I am busy setting up a lab to test various MPLS configurations. You can check my results so far on the WIKI: http://wiki.mikrotik.com/wiki/MPLS_Lab_Setup

I am hoping to start up a conversation around this to test the capabilities of MPLS and see how it operates with regard to RouterOS

Please give me any feedback and also participate in the wiki as much as you can.

Does anyone know if there is any information on the Traffic Engineering support included in the 3.7 version of ROS? A command line reference maybe or any kind of documentation.

Support for TE in 3.7 mpls package is rather limited - although it supports TE
tunnels, there are limited abilities to direct traffic over them (e.g. no way
to direct VPLS or MPBGP IP VPN traffic over TE tunnels). As of version 3.8
new mpls-test package will be introduced that together with routing-test will
support more TE and IP VPN features. Wiki article on these features is in
progress.

Support for TE in 3.7 mpls package is rather limited - although it supports TE
tunnels, there are limited abilities to direct traffic over them (e.g. no way
to direct VPLS or MPBGP IP VPN traffic over TE tunnels). As of version 3.8
new mpls-test package will be introduced that together with routing-test will
support more TE and IP VPN features. Wiki article on these features is in
progress.

Thats great - at this point I am not considering this for a production environment. The lab setup is for testing purposes to see how it might scale up for future use. I look forward to the Wiki article!

I too have setup an MPLS test lab, but I'm struggling with getting TE working. Has there been any progress since the last post to this thread in April? Any sign of the promised documentation yet? The documentation in the Wiki on MPLS and VPLS is good, but it doesn't detail how to setup TE tunnels.

RouterOS v3.13 seems to have MPLS built into the distribution and maybe with earlier releases too. Is there any difference between the bundled MPLS package and the MPLS-test/Routing-test packages that can be downloaded seperately?


Regards

Chris Macneill

please read uldis post on this thread - you will see where is the difference.

all development happens in mpls-test and routing-test packages, and you have to install them both to use them.

OK, I've got the mpls and routing test packages installed, but is there any quick information you can give me on how to configure TE tunnels? I'm not looking for full blown Wiki documentation, even if you can just give me a config dump of what you need to enter to get a TE tunnel to work. If you give me a text dump from a router with TE tunnels configured I'm sure I can work it out from there.

From information I've looked through it would appear that TE tunnels are independent of VPLS. Is this correct or do you have to configure VPLS first? The only documentation I can find is for Cisco routers and I can't work out how to implement this in RouterOS.

If I can get it working, I'm more than happy to contribute to a Wiki entry that will help others.

I've gone through all the CLI commands, but it's not obvious how to proceed.

Help please I'm desperate!!


Regards

Chris Macneill

Simple setup along with commands necessary to configure routers is described here:
http://wiki.mikrotik.com/wiki/MPLS_TE_Tunnels

Yes, you are right, TE tunnels are independent of VPLS. Just like LDP based label distribution, all TE tunnels do is - establish label switched path. LDP establishes LSP for every route in network, TE tunnels establish LSPs as required (configured). Established LSPs then can be used for forwarding data, one of potential "users" is VPLS.

If you want further assistance on configuring TE tunnels, please describe the setup you intend to have.

Many thanks Mplsguy, I don't know why I couldn't find this in the Wiki before!

The golden nugget is:-
I'd pretty much worked things out from the VPLS stuff and by trial and error, but nothing was connecting.

I'm not sure where you're located, I'm GMT+12 so it's late evening here currently. I'll try this out in the morning and let you know if I need any other assistance.

Currently I have a basic test lab scenario with 3 backbone nodes in a triangle, two "customers" connected to one node, the "ISP" connected to a different node and the third node acting as a backup link. I'm trying to model the behaviour when the primary link fails. Our live scenario has links with differing bandwidth and I want to determine whether TE Tunnels will apportion the available bandwidth proportionately when the customers are working over the slower backup link. The primary link is 100Mb/s, the backup link is only 10Mb/s.


