I'm providing oversight over a new network that has, IMO, too much complexity. N+1 or 2N is what I'm typically used to when it comes to network redundancy. This new network has tons of redundancy, and I don't think it buys them any availability. Imagine having two WAN routers with two MPLS providers. That seems pretty standard. However, each provider delivers its circuits on two different PEs. So four PEs in total. Then each of the WAN routers has connectivity to all four PEs. That's a total of eight connections when all that is needed is one. The LAN also has similar levels of redundancy. Two distros with three cores. Full mesh everything. While management and the engineers that designed it believes that it gives them exceptional reliability, I think that the additional redundancy is the culprit to the longevity of network incidents.
When does redundancy become too much? Is there any value to going above 2N?
How do you relay to management that network complexity is the issue when they believe that it's due to the operations staff's ineptitude?
I've been trying to find a design guide, a study, or some calculator that shows the trade-off of redundancy and complexity, but haven't had any success. I figured coming with pretty pictures and numbers would greatly improve my chances of convincing management to reduce redundancy and complexity to just what's absolutely necessary.
No comments:
Post a Comment