dslreports logo
site
 
    All Forums Hot Topics Gallery
spc

spacer




how-to block ads


Search Topic:
uniqs
783
share rss forum feed

raytaylor

join:2009-07-28
kudos:1

Microtik and IPv6

Here is a question for you

I hear microtik has some fancy routing protocol that doesnt require subnets to be equally divided and ip space wasted - eg. i can have odd numbers in each subnet. Does this protocol actually exist and what is it called?

Also will the 64 mb of ram in a RB2011L-IN be able to handle this protocol for the v4 subnet routing, as well as routing for the ipv6 when ubnt supports it in my rooftop radios?

I am building a routed loop / ring network for redundancy of 10 AP sites so that if one site goes down, data can flow the other way around the loop to get to its destination.

Is there a better way to do this with microtik hardware than manually programming the default 0.0.0.0 route gateway and its backup gateway / alternate path home? Like a dynamic routed mesh?
If this dynamic routing exists for microtik what is it called?
Will it support IPv6?

I have two RB1100's and an RB750 to test with so once i know exactly what protocols i need to read up on, i plan to hit the wiki's and youtube and get into this microtik stuff. Have been holding off for 3 years now but might be getting a government grant so my cheap bridged methods need to go for a more professional system and also to prioritise voip on the backhaul links rather than relying on airmax to do it - its failing.

Kinda wish UBNT was a bit more ahead than they are with their routers.

raytaylor

join:2009-07-28
kudos:1
I think it was ospf i was thinking of
Tell me if i am right in the way i am thinking:

I have only got a /22 of v4 ip space so i wanted to try and conserve ip addresses.
So i can have the sectors at each site bridged and just assign blocks of 32 ip's as i need them to each router (AP site) to issue to its local clients via its 3 sectors. Three ports on the router would be bridged for the sectors so the ip's can be used on any sector at that site without wasting IP's.

Now i am going to continue watching videos to see if OSPF will instantly re-route when a backhaul link goes down without a link state announcement.


Rhaas
Premium
join:2005-12-19
Bernie, MO
That is correct, you can route a /27 to each of your AP sites. You don't mention how you are giving the IP's to your end customer - Statically assigned, DHCP or PPPoE. You can create a bridge interface called AP'S and add the interfaces that your AP's are connected to to this bridge. Add your IP to the bridge interface an viola

I assume you mean link state being the physical interface state.

Re-convergence is slower without the physical interface being down as it must wait until the dead timer is exhausted. Most high end radio's will cause the ethernet port to shut off if the rf link is down. That will cause re-convergence to begin immediately.
--
I survived Hale-Bopp!

raytaylor

join:2009-07-28
kudos:1
reply to raytaylor
Thanks for your reply

They will be ubnt rockets for the WDS backhauls so if the wireless link goes down, the ethernet will stay up. I am hoping it has some way of detecting that it cant talk to the next router in the loop and then cut the link.
I have seen a feature request in the ubnt forums that if a wireless link goes down, they want the rocket to cut the ethernet port also - i assume this is what you are talking about for the physical interface.

He said this was because he didnt want to wait for OSPF to detect the link was down and needed it to respond instantly, where as i dont mind if it goes down for 30 seconds while it waits for some sort of ping? to fail and then start using another path, while rechecking the link every minute or so to see if it is back up.

Cant seem to find any specific info on how this would work if the ethernet port stays up (to the radio) but the radio link goes down.

jcremin

join:2009-12-22
Siren, WI
kudos:3
I use OSPF for failover for my two links to my upstream provider. One is a direct PTP shot which plugs into my core router, and the other is a shot to one of their other towers which runs over a VPN back to the same core router to make it look like both links are simple PTP shots.

I have the hello interval set to 1 second, and the router dead interval set to 5 seconds. That basically means that it will check the link every second to make sure it is still online, and if it doesn't respond for 5 seconds, it will failover to my other link.

I tested it down to 2 or 3 seconds, which worked, but caused some excessive flopping if I had any packet loss. 5 seconds has worked will for the last 2 years. You'll drop a few pings, but other than a little delay waiting for a page to load, the failover typically happens before most streaming video sites even run out of their buffer.


Rhaas
Premium
join:2005-12-19
Bernie, MO
reply to raytaylor
Click for full size
OSPF has a 'hello' packet that is sent at a set interval (10 seconds by default). If a 'hello' packet is not recieved by the time the 'dead' (40 seconds by default) time limit is reached the link is considered to be down and OSPF will reconverge. This is what happens if the radio's are up but the RF portion is not passing traffic. So it will work in the way you are describing - basically the link is dead for 40 seconds, then the traffic reroutes.

Attached is a picture of my active & planned wireless backhauls. I use OSPF for the IGP.
--
I survived Hale-Bopp!

raytaylor

join:2009-07-28
kudos:1
reply to raytaylor
Thank you so much - this perfectly answers it for me.


Inssomniak
The Glitch
Premium
join:2005-04-06
Cayuga, ON
kudos:2
reply to raytaylor
I do it differently.

For really efficient use of IPs I tunnelled all the traffic back to one core router via VPLS tunnels. Then bridged all the VPLS tunnels to a PPPOE concentrator.
I used a bridge horizon on the interfaces in the bridge to keep traffic from leaving the core router.

Was easier to administrator for sure.
--
OptionsDSL Wireless Internet
»www.optionsdsl.ca

raytaylor

join:2009-07-28
kudos:1
reply to raytaylor
That was my original idea with PPPoE but i decided against it.

I am thinking of doing it with just OSPF as the network i want to build needs to be fast - Hoping to have 70mbits running between towers and up to 20mbit out to the end users - So i will be making good use of caching.

For this reason, i dont really want to be tunneling otherwise there is extra overhead, and i will be offering WAN networking services so i want everything to be nice and efficient without having to pass centralized concentration points.

Also its going to be simpler for the installers if i decide to use powercode - no need to register a radio within powercode as well as a PPPoE or tunnel etc.

I think the powercode system will let them boot up a CPE radio, go to the billing website, login as the customer and click "register my radio" and it then grants access to the internet.

jcremin

join:2009-12-22
Siren, WI
kudos:3
said by raytaylor:

i want to build needs to be fast

For this reason, i dont really want to be tunneling otherwise there is extra overhead

I haven't pushed that much traffic over a vpn before, but I do often push 25 megs using the cheapo 411 routerboards and they don't even flinch.

Try doing a few tests. You might be surprised at how little overhead there really is. Or maybe once you get to that much throughput it will bottleneck. I can't say for sure, but I wouldn't rule it out so quickly.

I actually used to be 100% against PPPoE. Every experience with it (before starting a WISP) was awful and gave me the impression that it had a lot of overhead. I decided to give it a try and was surprised to find that setting a queue at 1 meg tested EXACTLY at 1 meg and it was just that the other ISP's using PPPoE sucked, not that PPPoE itself had any problems.


Inssomniak
The Glitch
Premium
join:2005-04-06
Cayuga, ON
kudos:2
EoIP does have a lot of overhead.

But VPLS tunnels are very efficient, and MPLS runs very well over an OSPF network. The tunnels dont fall apart on re-route. There wont be any speed decreases with VPLS tunnels.

I have a full MPLS cloud network, I have one router at each PoP, I bridge all the customer APs at a tower to a VPLS tunnel leaving a router back to my core router, so I have full layer2 to my core router from the customer. Then PPPoE on that! Magical I say.
--
OptionsDSL Wireless Internet
»www.optionsdsl.ca

soportec
Premium
join:2006-01-06
reply to raytaylor
hello Inssomniak, what Mtu are you using on your links for the MPLS and VPLS to work well? I have all older UBNT gear on my back hauls and I was under the impression that id did not support a high enough MTU for the MPLS to work well. But I might be changing out that gear soon, as I use OSPF on my network and then PPOE for CPE. Right now I am setting up PPPOE servers at each pop to stay away from EOIP tunneling. Some Pop I have Vlaned back to main POP to see what is more efficient.
--
HONDURAS WISP Ubiquiti Sectors With 5ghz -Ubnt Backhauls , MTCRE,MTCUME,MTCTCE,MTCWE- Central American Ubiquiti Distro

raytaylor

join:2009-07-28
kudos:1
reply to raytaylor
Do these tunnels just add a piece of extra info to each packet, or does it run as a single connection through a router?

I ask because some sites will have netequalisers and need to differentiate between voip and large downloads, from the same client.

eg. It needs to see voip as seperate and doesnt get slowed down while bit torrent does.

If it just saw connections within a single stream like a VPN going through a router - then it would slow everything down for the whole client's connection.


Inssomniak
The Glitch
Premium
join:2005-04-06
Cayuga, ON
kudos:2
The net equalizer wouldn't see inside of the tunnels.


TomS_
Git-r-done
Premium,MVM
join:2002-07-19
London, UK
kudos:5

1 edit
reply to raytaylor
If MT had something like that, I think they'd be pretty valuable right now.

An existing protocol that can do something similar is PPP, which allows you to assign any arbitrary IP address to a PPP session, and use a routing protocol to distribute reachability information for that IP around the network. You can get about as close as possible to 100% efficiency that way.

PPPoE is an extension of, and allows you to do precisely that.

edit:
As for OSPF and IPv6, you need to run two instances of OSPF to distribute IPv4 and IPv6 around your network. OSPFv2 for IPv4, and OSPFv3 for IPv6.

IS-IS, which works pretty much exactly the same way as OSPF (same algorithm), can do both protocols with a single instance.