Regards

Chris Macneill

You should take into account that TE tunnels themselves do not limit/manage traffic. The "bandwidth" that is specified for TE tunnel is more like administrative value, it affects if bandwidth reservation on some router can be done or not. If some TE tunnel has reserved bandwidth 10Mbps, routers along the way do not limit bandwidth for particular stream to 10Mbps. If you want to really hard limit amount of data sent over TE tunnel, you should make sure that no more data than that limit enters TE tunnel (you can do this by queues on TE ingress router)

If in your setup both customers have their traffic forwarded along TE tunnels (over main backbone link) that each has bandwidth 10Mbps, once you will break main backbone link, only one tunnel will be rerouted over backup link (because first tunnel will "reserve" 10Mbps and there will be no more "bandwidth" for second tunnel).

Another thing you should consider is that RouterOS does not yet implement MPLS local protection (fast reroute) - change of traffic to backup link will only happen when IGP state will change and tunnels will get renegotiated.

I'd pretty much come to the conclusion that TE tunnels weren't really going to achieve the throttling objective over the slow backup link.

I guess the time for IGP to re-establish routing following a link failure is in the order of a few minutes, TE fast reroute I suppose will reduce that to seconds? Even if a few minutes, this is probably quicker/preferable to the current arrangement of manual bypass!

I only got my test lab working properly late this afternoon, a quick test by unplugging the active link did appear to cause a re-establishment faster than I could run the test, i.e. in the order of a few seconds. I need to do more testing to confirm my initial results.

Would a bridged mesh of VPLS tunnels solve the problem of the low bandwidth backup link? Could queues be setup with different bandwidth allocations to apportion the available bandwidth by link? This would use (R)STP to prevent a loop and weighting factors should force the slow speed link to be the hot standby.

The Wiki page for TE tunnels has a confusing statement in the section "Forwarding Traffic onto TE tunnels"

Note that RSVP TE tunnels are unidirectional - it is not necessary to have matching tunnel for reverse direction on tail-end router.

Surely if the tunnels are unidirectional then a matching tunnel MUST be created for the reverse direction on the tail-end router?

Or does this mean TE tunnels are unidirectional in setup, but bidirectional in data transfer?


Regards

Chris Macneill

Fast reroute (remember that RouterOS does not implement this yet) should forward traffic over backup tunnel in milliseconds, not seconds or minutes. Also take into account that fast reroute is not re-establishing existing tunnels - you have to set up backup tunnels around "vulnerable" links or nodes in tunnel path and on link/node failure traffic will be sent over detour path (actually main tunnel will also be maintained over this backup link so that it does not time out).

Time until TE tunnels converge to new topology depends on type of failure. E.g. if router detects removal of cable (supposedly your test case), RSVP/IGP can react immediately - in this case tunnel will be re-established in few seconds. If failure detection will happen based on protocol timeout (e.g. OSPF or RSVP), it will take more.

What exactly is the problem with slow backup link? In order to control how bandwidth is divided among traffic streams when sending over slower backup link you can mark packets based on which customer they came from and setup queue on backup interface (using proper limit-at and max-limit values). I do not see how VPLS mesh fits here. As to establishing TE tunnels - you just have to make sure that they both have chance to get established over backup link (configured bandwidth).

TE tunnels are unidirectional both - in setup and in data transfer. The confusing statement is supposed to say that you do not need to have tunnel in reverse direction in order for first tunnel to work and transfer data. Of course you should set up tunnel in reverse direction if you want data to be transferred both directions over TE tunnels, but that is not a requirement to have bidirectional communications - e.g. data can be sent in reverse direction over LDP established LSP.

Thanks for clarifying that, I'll play around a bit with various scenarios now that I have a semi stable test platform and I'll post here again if I need further help.

Regards

Chris Macneill

I have some questions regarding the examples in the Wiki on setting up MPLS, these are more to do with OSPF, but they do arise from the MPLS examples:-

1. The IP addresses used as loopback addresses are public addresses, 9.0.0.0/8 belongs to IBM, should these not actually be in a private address range, i.e. 10.0.0.0/8, 172.16.0.0/16 or 192.168.0.0/24 ? I can understand that these are examples, but shouldn't they be correct examples?
2. Setting up the test lab as per the Wiki examples I have some routes shown in the OSPF routing tables with Area as "unknown", most notably the loopback addresses, is this OK or should the address ranges have Area definitions? Everything appears to work OK, it just looks wrong!

