

>I believe Teksavvy has ICMP packets set to low priority, so when you ping/trace, and the router is busy it will show as timed out.

OK, this would mean to me then that they are, in effect, doing traffic shaping or throttling of certain types of IP traffic (ICMP is an IP protocol). My understanding was that they weren't doing any such thing, treating certain IP packets differently.

>This doesn't mean it is down, if you never get dropped downloads/uploads then your line is fine.

Yes, I agree, nothing is "down". TCP makes up for lost packets and most of us would never notice if 2% of packets are lost, assuming the time distribution is random. Even VOIP should be OK with such packet loss, but 0% is better.

Don't get me wrong, Internet still works OK at 2% loss rate; Even google.com is showing me 1-3% packet loss (on my Videotron cable which has 0% loss for other sites). I'm just trying to figure out if this is normal for TekSavvy or perhaps due to a sometimes, somewhat congested local router.

If ICMP packets truly are being treated differently on TekSavvy network, I'm sure I can find another IP or UDP protocol to do some latency and packet loss testing with.

reply to hiremichael
You're not the only one:

»Peer1 is junk

I still haven't figured out if it's actually TekSavvy or Peer1 that's causing the packet loss though.

Anyone who sees these IPs in a trace route will more than likely see packet loss during peak times:

65.39.134.1
65.39.198.250
216.187.114.161
216.187.114.141

I haven't been able to narrow down the source of the packet loss further due to the fact that 65.39.134.1 sets ping packets to low priority, so it's showing inaccurate numbers. However I'm pretty sure the packet loss starts at either 65.39.134.1 or 65.39.198.250.

i lost connection to the netw for a few hours last night

reply to hiremichael
Try UDP traceroute

reply to Angelo
>i lost connection to the netw for a few hours last night

I suspect that is not related to this packet loss. As far as I can tell, this packet loss is normal for TekSavvy, and probably not a major concern for most people, unlike those who lose all internet connectivity.

you can try path ping also...

reply to sMURF
>65.39.134.1

Yeah, thats router number 3 in all my traceroutes to the Internet, and that's where I think my packets are being lost:

Number 1 is my router and number 2 is also used for direct routes to teksavvy.com, which don't have this packet loss.

traceroute to linux.org (198.182.196.48), 64 hops max, 40 byte packets
1 d (192.168.9.1) 15 ms 99 ms 104 ms
2 206.248.154.102 (206.248.154.102) 23 ms 24 ms 24 ms
3 65.39.134.1 (65.39.134.1) 18 ms 24 ms *
4 65.39.198.250 (65.39.198.250) 17 ms 17 ms 17 ms
...

traceroute to teksavvy.com (206.248.154.253), 64 hops max, 40 byte packets
1 d (192.168.9.1) 94 ms 100 ms 100 ms
2 206.248.154.102 (206.248.154.102) 18 ms 17 ms 18 ms
3 206.248.155.236 (206.248.155.236) 17 ms 17 ms 17 ms
4 206.248.154.66 (206.248.154.66) 25 ms 25 ms 24 ms
5 * * *

I'd be happy to hear from some TSIer(s) here, but my suspicion is that this is business as usual for Teksavvy. It helps explain to me how Teksavvy can offer good support for $30/month unlimited/hi-cap, when Bell and the cable co's charge more for less skilled support.

I hope noone here thinks I'm trying to "bash" Teksavvy, just trying to compare their DSL to Videotron cable to explore my options. Videotron support, repair and customer service is certainly not the best, although I think they are better than Rogers.

Tracing route to 65.39.134.1 over a maximum of 30 hops

1 16 ms 99 ms 99 ms dsldevice.lan [192.168.1.254]
2 12 ms 14 ms 13 ms 206.248.154.102
3 12 ms 12 ms 12 ms 65.39.134.1

Trace complete.

works perfect here...

yours seems good also

reply to hiremichael
also please note this is with my connection currently being used for downloading at the same time...

