11-08-2006 07:50 AM
Hello
I am currently investigating a dual data centre design running
in active/active mode. The data centres will each have connectivity to
our WAN (MPLS) and to the Internet. They will have also have dedicated
links to each other for site replication etc.
Having read a few of the Cisco SRND's what i am still a little unclear
about is whether it is better to connect the two data centres over the
dedicated link using layer 2 or layer 3 and what the pros and cons are of
each. I would appreciate any experiences (good and bad) that people have had
in this area.
My instinct is to go layer 3 eliminating a potential spanning tree issue
that could affect both data centres but i am sure there are more issues
than this to take into account.
Many thanks
11-08-2006 09:51 AM
Hi ,
I think you will be using diff set of ip's for replication purpose . If its going to be like that , then L3 connectivity would be fine because in case of your dedicated link failure you can make ur replication work via the WAN .
I think there would not be any issues with L2 also other than backup, unless there is a ring , there is no spanning tree coming into picture.
Hope this helps
regards
vanesh k
11-14-2006 06:14 AM
i have redundant data centers and they have been setup as follows for specific reasons:
(these data centers are not separated by a WAN, if they were, a T3 or better would be required in my case but i'd opt for a metro fiber type of solution to provide GB+)
using the 3 hierarchial network design: core, distribution, access
1) the CORE is L3/routed; we do not want a L2/switched core for a few reasons. one is to allieviate STP and its inherent problems.
(the core should be moving packets as fast and predictable as possible; stp can interrupt this and cause complete packet forwarding delay or worse; with todays routers, they can route packets just as fast as switching them, or faster in some cases)
2) the distribution layer is switched with fully meshed GB or greater trunks to both the cores. also provides redundant intra VLAN routing for all the VLANs controlled in their specific 'distribution blocks'; i have 5 fully redundant distribution blocks with VLAN routing and VLAN load balancing via HSRP.
(i channel upto 6 GB trunks in a given link)
3) the access layer is switched with fully meshed GB or greater trunks to at least two distribution switches per access switch; one trunk to each core, at least.
(there is no routing performed at the access layer)
other reasons such as the routing operation, location and number of distribution switches, administration and speed affect the design.
12-18-2006 01:37 PM
Go with your instincts and do layer 3. You will appreciate it much more when it comes down to troubleshooting. You are already using WAN w/ MPLS which is already transparent enough to deal with. Imagine an L2 problem.. skip the hassle of having to deal with L2 as well.
Find answers to your questions by entering keywords or phrases in the Search bar above. New here? Use these resources to familiarize yourself with the community: