Forum - Using multiple ethernet connections :: SixXS

Using multiple ethernet connections

Shadow Hawkins on Monday, 27 May 2013 20:19:35

My network: Cox cable internet connected to a cable modem which is connected to a Cisco RVS4000 router (wired only). I have 4 computers running off this router. However, I'm only interested in using one with SixXS, this one is a Shuttle media server (has two eth ports) with a standard Ubuntu 12.04 install. I'm currently using Enlightenment 17.3 desktop. I would like to be able to use my sixxs tunnel on one interface and my normal ipv4 connection on the second interface. The sixxs tunnel would be used to serve my web sites while I could use the standard ipv4 connection to watch movies. This is preferable because using the sixxs tunnel results in a radical download speed drop. This speed drop wouldn't affect my web server much, but it would impact other services (such as streaming media). When I use Hurricane I put the ip of specific eth interface in the interfaces configuration, so it seems likely that I could seperate the functionality using Hurricane. However, when I use sixxs (using heartbeat) I don't see a way to accomplish this seperation. Maybe I need two tunnels, but how to direct them to a specific local network ip would be the problem. Perhaps that could be done with my router configuration? If you've ever tried this or just have any ideas about trying it, I'd appreciate your comments. Thanks, Gary

Using multiple ethernet connections

Jeroen Massar SixXS Staff

on Tuesday, 28 May 2013 01:59:35

I would like to be able to use my sixxs tunnel on one interface

Please note that a tunnel is a separate interface by itself. The only relation it has further to the 'ethernet' then is just where the IPv4 that the tunnel runs over is configured on.

This is preferable because using the sixxs tunnel results in a radical download speed drop.

Please see the FAQ for 'slow' and detail what you mean with a speed drop, there are many factors that can cause this. You mention a 'Shuttle media server', if that is something with a low-power cpu, and you use it in combination with AYIYA, then indeed it might be slow as AYIYA is high on CPU overhead (due to the crypto calculations used to verify if the packet is really coming from where it should be and also for signing them when sending them).

However, when I use sixxs (using heartbeat) I don't see a way to accomplish this seperation.

Ah, you are using heartbeat, then at least the above mentioned AYIYA signing/verifying cpu-overhead would not be the case. I honestly do not really see what you are trying to accomplish (maybe a picture of what you want would help). Please note that IPv4 and IPv6 are two different routing domains, and as noted above, a tunnel get it's own interface, as it is a tunnel. Also note that your "router" might already support IPv6 itself and that you could terminate the tunnel there.

Using multiple ethernet connections

Shadow Hawkins on Wednesday, 29 May 2013 05:40:03

Thank you for your response. Jeroen Massar wrote:

Ah, you are using heartbeat, then at least the above mentioned AYIYA signing/verifying cpu-overhead would not be the case.

I have read the FAQ on slow. I can contact Cox (isp), but I don't expect any cooperation (from past experiance). I switched to heartbeat, hoping that it would improve the speed issue. By speed drop I mean, for example (I have a 18Mb/sec isp cable connection): ipv4 ping at google.com reports 20ms time ipv6 ping at ipv6.google.com reports 300+ms time ipv6-test.com speed tests ipv4: 17.4 Mbps ipv6: 2.4 Mbps

I honestly do not really see what you are trying to accomplish (maybe a picture of what you want would help). Please note that IPv4 and IPv6 are two different routing domains, and as noted above, a tunnel get it's own interface, as it is a tunnel.

What I want to do is get an acceptable speed for media streaming and ftp remote backup. It doesn't seem that I can do that with a ipv6 tunnel with my current setup. If I can't then I need to use ipv4 for those types of needs and restrict ipv6 use to other less demanding needs (like serving web pages).

Also note that your "router" might already support IPv6 itself and that you could terminate the tunnel there.