Tracing route to teksavvy.com [206.248.154.253]
over a maximum of 30 hops:

1 19 ms 98 ms 99 ms dsldevice.lan [192.168.1.254]
2 43 ms 56 ms 35 ms 206.248.154.102
3 46 ms 39 ms 13 ms 206.248.155.236
4 32 ms 21 ms 55 ms 206.248.154.66
5 59 ms 72 ms 35 ms www.teksavvy.com [206.248.154.253]

Trace complete.

C:\Windows\system32>tracert linux.org

Tracing route to linux.org [198.182.196.48]
over a maximum of 30 hops:

1 29 ms 99 ms 99 ms dsldevice.lan [192.168.1.254]
2 14 ms 12 ms 14 ms 206.248.154.102
3 44 ms 48 ms 54 ms 65.39.134.1
4 56 ms 54 ms 51 ms 65.39.198.250
5 47 ms 53 ms 51 ms gig4-0.tor-gsr-a.peer1.net [65.39.198.225]
6 47 ms 18 ms 15 ms 216.191.190.105
7 73 ms 20 ms 45 ms pos8-0-0.hcap1-ott.bb.allstream.net [199.212.172
.22]
8 29 ms 57 ms 49 ms 216.191.132.150
9 41 ms 55 ms 41 ms router.invlogic.com [207.245.34.122]
10 60 ms 52 ms 52 ms linux.org [198.182.196.48]

Trace complete.

C:\Windows\system32>pathping linux.org

Tracing route to linux.org [198.182.196.48]
over a maximum of 30 hops:
0 Angelo-PC [192.168.1.11]
1 dsldevice.lan [192.168.1.254]
2 206.248.154.102
3 65.39.134.1
4 65.39.198.250
5 gig4-0.tor-gsr-a.peer1.net [65.39.198.225]
6 216.191.190.105
7 pos8-0-0.hcap1-ott.bb.allstream.net [199.212.172.22]
8 216.191.132.146
9 router.invlogic.com [207.245.34.122]
10 linux.org [198.182.196.48]

Computing statistics for 250 seconds...
Source to Here This Node/Link
Hop RTT Lost/Sent = Pct Lost/Sent = Pct Address
0 Angelo-PC [192.168.1.11]
0/ 100 = 0% |
1 0ms 0/ 100 = 0% 0/ 100 = 0% dsldevice.lan [192.168.1.254]
0/ 100 = 0% |
2 31ms 0/ 100 = 0% 0/ 100 = 0% 206.248.154.102
0/ 100 = 0% |
3 32ms 1/ 100 = 1% 1/ 100 = 1% 65.39.134.1
0/ 100 = 0% |
4 34ms 2/ 100 = 2% 2/ 100 = 2% 65.39.198.250
0/ 100 = 0% |
5 36ms 1/ 100 = 1% 1/ 100 = 1% gig4-0.tor-gsr-a.peer1.net [65.39.
198.225]
0/ 100 = 0% |
6 --- 100/ 100 =100% 100/ 100 =100% 216.191.190.105
0/ 100 = 0% |
7 --- 100/ 100 =100% 100/ 100 =100% pos8-0-0.hcap1-ott.bb.allstream.ne
t [199.212.172.22]
0/ 100 = 0% |
8 47ms 1/ 100 = 1% 1/ 100 = 1% 216.191.132.146
0/ 100 = 0% |
9 49ms 0/ 100 = 0% 0/ 100 = 0% router.invlogic.com [207.245.34.12
2]
1/ 100 = 1% |
10 48ms 1/ 100 = 1% 0/ 100 = 0% linux.org [198.182.196.48]

Trace complete.

C:\Windows\system32>pathping teksavvy.com

Tracing route to teksavvy.com [206.248.154.253]
over a maximum of 30 hops:
0 Angelo-PC [192.168.1.11]
1 dsldevice.lan [192.168.1.254]
2 206.248.154.102
3 206.248.155.236
4 206.248.154.66
5 www.teksavvy.com [206.248.154.253]

