MPLS Basics and Disaster recovery

Unanswered Question

Having recently initiated a global MPLS backbone network across our three offices in UK, India and ME, a telecom company based in UK, has been choosen as the service provider and single Point of contact for all issues. Conectivity bandwidth is ranging from 1-2MB across 3 sites.

I have heard a lot about MPLS and one of it's benefits appears in the form of ability to define COS.

One more that's often talked about is ease of planning 'disaster recovery'.

I have read that In a MPLS network, traffic can be easily diverted to any other location on the network in the event a particular network node becomes unavailable. Of course there has to be a host available at the other selected node, but that comes with disaster recovery planning.

Can someone throw some details on how to really go about planning an efficient DR over an MPLS backbone?

Any/all comments are welcome.

I have this problem too.
0 votes
  • 1
  • 2
  • 3
  • 4
  • 5
Overall Rating: 0 (0 ratings)
Loading.
mounir.mohamed Tue, 02/19/2008 - 01:09

Are you talking about DR from the CE side or disaster recovery within the MPLS backbone?

Mounir, I tried to answer your question but really couldn't get it, pardon the ignorance.

We are the CE, and the MPLS connectivity comes with an SLA through ISP.

I am essentially looking at exploiting the existing MPLS link for something more than the routine mail / file exchanges across our branch offices spread across three continents.

Disaster recovery within MPLS backbone is probably not what I am looking for, at least to my understanding.

Joseph W. Doherty Sat, 02/23/2008 - 04:57

Are you performing MPLS end-to-end, or are you just running across an MPLS backbone which you see as IP? If the latter, who controls the CE router and what routing protocols are being used on the CE facing both in and out?

MPLS isn't end to end. We (the CE) have an SDSL line coming at the last mile. The ISP says it doesn't really matter if the terminating circuit (around 20 miles away in the next city) is MPLS.

Internal equipment is designed around Nortel edge and core switches and we are planning to run SMLT, a proprietary protocol, equivalent of HSRP/VRRP combine.

- So, effectively we are just running MPLS backbone which we see as an IP.

- ISP controls the router at our premises.

- Dunno about out, but inside, we can yet decide and get implemented.

Do we have a chance at better DR design??

Joseph W. Doherty Sun, 02/24/2008 - 12:58

I just been working on an answer, when it occurred to me, I'm unsure what specific DR you're trying to design for. DR as in the whole site is physically lost, DR as in network connectivity to a site is lost? Permissible duration of DR event, i.e. acceptable downtime? Level of performance while in DR mode?

Joseph, nice that you asked. Looking forward to your answer.

DR, in the event the whole site's infrastructure (Servers, Applications, Servcies etc., are lost, but routers linking to the MPLS backbobe are still up and running.

As I related earlier, I have three site offices on the MPLS backbone, UK, India and Middle East.

Maintaining core servers data at all three locations and undertaking routine replication so that if a server crash happens at any of the locations, users can be rerouted / load balanced; at any of the remaining two locations?

If cost of such a design is not a criteria, what all do I need to design technically?

Also, you asked, DR in case the network connectivity to a site is lost, then does that involve redundant links too or it can be accomplished with the same MPLS link?

Permissable duration is not yet defined might not be an issue?

Level of performance can be downgraded?

So....Pls advise!!

Joseph W. Doherty Mon, 02/25/2008 - 18:56

What MPLS backbone does for you concerning DR, it often eases any-to-any connectivity. With only three sites, you're not really leveraging this advantage. However, I suspect within the current international market, it might be the most cost effective dedicated WAN technology.

When you describe, ". . . the event the whole site's infrastructure (Servers, Applications, Servcies etc., are lost, but routers linking to the MPLS backbobe are still up and running.", this would seem most unlikely, since if only the MPLS connected routers are working, what are you left with at that site? Perhaps, client hosts, and their supporting infrastructure, are still on-line and still have connectivity to other sites via MPLS backbone. Then it's just a matter to have them use the DR resources at one, or both, of your other sites. (How the latter is accomplished is often not dependent upon the network beyond providing a functional traffic path.)

More likely is failure with your connectivity to/from the MPLS backbone. Also, issues within the MPLS backbone can make for problems between some or all of your sites. Both these points of failures are handled by providing an independent and redundant paths between your sites. My suggestion would be to use VPNs across the Internet. If your sites already have local Internet connections, you could piggy back across them. Just remember to have sufficient additional bandwidth to cover for loss of your MPLS path and insure, if possible, nothing is shared with the MPLS physical path.

[edit]

PS:

The sticky part of having an alternate WAN path is for your network to "know" whether to use it or not. Technically not too difficult, but if a 3rd party is managing your MPLS CE routers, either they should also manage the alternate path or you need some good cooperation.

Actions

This Discussion