North American Network Operators Group
Date Prev | Date Next | Date Index | Thread Index | Author Index | Historical

Re: Anycast and windows servers

  • From: Steve Francis
  • Date: Fri Feb 20 13:00:30 2004

Given your initial question was, I think, about the OSPF implementation on windows - I used it on NT 4.0, when it was part of the routing and remote access option, to implement fault tolerant routing through some windows based firewalls.
It worked fine then. So long as you minimize the services running on the windows box, it was stable enough.
Have not used window servers since NT 4.0, but I don't imagine its gotten worse.


Buhrmaster, Gary wrote:

Depending on the service being provided, Microsoft
has their own clustering solution which will
perform failover. Sometimes choosing full vendor
supported technologies is the easiest path.
With Windows 2003 Server they even support
geographically disperses failover. Info at:
http://www.microsoft.com/windows2000/technologies/clustering/default.asp

Gary


-----Original Message-----
From: Daniel Senie [mailto:[email protected]]
Sent: Friday, February 20, 2004 6:39 AM
To: Sean Donelan
Cc: [email protected]
Subject: Re: Anycast and windows servers



At 05:43 AM 2/20/2004, you wrote:


On Thu, 19 Feb 2004, Patrick W.Gilmore wrote:

Honestly, I do not know about OSPF (or BGP) on Windows,
however, you

can just static route to the Windows box(es). Sure, if
the OS hangs,

the interface will stay up and the static route will
still push bits at

the dead box, but it will work (FSVO "work").

Besides, how often does Windows crash? <snicker>

Hence the reason why I want the route to cease being
advertised if the box

"fails."

Connect the server(s) to APC MasterSwitch or equivalent hardware. Monitor the server box(es) for responsiveness. If/when it fails, the monitoring station can instruct the MasterSwitch to reboot (power cycle, really) the box. Stuff is pretty inexpensive (certainly less so than load balancers).



I'm trying to avoid putting yet another server load balancer
box in front

of the windows box to withdraw the route so a different
"working" box will

be closest. It may be an oxymoron, but I'm trying to make
the windows

service (if not a particular windows box) as "reliable" as possible
without introducing more boxes than necessary.

My initial thought last night was in fact the use of load balancers. But then you need to think about redundant load balancers and so on.