Computing statistics for 125 seconds...
Source to Here This Node/Link
Hop RTT Lost/Sent = Pct Lost/Sent = Pct Address
0 Angelo-PC [192.168.1.11]
0/ 100 = 0% |
1 0ms 0/ 100 = 0% 0/ 100 = 0% dsldevice.lan [192.168.1.254]
0/ 100 = 0% |
2 33ms 0/ 100 = 0% 0/ 100 = 0% 206.248.154.102
0/ 100 = 0% |
3 32ms 0/ 100 = 0% 0/ 100 = 0% 206.248.155.236
0/ 100 = 0% |
4 41ms 0/ 100 = 0% 0/ 100 = 0% 206.248.154.66
0/ 100 = 0% |
5 42ms 0/ 100 = 0% 0/ 100 = 0% www.teksavvy.com [206.248.154.253]

Trace complete.

as you can tell there is packet loss but at linux.org's side... not teksavvy's

good luck in your tests

reply to Angelo
>works perfect here...

One traceroute is not statistically significant. Try 100, or 1000 or more.

For the last 48 minutes, from 2:50PM - 3:38 PM EST today (Sunday), sending pings every 5 seconds for a total of 576 pings:

I'm seeing packet loss rates of 0.3% (yahoo.com) to 0.8% (linux.org) on Teksavvy DSL while I get 0.0% (no loss) to the same hosts on Videotron cable.

Here's what I get from mtr on DSL:

mtr -r -c 100 yahoo.com
HOST: s Loss% Snt Last Avg Best Wrst StDev
1. d 0.0% 100 61.2 47.2 11.2 95.8 13.5
2. 206.248.154.102 0.0% 100 16.5 18.0 16.0 34.9 2.8
3. 65.39.134.1 21.0% 100 16.3 17.5 16.0 39.2 2.9
4. 65.39.198.250 1.0% 100 18.0 18.6 16.9 24.7 1.6
5. oc48-po7-0.tor-1yg-cor-1.pee 1.0% 100 17.7 18.7 17.0 24.6 1.6
6. 216.187.114.141 1.0% 100 29.9 29.6 28.1 40.1 1.9
7. exchange-cust1.ch1.equinix.n 1.0% 100 29.2 29.8 28.1 36.2 1.7
8. ge-2-0-9.p551.pat2.dce.yahoo 0.0% 100 52.0 43.6 39.9 59.9 3.5
9. ge-1-0-0-p100.msr1.dcn.yahoo 1.0% 100 42.3 43.1 39.7 56.0 2.7
10. ge9-3.bas2-m.dcn.yahoo.com 1.0% 100 45.1 43.0 39.8 55.2 2.5
11. w2.rc.vip.dcn.yahoo.com 0.0% 100 40.4 41.7 39.4 48.4 1.6

Note the 21% loss at 65.39.134.1 .

Here's the same from Videotron: (different host)

mtr -r -c 100 69.28.239.62
HOST: f Loss% Snt Last Avg Best Wrst StDev
1. i 0.0% 100 0.4 1.0 0.4 11.8 2.2
2. 10.192.128.1 0.0% 100 8.7 8.4 5.6 19.4 2.7
3. 24.200.230.166 0.0% 100 6.9 8.1 5.6 17.3 2.4
4. 24.200.250.21 0.0% 100 17.4 11.7 8.4 27.3 3.8
5. 216.113.122.41 0.0% 100 11.2 11.7 8.3 26.6 3.4
6. ia-cnnu-bb04-ge14-0-0.vtl.ne 0.0% 100 12.2 11.8 8.8 27.8 2.9
7. videotron-gw.peer1.net 0.0% 100 9.9 12.0 8.9 21.4 2.8
8. 216.187.90.57 0.0% 100 13.2 13.2 10.0 30.1 3.0
9. peer1-mtl-gw.metrooptic.com 0.0% 100 13.3 18.5 11.8 114.1 15.7
10. ATM-OC3POS1-0.gw-exchange.el 0.0% 100 13.0 16.2 12.1 81.2 9.6
11. ip62-dc.electronicbox.net 0.0% 100 12.9 18.3 12.0 120.6 13.3

Looks clean as a whistle on Cabletron.

