Really slow connection to GNAPS?

[Derek Atkins <warlord@…>]

Hi,

Trying to get from here to my home network (204.107.200.x) I seem to only get about 20KB/s even though I know the network there is significantly better than that. Looking at an MTR trace the problem looks like the link from 86.43.238.9 to res1-avici1-g1-15-1.gnaps.net is overloaded:

                                              Packets               Pings
 Host                                       Loss%  Last   Avg  Best  Wrst StDev
 1. rtrexta-wireless-open.meeting.ietf.org   0.0%   1.5   9.1   1.3 243.0  35.9
 2. 86.43.238.9                              2.2%   1.1  17.0   1.0 196.2  42.7
 3. res1-avici1-g1-15-1.gnaps.net            2.3% 112.7 106.6 104.6 120.8   3.3
 4. at-1-0-0-000.br1.res1.va.gnaps.NET       2.3% 105.9 107.4 105.3 119.7   3.4
 5. so-5-0-0-001.br1.nyc1.ny.gnaps.NET       0.0%  97.5  93.3  91.4 102.6   2.8
 6. so-7-0-0-001.br2.qcy1.ma.gnaps.NET       2.3%  99.7 100.0  97.2 140.1   6.8
 7. ge-2-1-0-000.ar1.qcy1.ma.gnaps.NET       0.0% 116.7 115.9  96.9 148.0  19.0
 8. link42-14.CENT.NET                       0.0%  94.9 119.6  94.9 296.2  43.8
 9. DOGBERT.IHTFP.ORG                        0.0% 115.8 135.7 107.5 356.4  40.1

Whereas if I try to get to MIT I get a much better routing via cogent:

                                              Packets               Pings
 Host                                       Loss%  Last   Avg  Best  Wrst StDev
 1. rtrexta-wireless-open.meeting.ietf.org   0.0%   3.2   2.9   1.2  18.3   2.8
 2. 86.43.238.9                              2.4%   3.0  11.9   1.0 137.1  30.9
 3. 195.66.224.185                           2.4%  15.4  18.2  14.7 103.0  11.3
 4. te4-4.mpd02.lon01.atlas.cogentco.com     1.2% 180.5  39.2  12.4 212.8  41.4
 5. te2-2.ccr04.jfk02.atlas.cogentco.com     2.4%  88.2  88.6  85.7 147.9   7.9
 6. te2-2.ccr02.bos01.atlas.cogentco.com     1.2%  91.7  94.0  89.8 314.6  24.8
 7. gi11-0-0.core01.bos01.atlas.cogentco.co  1.2%  94.6  91.7  89.8 139.2   5.6
 8. 38.20.40.22                              4.8%  91.6  93.8  90.9 135.7   5.2
 9. MIT.demarc.cogentco.com                  1.2% 383.3 140.7  94.6 565.5 106.9
10. amtgw.media.mit.edu                      1.2%  96.4 100.5  94.0 154.8  14.1

Is there some way we can control this routing or perhaps check on that link? It's been this way since I arrived on Saturday evening, and doesn't seem to change at all throughout the day.

Thanks!

-derek

[llynch@…]

Andrew -

Can you take a look at this?

• lucy

[rchandler@…]

Replying to Derek Atkins <warlord@MIT.EDU>:

Hi Derek,

Could you repeat your mtr from a wired connection. I want to isolate how much of this is due to the WAN versus wireless network. The link out is 200Mbps and there's only about 40Mbps on it. I'm seeing sone packet loss on the Internet but let's isolate how much of this is due to the IETF network first.