Regards


Chris Macneill

As to the addresses in wiki examples - yes, those are provided just as an example, the particular addresses used do not have any meaning, the only intention is so that it is easier to distinguish the networks and to make it easier to relate console printouts to network diagram. Of course, the same setup can be implemented using "correct" private space addresses, still I don't think examples can be considered wrong.

Please post OSPF related console outputs that "look wrong" so that it can either be explained and documented or fixed if necessary.

It may just be necessary to make it clearer that the 9.x.x.x loopback addresses should be changed to something meaningful in a private address space when migrating from a test to production environment. Someone somewhere is bound to leave the 9.x.x.x addresses and some time down the line may run into weird routing problems! I realise the kind of people that are likely to be working with MPLS are going to be pretty advanced and work this out for themselves, but there will always be one that doesn't!

I've attached a JPEG of the OSPF routing screen for one of my nodes. The problem is only apparent in Winbox and may be that it's just a bug there with the routing-test package, I'm running v3.14. The CLI version of the data doesn't show Area information at all.

I have configured Area entries for the 9.0.0.0/8 network and 192.168.3.0/24, but they still show up as "unknown" even after a reboot. Node 1 is connected via another RB532 to the company's local network (192.168.3.0/24) via 192.168.155.0/24. 192.168.55.0/24 is also on the "gateway" RB532 as a local test network. The networks that connect my three test nodes are 192.168.12.0/24, 192.168.23.0/24 and 192.168.13.0/24. I'm using NAT on the "gateway" node so that I appear as a single 192.16.3.0/24 address to the company network, everything else is routed.

Regards

Chris Macneill

hi, just playing around with mpls and routerOS, its working great but i have some questions..

how can i direct traffig from R1 to R5 through the TE tunnel ( without VPLS/VPN)

for an example. redirect udp traffic with high priority trough TE tunnel.

thanks!

gustkiller, as TE tunnels are interfaces, you can forward traffic onto them using routing (or policy routing). For this you will have to add IP address to head end of tunnel (just to enable it to be used for IP routing) and then use "gateway=<TE tunnel>" setting for route.

Hello

Is there anybody who can assist me with Forwarding Traffic into VPLS PWs/TE tunnels.

Ive attached a screen shot of my provider edge (PE router) node which is part of a VPLS network.

A computer was attached to CE router and was streaming a video. The video was recieved at the destination node but as you can see from the screen shot there is no traffic traversing the VPLS link (VPLS1) between _pe1 and _pe2.

And yet there is traffic at the _pe - _ce interfaces.

Please assist me with forwarding the traffic into the VPLS PWs/TE tunnels.

Please help.

Kind regards

Allano

Apologies for resurrecting such an ancient thread, however I'm seeing exactly the same problem here and struggling to get my head around the configs or the wiki descriptions

This is using 6.13 and a mixture of RB450 and RB750 routers in a lab (with a view to rolling out multiple 1016, 1036 and 1072's on our backbone if we ever get this working)

We have in effect multiple links between 3 sites (triangle/ring) which we want to use as primary and secondary routes (we actually have up to 18 x waves (at up to 10 Gig each) between each location currently which we can LAG if necessary)

I'm trying to configure MPLS between sites, with the 3 core sites and then PE at each location (1016/1036) feeding customer racks

At the moment (in our lab) I can't get this to work ... as soon as I bring up the VPLS tunnels traffic stops routing. If/when I get the VPLS tunnels working I still can't get the TE to work sensibly (shows no traffic traversing the TEs)

What I'm trying to implement is something very similar to the design from the 2009 MUM presentation on MPLS however the Wiki all seem to talk about 3.8 and mpls-test and there doesn't seem to be a defacto / definitive guide on how to do this 5 years after the testing started? http://mum.mikrotik.com/presentations/U ... s_mpls.pdf