Data Center

IMPORTANT MODERATION NOTICE

This community is currently under full moderation, meaning  all posts will be reviewed before appearing in the community. Please expect a brief delay—there is no need to post multiple times. If your post is rejected, you'll receive an email outlining the reason(s). We've implemented full moderation to control spam. Thank you for your patience and participation.



  • 1.  Best option for physical connectivity

     
    Posted 05-05-2021 07:20
    Good Day,

    I have attached a diagram with 3 examples and would like to know which would be considered the best for physical connectivity between the fabric and core.  The network consist of an MX MPLS core and then two data center fabrics. The fabric would be a combination of bridge overlay and routed bridge overlay at the BL. The fabric does not extend to the MX.  The links are 10G and we are using AE not only for redundancy but the ability to increase the bandwidth. I feel example 3 is a no go do to possible complexity with using MCLAG etc. but from the other two examples would be the best for redundancy, traffic flow and removes complexity from the overall solution without compromising redundancy etc.



  • 2.  RE: Best option for physical connectivity

    Posted 05-06-2021 08:33
    Hi ​MFB,
    Might I offer the suggestion that, since you're going to the effort in example 3 to use ESI LAG on the fabric​ side, you could also go ahead and use ESI LAG on the MPLS network? This could get you that active/active multihoming across the MPLS network, as it sounds like bandwidth is a concern. Depends on perspective, but you certainly could argue that this also reduces the complexity of example 3.

    ------------------------------
    Jason Rokeach
    ------------------------------



  • 3.  RE: Best option for physical connectivity

     
    Posted 05-06-2021 08:57
    Good Day,

    Thanks for the feedback.  We will not be using EVPN signaling in the core MPLS network.  Most services coming from the fabric will be  offloaded into a L3VPN at the core.

    Thanks,


  • 4.  RE: Best option for physical connectivity

    Posted 05-06-2021 10:02
    That helps. I had assumed that you had some L2 you needed to cross the MPLS network based on the MC-LAG and ESI-LAG design options. If you're strictly operating L3 at the core, then I'm with you completely in terms of complexity and would likely avoid any MC-LAG (or even ESI-LAG - the BLs could serve separate LAGs if you want since it's L3 termination). You likely already know this, but ensure you use IP-based RDs on the L3VPN for this use case to allow for multipath.
    You've outlined what appear to be 3 valid designs, and what likely matters most at this point is the ease of you and others who would be responsible for this network to implement and manage it.

    ------------------------------
    Jason Rokeach
    ------------------------------



  • 5.  RE: Best option for physical connectivity

     
    Posted 05-06-2021 11:09
    Thanks for the feedback Jason,

    So both option 1 and 2 all traffic get concentrated at PE1 MX due to L3 GW being primary for a particular service. The same could be reversed for PE2 for another service. The benefit with option 1 is just normal AE between the BL and PE direct. If i need to add capacity ,i just add more members.  Also there is more than enough capacity inside the fabric so flows do not need to be spread across BL before going to the primary PE (per option 2) so i am happy that everything goes through the same BL on exist and it kind of makes traffic flow predication much easier and symmetric .  The only problem with option 1 is the unlikely failure of cross opposite devices. For L2 stretch between deterrent DC leafs i will just be dong DCI via a L3VPN as well and the VTEP will be between all DC leafs.