According to the sixxs wiki my router (Linksys RVS4000) doesn't support ipv6 over ipv4 tunneling. But if terminating the tunnel at the router meant that all my computers on my lan would have to use ipv6, then I wouldn't want to do that at this point anyway. I don't mind experimenting with ipv6, but I wouldn't want the other household members to experiance problems as a result of my experimentation. Thanks again for trying to help. I think I need to think this through a bit more. I understand most of the concepts here but I'm not sure about how I can use them. If you have any more questions or ideas I'd be happy to here them. Gary

Using multiple ethernet connections

Carmen Sandiego on Sunday, 02 June 2013 19:03:12

Gary Witscher wrote:

It looks like you've just poorly chosen the PoP - when you ping google with v4 it goes you->ISP->google, with v6 tunnel broker it goes you->ISP->PoP->google, so if RTT from ISP to PoP is 300ms - add this delay to all your communication. Not only delay is added but also other PoP limitations. So just try another one. Otherwise HE is absolutely similar to heartbeat - SIT tunnel. Yes, HE is GRE + dynamic conf, while SIXXS' heartbeat is dynamic conf + GRE, but otherwise... you can always use linux advanced routing, like add the specific route with src IP to PoP on your box with aiccu - smth like # ip ro add <pop.ip.address> via <gateway.ip> src <ip.addr.of.eth1> or you can use rules, like # ip ru add to <ip.of.pop>/32 table 6 # ip ro add default via <gw> src <ip.of.eth1> table 6

Using multiple ethernet connections

Jeroen Massar SixXS Staff

on Sunday, 02 June 2013 21:20:52

Otherwise HE is absolutely similar to heartbeat - SIT tunnel.

I assume with "HE" you mean Hurricane Electric / Tunnelbroker.net.

Yes, HE is GRE + dynamic conf,

As you wrote, correctly, in the above line, it is SIT, aka Simple IP Tunnel, or better noted Protocol 41, directly putting IPv6 packets inside IPv4. GRE (protocol 46) does not come anywhere in this picture. The only "dynamic" thing HE has is that one can use a dyndns-style pull for updating the tunnel endpoint, still that is a static tunnel.

while SIXXS' heartbeat is dynamic conf + GRE,

Not at all. Again GRE does not come into the picture. Heartbeat tunnels are protocol 41 with next to it a system that notifies the PoP side where the tunnel resides. The advantage here being that the PoP knows where that endpoint is and if that endpoint is still there (no heartbeat and it disables the tunnel).

you can always use linux advanced routing,

Applying something like that does not solve the problem. The latency is there and will not go away unless a better path is established.

Using multiple ethernet connections

Shadow Hawkins on Monday, 03 June 2013 06:13:02

Jeroen Massar wrote:

Applying something like that does not solve the problem. The latency is there and will not go away unless a better path is established.

Is the latency PoP specific? If latency is related to the signal route then perhaps I should change my PoP. Following is a trace route of three USA PoP's from my laptop. The last (Pheonix), looks way better than the one I'm using. Would it help my "speed" if I was using this PoP?

