[afnog] Need help on Disaster Recovery Site Setup

[Bernard Wanyama]

Hi Yasini,

>From your mail, I think you are looking at network-level disaster recovery.
You can start by clarifying your requirements and then developing a
design that meets your requirements.
>From there, implementation and then operation follows.

A simple plan & design comes to mind:

Planning:
- Choose a LIVE SITE and a DR SITE
- Size the traffic, interfaces and growth requirements so that you can
deploy the correct equipment model for core routers
- Confirm that ALL ISPs can deliver a high-capacity trunk link to the
LIVE SITE and DR SITE
- WAN service preferred to be Layer 3 VPN (MPLS)

Design:
- LIVE SITE - Core router with 4 trunk links to WAN providers and 2
point-to-point links to DR SITE
- DR SITE - Core router with 4 trunk links to WAN providers and 2
point-to-point links to LIVE SITE
- BGP routing with WAN service providers
- BGP and OSPF (or IS-IS) in the core between LIVE and DR SITEs across
the 2 point-to-point links

Improvements & suggestions:
- Four WAN providers can work very well technically, but from an
operations and management view, 2 are better.
- Fiber access for the core sites will make life easier
- BGP and dynamic routing in general with the WAN providers will also
make life much easier.
- If you are able to get high speed fibers (or even dark fiber)
between the core sites, then you can leverage application-level DR
technologies like Vmware Vmotion, etc

I hope the above helps.

Kind regards,
Bernard



On 4 October 2011 10:18, Mark Tinka <mtinka at globaltransit.net> wrote:
>
> On Tuesday, October 04, 2011 12:39:34 PM Yasini Kilima
> wrote:
>
> > The scenario is not an issue to design but the challenge
> > is how should I configure the failover or load balancing
> > for the sites that are distant and connected by MPLS
> > Layer 3 links? I have 4 WAN service providers and it is
> > required all should be involved as our regional sites
> > are connected on either or two of the four where
> > multihoming is used. I have Cisco routers throughout. I
> > have googled but didn't find any useful information.
>
> Typically, you're going to consider:
>
>        a) One data centre handles all traffic. Other data
>           centre comes online if the main data centre dies.
>
>        b) Both data centres handle all traffic, load
>           sharing as equally as possible.
>
>
> There are two paradigms in which you can solve this problem
> - at the Network level and at the Application level.
>
> At the Network level tends to work if your budget and
> requirements are simple, using basic IP routing to attain
> your goal.
>
> Application level solutions tend to be more granular, simple
> or complex. Solutions range from open source (as some have
> mentioned on this list) to commercial options, not
> discounting what role virtualization can play here. Things
> can be as simple as DNS, or as complex as geographically-
> aware site load balancers.
>
> I'd defer to others on the application side of things, as my
> thoughts would be centred more around solving the problem at
> the routing level (I'm biased, hehe), which may not
> necessarily be what you're looking for :-).
>
> Mark.
>
> _______________________________________________
> afnog mailing list
> http://afnog.org/mailman/listinfo/afnog



--
Bernard Wanyama
Technical Manager
SYNTECH ASSOCIATES Ltd
Cell: +256 712 193979
Fixed: +256 414 251591
Web: www.syntechug.com
Email: bwanyama at syntechug.com