Here's yahoo.com from Videotron:

mtr -r -c 100 yahoo.com
HOST: f Loss% Snt Last Avg Best Wrst StDev
1. i 0.0% 100 0.4 0.8 0.4 7.8 1.5
2. 10.192.128.1 0.0% 100 6.6 7.9 5.5 42.5 4.1
3. 24.200.230.166 0.0% 100 9.7 8.0 5.6 20.0 2.9
4. 24.200.250.21 0.0% 100 9.4 11.1 8.2 23.9 2.9
5. ia-piex-bb04-ge6-1-0.vtl.net 0.0% 100 11.6 10.7 8.5 24.0 2.6
6. ia-piex-bb04-pos11-0-0-cpe08 11.0% 100 20.9 29.6 17.4 224.7 37.6
7. sl-bb26-nyc-9-0.sprintlink.n 0.0% 100 21.6 23.5 17.3 197.4 22.1
8. 144.232.8.194 0.0% 100 18.4 27.2 17.3 220.7 31.9
9. tbr2.n54ny.ip.att.net 0.0% 100 32.5 36.1 31.4 224.8 19.3
10. 12.122.16.213 0.0% 100 31.7 32.7 30.6 42.4 1.9
11. cr2.wswdc.ip.att.net 0.0% 100 32.1 33.0 30.7 48.0 2.8
12. 12.122.16.6 0.0% 100 32.4 34.1 30.9 50.3 3.4
13. 12.122.113.17 0.0% 100 31.8 36.3 30.7 206.9 19.2
14. 12.86.111.22 0.0% 100 23.8 26.4 22.8 159.4 13.7
15. ge-0-0-0-p130.msr2.dcn.yahoo 0.0% 100 24.0 26.0 22.9 102.2 8.1
16. ge7-2.bas2-m.dcn.yahoo.com 0.0% 100 42.7 25.5 23.0 42.7 3.6
17. w2.rc.vip.dcn.yahoo.com 0.0% 100 25.1 26.5 22.7 137.3 11.6

Hmm, looks like ia-piex-bb04-pos11-0-0-cpe08 is dropping 11%. (?)

you need to remember that up to 15% loss is good. Anything more shows a problem with the path / over flooded router. Teksavvy doesn't proitize icmp... to my knowledge anyways. You could try teksavvy's prem service which gives you the best pings possible.

reply to hiremichael

reply to Angelo
>you need to remember that up to 15% loss is good. Anything more shows a problem with the path / over flooded router.

Personally, although TCP is robust and can handle 15% loss, I think 15% is too high, especially for VOIP. In any case, the most I've seen since Friday at noon was about 2.5% randomly distributed over 48 minutes and that's not TOO bad, IMO.

>You could try teksavvy's prem service which gives you the best pings possible.

Apparently, I am already on Premium.

then i dunno what to tell you.. every isp will have some loss. Good luck

>then i dunno what to tell you.. every isp will have some loss. Good luck

Thanks for your info and feedback. Yeah Videotron has virtually 0% packet loss, at least until something goes wrong, then it's probably 20-100%.

Since Friday noon I see average Internet loss rates of 0.3-0.4% on DSL, peaking at 2.5% last night during Internet download prime-time. Of course I've also had 14 Bell blameable DSL outages of 2-36 hours over the last 4 months where my loss rate was 100%.

I'm pondering keeping both internet connections, for continuous uninterrupted Internet and VOIP. They all go down sooner or later, planned or unplanned, for a few minutes or days-weeks.

Based on packet loss and latency though, if I went with one provider I think I'd keep Videotron. If I keep 2 providers, I will likely make Videotron primary and Teksavvy DSL for backup.

The DSL COULD also be used for high-volume transfers if I went with a Videotron capped package, but I think will go for the $40/month ($30/month for first year) business 1.5 MBps unlimited package.

1.5 MBps is fine for my needs, and I'd prefer that to the 7 MBps with 20/10 gb caps, where I'd constantly have to babysit my downloads to avoid the outrageous Videotron per GB over-cap charges.

AFAIK Teksavvy doesn't shape, their router (as well as Peer1's and some other large ISP's ISP out there) have a max rate for ICMP packets hitting that specific router. So if you ping your destination, that would be an accurate indication of packet loss, unless your destination also rate limits ICMP packets.

The real question is: are you seeing actual effects of packet loss or just seeing what you think is loss on pings/traceroutes?