traceroute uslax02.sixxs.net Los Angeles, California HopHostnameIPTime 1 1ThinkPad-T430.local192.168.1.102 0.110 1192.168.1.1192.168.1.1 1.822 1192.168.1.1192.168.1.1 1.239 210.143.0.110.143.0.1 10.021 3ip68-6-12-188.sd.sd.cox.net68.6.12.188 9.777 5fed1dsrj01-xe130.0.rd.sd.cox.net68.6.8.0 12.011 6langbprj01-ae1.rd.la.cox.net68.1.1.13 22.627 7ge3-0.4.cr01.lax04.pccwbtn.net63.218.51.13 15.418 8burstnet.tenge11-1.br01.lax05.pccwbtn.net63.218.42.102 88.246 9ec0-59.1a0201.laca02.hostnoc.net64.120.246.90 90.516 10uslax02.sixxs.net184.22.85.226 98.211 usmia01.sixxs.net Miami, Florida HopHostnameIPTime 1 1ThinkPad-T430.local192.168.1.102 0.082 1192.168.1.1192.168.1.1 1.822 1192.168.1.1192.168.1.1 1.369 210.143.0.110.143.0.1 10.490 3fed1sysc02-vlan500.sd.sd.cox.net68.6.11.102 10.179 5fed1dsrj02-xe130.0.rd.sd.cox.net68.6.8.4 9.253 6langbprj01-ae1.rd.la.cox.net68.1.1.13 17.381 7ge3-0.4.cr01.lax04.pccwbtn.net63.218.51.13 25.500 8arbinet.ge4-9.br01.mia02.pccwbtn.net63.218.113.170 104.146 9ec0-64.1a0201.mifl01.hostnoc.net64.120.243.130 106.810 10usmia01.sixxs.net184.22.119.138 104.754 usphx01.sixxs.net Phoenix, Arizona HopHostnameIPTime 1 1ThinkPad-T430.local192.168.1.102 0.063 1192.168.1.1192.168.1.1 2.059 1192.168.1.1192.168.1.1 1.988 210.143.0.110.143.0.1 14.012 3fed1sysc02-vlan500.sd.sd.cox.net68.6.11.102 9.882 5fed1dsrj02-xe130.0.rd.sd.cox.net68.6.8.4 11.639 7unknown.hwng.net69.16.191.65 18.607 8ve3211.r1.ph.hwng.net209.197.0.45 29.130 91-4.r2.ph.hwng.net69.16.191.74 25.170 10usphx01.sixxs.net209.197.5.66 25.492

TIA Gary

Using multiple ethernet connections

Carmen Sandiego on Monday, 03 June 2013 18:57:55

Jeroen Massar wrote:

> Otherwise HE is absolutely similar to heartbeat - SIT tunnel. I assume with "HE" you mean Hurricane Electric / Tunnelbroker.net.

Yes, HE is GRE + dynamic conf,

while SIXXS' heartbeat is dynamic conf + GRE,

Not at all. Again GRE does not come into the picture.

Of course, just a typo.

> you can always use linux advanced routing, Applying something like that does not solve the problem. The latency is there and will not go away unless a better path is established.

Latency I'm sure is bound to the PoP. SIT is done in a kernel and poses a little overhead over packet forwarding. Advanced routing simply allows you to pin tunnel endpoint to certain IP if your TIC info does not have this field (sIPv4_Local) which is always the case for heartbeat. As for when this could be needed - no idea in this particular case but i can imagine a handful of cases, including some which could change the path (ISP selection).

Using multiple ethernet connections

Jeroen Massar SixXS Staff

on Monday, 03 June 2013 21:41:35

you can always use linux advanced routing,

Applying something like that does not solve the problem. The latency is there and will not go away unless a better path is established.

Latency I'm sure is bound to the PoP.

I can only hope to assume that you mean that the latency is 'bound' to the PATH roundtrip back and forth to the PoP.

SIT is done in a kernel and poses a little overhead over packet forwarding.

Yes, so little, that it has no relevance at all when the latency is on the path between two endpoints.

Advanced routing simply allows you to pin tunnel endpoint to certain IP if your TIC info does not have this field (sIPv4_Local) which is always the case for heartbeat.

As for when this could be needed - no idea in this particular case but i can imagine a handful of cases, including some which could change the path (ISP selection).

You are assuming that someone has multiple distinct internet uplinks. There are very few people who do have this. As such, if you would have asked the original poster "do you maybe a multiple links, one which has a better path than there currently is", then in that case, it would have made sense to reply with "use advanced kernel routing", though actually one does not need that. Just setting an more specific route toward the PoP over that uplink would be sufficient, no special magic required. Very few people have multiple distinct uplinks though, and the ones that do typically know how to resolve issues like these.

Please note Posting is only allowed when you are logged in.