Regards,
Ross

Hi,

Trying to get from here to my home network (204.107.200.x) I seem to only get about 20KB/s even though I know the network there is significantly better than that. Looking at an MTR trace the problem looks like the link from 86.43.238.9 to res1-avici1-g1-15-1.gnaps.net is overloaded:

                                              Packets               Pings
 Host                                       Loss%  Last   Avg  Best  Wrst StDev
 1. rtrexta-wireless-open.meeting.ietf.org   0.0%   1.5   9.1   1.3 243.0  35.9
 2. 86.43.238.9                              2.2%   1.1  17.0   1.0 196.2  42.7
 3. res1-avici1-g1-15-1.gnaps.net            2.3% 112.7 106.6 104.6 120.8   3.3
 4. at-1-0-0-000.br1.res1.va.gnaps.NET       2.3% 105.9 107.4 105.3 119.7   3.4
 5. so-5-0-0-001.br1.nyc1.ny.gnaps.NET       0.0%  97.5  93.3  91.4 102.6   2.8
 6. so-7-0-0-001.br2.qcy1.ma.gnaps.NET       2.3%  99.7 100.0  97.2 140.1   6.8
 7. ge-2-1-0-000.ar1.qcy1.ma.gnaps.NET       0.0% 116.7 115.9  96.9 148.0  19.0
 8. link42-14.CENT.NET                       0.0%  94.9 119.6  94.9 296.2  43.8
 9. DOGBERT.IHTFP.ORG                        0.0% 115.8 135.7 107.5 356.4  40.1

Whereas if I try to get to MIT I get a much better routing via cogent:

                                              Packets               Pings
 Host                                       Loss%  Last   Avg  Best  Wrst StDev
 1. rtrexta-wireless-open.meeting.ietf.org   0.0%   3.2   2.9   1.2  18.3   2.8
 2. 86.43.238.9                              2.4%   3.0  11.9   1.0 137.1  30.9
 3. 195.66.224.185                           2.4%  15.4  18.2  14.7 103.0  11.3
 4. te4-4.mpd02.lon01.atlas.cogentco.com     1.2% 180.5  39.2  12.4 212.8  41.4
 5. te2-2.ccr04.jfk02.atlas.cogentco.com     2.4%  88.2  88.6  85.7 147.9   7.9
 6. te2-2.ccr02.bos01.atlas.cogentco.com     1.2%  91.7  94.0  89.8 314.6  24.8
 7. gi11-0-0.core01.bos01.atlas.cogentco.co  1.2%  94.6  91.7  89.8 139.2   5.6
 8. 38.20.40.22                              4.8%  91.6  93.8  90.9 135.7   5.2
 9. MIT.demarc.cogentco.com                  1.2% 383.3 140.7  94.6 565.5 106.9
10. amtgw.media.mit.edu                      1.2%  96.4 100.5  94.0 154.8  14.1

Is there some way we can control this routing or perhaps check on that link? It's been this way since I arrived on Saturday evening, and doesn't seem to change at all throughout the day.

Thanks!

-derek

[anonymous]

I will attempt to do so this afternoon. However I was getting similar performace both from the 802.11(a) network and also from the network in my hotel room.

[Derek Atkins <warlord@…>]

Er, that was from me, obviously.

-derek

[rchandler@…]

Replying to anonymous:

I will attempt to do so this afternoon. However I was getting similar performace both from the 802.11(a) network and also from the network in my hotel room.

There's a problem in part of the hotel room network that's been worked on separately. Have you tried from the wired connection in the terminal room?
That should isolate any issues down to just the Internet.

Ross

[anonymous]

Okay, from the IETF wired network:

                                              Packets               Pings
 Host                                       Loss%  Last   Avg  Best  Wrst StDev
 1. rtrexta-wired.meeting.ietf.org           0.0%   1.0   2.1   0.9   6.6   1.8
    130.129.48.30
 2. 86.43.238.9                              0.0%   0.7   0.6   0.6   1.0   0.1
 3. res1-avici1-g1-15-1.gnaps.net            0.0% 104.2 104.3 104.0 104.8   0.2
 4. at-1-0-0-000.br1.res1.va.gnaps.net       0.0% 105.2 105.2 104.8 105.5   0.2
 5. so-5-0-0-001.br1.nyc1.ny.gnaps.NET       0.0%  89.0  89.1  88.7  90.9   0.5
 6. so-7-0-0-001.br2.qcy1.ma.gnaps.NET       0.0%  96.0  94.8  94.3 100.2   1.2
 7. ge-2-1-0-000.ar1.qcy1.ma.gnaps.net       0.0%  96.6  96.8  96.4  97.5   0.3
 8. link42-14.cent.net                       0.0%  97.0  96.8  96.4  97.3   0.2
 9. DOGBERT.IHTFP.ORG                        0.0% 106.5 107.3 106.5 114.9   1.7

