Comware

 View Only
  • 1.  5300 routing

    Posted Mar 13, 2013 03:42 PM

    Please forgive im new to networking. We recently leased metro Ethernet. We have two locations (Site A, Site B) with 5300 series switches Routing is enabled at site A and site B. we have a server VLAN and PC VLAN at both sites. The uplink ports to the metro E are tagged for both VLANs. We have servers at both locations. Lately we’ve been having latency issues on the server VLAN.  Should I create a new VLAN that has the unlinks port untagged to the metro E network and create static routes to each location

    Site A:

    vlan 12

       name "PC"

       untagged A8-A16,A19-A23

       ip address 192.168.2.254 255.255.255.0

       ip helper-address 192.168.99.5

       tagged B15-B16,C1,C4,D4

       ip igmp

       exit

    vlan 13

       name "Servers"

       untagged B3-B6,B8-B11,B13

       ip address 192.168.99.254 255.255.255.0

       tagged B15,D4

       ip igmp

       exit                        

     

     

     Site B:

    vlan 13

       name "Servers"

       untagged B5,B13-B14

         tagged B16

       ip igmp

       exit

    vlan 12

       name "PC"

       no ip address

       tagged B12,B15-B16

       ip igmp

       exit



  • 2.  RE: 5300 routing

    Posted Mar 17, 2013 11:43 PM

    Generally, if you have a new site, I find it best to design the subnets for that site as independent ones - then it can function independently of Site A if required.

     

    As you say both routers have routing enabled, it seems even more obvious that SIte B's local hosts have their default gateway locally, with routing to get them off-site.

     

    However, often a link between sites used by *servers* requires layer2 connectivity for the server subnet (VMotion, etc...).

     

    In any case, "latency issues on the server VLAN"?

     - Sounds like a Server Team trying to pass off their problems as being a "network problem".

     

    --> Work with them to pinpoint the exact location of the issue: what you will be trying to do is prove exactly which server/switch interfaces are on the path of the "latency issues" and demonstrate that each relevant switchport is always running with lots and lots of spare capacity.