Hurricane Electric's IPv6 Tunnel Broker Forums

Tunnelbroker.net Specific Topics => Questions & Answers => Topic started by: simplepanda on November 30, 2014, 08:43:49 PM

Title: Rogers / Canada performance issues
Post by: simplepanda on November 30, 2014, 08:43:49 PM
All,

I (and others) are having a lot of performance problems with IPv6 tunnels from Rogers here in Canada. Rogers is a pretty big DOCSIS ISP based in Ontario (about 2M customers).

The issue we're all having is that while IPv6 tunnelling works fine, performance is fairly poor. For example, on 250Mbps DOCSIS connections speeds over IPv4 test at ~300Mbps (Rogers has some local node speed boosting). IPv6 tunnelling, on the other hand, rarely exceeds 20Mbps or so regardless of what tunnelbroker.net server being used.

This behaviour can't be reproduced on other Canadian ISP's. Bell Canada, TekSavvy, Videotrom etc all have tunnel performance that matches IPv4.

More information in this thread:
http://www.dslreports.com/forum/r29521047-Internet-Rogers-IPV6-6rd-

Doing some poking around, I'm seeing some odd routing if you're a Rogers customers:

Rogers (Toronto HFC 60/10) to tunnelbroker.net Toronto:

1. pandacore                                                  0.0%     7    1.0   1.3   0.8   3.4   0.7
2. ???
3. 209.148.245.173                                            0.0%     7   14.1  17.3   8.8  36.6   8.9
4. 69.63.248.181                                              0.0%     6   23.8  16.8  12.7  23.8   4.8
5. 69.63.249.26                                              16.7%     6   24.6  27.5  24.6  30.4   2.3
6. 10gigabitethernet2-2.core1.ash1.he.net                     0.0%     6   29.5  44.6  29.5 104.0  29.3
7. 100ge5-1.core1.nyc4.he.net                                 0.0%     6   37.4  33.7  27.2  40.4   5.5
8. 100ge9-1.core1.tor1.he.net                                 0.0%     6   27.8  26.5  24.2  31.1   2.5
9. tserv1.tor1.he.net                                         0.0%     6   30.2  28.8  26.1  31.0   1.7

Note that 69.63.249.26 is a Rogers IP. So getting to Toronto is slow because it's going to Ashburn, then NYC, then back to Toronto. I thought Rogers peered with HE.net at TORIX, but maybe not.

That said, this looks like Rogers peers with HE in Ashburn. Great! I'll use the Ashburn tunnelbroker.net server. However...

Rogers (Toronto HFC 60/10) to tunnelbroker.net Ashburn:

1. pandacore                                                  0.0%     6    1.0   2.2   0.9   5.8   1.9
2. ???
3. 209.148.245.177                                            0.0%     6   32.9  21.2  12.9  32.9   8.9
4. 69.63.248.185                                              0.0%     6   23.7  18.6  13.1  24.3   5.8
5. 24.156.144.182                                             0.0%     5   20.7  22.7  20.7  25.3   1.6
6. 10gigabitethernet4-1.core1.chi1.he.NET                     0.0%     5   34.4  28.2  22.0  36.6   6.8
7. 10ge15-7.core1.ash1.he.net                                 0.0%     5   47.0  44.0  32.5  52.3   7.3
8. tserv2.ash1.he.net                                         0.0%     5   33.1  36.3  30.2  45.3   6.3

Lovely. Note that 24.156.144.182 is a Rogers IP, so it looks like Rogers also peers with HE in Chicago, and for whatever reason, getting to Ashburn on Rogers involves going through Chicago. Strange, because as above, you can get to Ashburn directly.

Fine then, I'll use the Chicago peer point.

Rogers (Toronto HFC 60/10) to tunnelbroker.net Chicago:

1. pandacore                                                  0.0%     7    0.9   1.0   0.8   1.4   0.0
2. ???
3. 209.148.245.173                                            0.0%     7   17.2  15.1  11.5  24.1   4.4
4. 69.63.248.181                                              0.0%     7   12.8  15.9  12.1  21.5   3.1
5. 69.63.249.26                                              50.0%     7   34.0  32.3  26.8  36.1   4.8
6. 10gigabitethernet2-2.core1.ash1.he.net                     0.0%     7  114.8  84.4  26.4 198.2  61.3
7. 10ge1-2.core1.cmh1.he.net                                  0.0%     7   36.9  66.1  36.9 153.7  42.5
8. 10ge1-2.core1.chi1.he.net                                  0.0%     7   40.4  48.3  36.5  86.6  17.5
9. 184.105.253.14                                             0.0%     6   37.7  37.4  34.5  42.4   2.6

So when you try to get to directly to HE.net's Chicago POP, you go back through Ashburn.

So none of this is terribly useful. Let's look at hop #7 in the initial trace to Toronto. That's HE.net in NYC.
Let's trace to tunnel broker in NYC then, that seems to actually be fairly close to us...

1. pandacore                                                  0.0%     4    0.8   1.6   0.8   3.5   1.0
2. ???
3. 209.148.245.173                                            0.0%     4   13.9  14.4  12.6  16.4   1.5
4. 69.63.248.181                                              0.0%     4   11.6  13.6  11.6  14.7   1.3
5. ???
6. 10gigabitethernet2-2.core1.ash1.he.net                     0.0%     4   35.7  34.5  25.3  38.6   6.2
7. 100ge5-1.core1.nyc4.he.net                                 0.0%     3   30.6  30.6  29.4  31.8   1.0
8. tserv1.nyc4.he.net                                         0.0%     3   29.7  30.6  29.7  31.1   0.7

Nice. First we go to Ashburn.

So unless there is something I'm missing here, I'm wondering if the performance problems to tunnelbroker.net aren't because of some of the curious routing that's going on here.

Thoughts? I'd really like to get good performance to tunnelbroker.net and Rogers has been less than helpful on this issue (customer support declares this issue to not be something they support and they refuse to escalate).