(let me know if you want the trace to MIT as well)

[rchandler@…]

Replying to anonymous:

I see no loss on that one to your home network. Is is slow to it still from the wired net?

From the wired net to MIT I see the loss confined to Cogent's network.

My traceroute [v0.73]

noc.meeting.ietf.org (0.0.0.0) Mon Jul 28 13:25:17 2008
Keys: Help Display mode Restart statistics Order of fields quit

Packets Pings

Host Loss% Snt Last Avg Best Wrst StDev?

1. rtrexta-server.meeting.ietf.org 0.0% 33 1.1 2.5 0.3 20.6 3.9
2. 86.43.238.9 0.0% 33 1.1 20.2 0.6 202.2 50.9
3. 195.66.224.185 0.0% 32 16.9 15.7 14.3 28.8 2.6
4. te2-2.mpd03.jfk02.atlas.cogentco 0.0% 32 82.2 84.9 82.1 118.9 8.5
5. te2-4.mpd01.bos01.atlas.cogentco 12.5% 32 86.4 90.3 86.1 158.0 14.2
6. gi2-0-0.core01.bos01.atlas.cogen 0.0% 32 86.2 86.2 84.9 87.7 0.4
7. 38.20.40.30 0.0% 32 87.4 93.5 86.8 254.9 29.5
8. MIT.demarc.cogentco.com 0.0% 32 91.1 162.2 90.8 718.9 154.7
9. amtgw.media.mit.edu 0.0% 32 91.2 92.1 90.9 103.3 2.4

I've asked Eircom to raise a ticket with Cogent.
Regards,
Ross

Okay, from the IETF wired network:

(let me know if you want the trace to MIT as well)

[Derek Atkins <warlord@…>]

Right now it actually looks like there's a problem within gnaps:

                                              Packets               Pings
 Host                                       Loss%  Last   Avg  Best  Wrst StDev
 1. rtrexta-wireless-open.meeting.ietf.org   3.2%  12.3  18.1   1.5 423.9  53.4
 2. 86.43.238.9                              2.4%   1.6  24.4   1.2 312.9  53.8
 3. res1-avici1-g1-15-1.gnaps.net            2.9% 105.2 116.3 104.8 379.2  29.6
 4. at-1-0-0-000.br1.res1.va.gnaps.net       1.8% 106.6 117.6 105.5 872.9  47.1
 5. so-5-0-0-001.br1.nyc1.ny.gnaps.NET      60.9% 100.5 103.2  92.0 281.1  21.0
 6. so-7-0-0-001.br2.qcy1.ma.gnaps.NET      60.6%  99.3 108.2  97.4 216.4  21.2
 7. ge-2-1-0-000.ar1.qcy1.ma.gnaps.NET      59.8%  98.7 117.4  97.3 408.3  41.5
 8. link42-14.CENT.NET                      60.6% 134.2 123.5  95.0 629.6  79.1
 9. CBC-GW.IHTFP.ORG                        60.1% 100.3 130.0  99.2 603.3  67.8

But I dont think this is our fault unless our reverse route is what's flapping between va and ny? I doubt this is eircom's problem. :-(

[anonymous]

FYI, I have verified that the network slowness is now gone. You can close this ticket out. Thank you!

[Derek Atkins <warlord@…>]

• grumbles about Trac not keeping login state *

-derek

[rchandler@…]

Replying to anonymous:

Hi Derek,

Closing this ticket.

Thank you.

Ross

[(none)]

Milestone IETF Week ieft 72 deleted