Wired Intelligent Edge

 View Only
Expand all | Collapse all

Replace Comware 5700 with AOS-CX 8325 core/server switches

This thread has been viewed 10 times
  • 1.  Replace Comware 5700 with AOS-CX 8325 core/server switches

    Posted Oct 23, 2023 11:19 AM

    Hi,

    We're currently having the following switches in our network, which will need to be replaced by newer switches.

    Core/server switches:
    4x HPE 5700 32XGT 8XG 2QSFP+ switches in a 4-node IRF stack, physically deviced over 2 datarooms (2 units per dataroom). The switches are inter-connected with QSFP+ (DAC cables locally in the dataroom, SR fiber transceivers over the 2 datarooms).

    No routing or VRRP in use. All L3 things happen on the firewall.

    All access switches are connected to the core/server switches.
    Also all servers are connected to the 5700 IRF stack with LACP link-aggregations.

    Access switches:

    7x 2930M switches, single units (no stacking), connected to both datarooms with SR fiber (SFP+ 10G) with LACP to the 4x 5700 IRF stack.



    In the new situation we're planning to replace the 4x 5700 switches with 4x 8325 switches to be future-ready for higher speed connections to servers with 25G (instead of 10G nowadays). The 2930M access switches will not be replaced at this time.

    I was wondering how to get the 4x 8325 switches configured correctly, because we will have to use VSX instead of IRF stacking now.
    I read it's only possible to have 2 units in a VSX 'stack'. 
    Will I have to form a VSX stack per dataroom (VSX 1: unit 1 + 2, VSX 2: unit 3 + 4), or a VSX stack over the two datarooms (VSX 1: unit 1 + 3, VSX 2: unit 2 + 4)?
    I still want the access switches connected to both datarooms in a 2x 10G LACP with both links active (so no active/standby with STP).

    Thanks in advance for your help!



  • 2.  RE: Replace Comware 5700 with AOS-CX 8325 core/server switches

    Posted Oct 23, 2023 11:29 AM

    Hi

    Yes, it's correct that VSX can only form a cluster with exactly 2 members. So you will end up in the decision whether you want to form two VSX clusters with the two switches in the same dataroom or two clusters with one switch per room each. Both is suboptimal for either the servers which I assume should be connected "locally" to the two core switches or the access switches which should be connected to a switch in each dataroom. 

    In my projects I always ended up in having a third pair of VSX switches so I could cover all the use cases and avoid any sort of suboptimal redundancy or loops/STP. I had a pair in dataroom 1, another in dataroom 2 and a third across both rooms.

    The other solution would be to only deploy 1 pair in total, 1 node per room, save on some cost (if possible given by the port capacity) and connect the server using 1 DAC and 1 SR connection to the other room. 

    I hope that helps. 

    Kind regards, 
    Thomas