Routing

Expand all | Collapse all

Frame loss test 10G when using using frame size below 512 byte through L2VPN on ACX5448..

  • 1.  Frame loss test 10G when using using frame size below 512 byte through L2VPN on ACX5448..

    Posted 13 days ago
    Hi all,


    Currently i'm testing 10g traffic transit through L2VPN using ACX5448 and PE. All the jumbo MTU already configured properly and no issue in reachibility. The issue is when i generate traffiic 10g from Tester 1 to Tester 2 using frame size below 512 byte then we will detect frame loss on tester itself that look like frame loss start when traffic on 7g. On ACX itself we not see any traffic drop when execute command "show interface extensive". Is there any special knob to make sure it can success 10g traffic through MPLS? Below is my test topology.


    Tester 1 (generate traffic 10g) -----> (L2VPN) PE1 (MPLS) ---------- (MPLS) PE2 (L2VPN) --------> Tester 2 (receiver should recevied 10g).


    Thanks and appreciate your feedback.


  • 2.  RE: Frame loss test 10G when using using frame size below 512 byte through L2VPN on ACX5448..

    Posted 10 days ago
    Hi, are you using LDP/RSVP or Labeled-ISIS  to exchange label information across MPLS network? Any bandwidth reservations or traffic congestion  on the MPLS path?

    What type of L2VPN is being tested? Could you please provide configuration snip of that part from both sides and maybe overall topology to better understanding how devices are connected.

    Please provide the following output from both sides: show mpls lsp extensive

    Have you also tried to swap SFP modules and re-check after that? Any possible issues checked on physical part?



    ------------------------------
    Regards,
    Elchin
    ------------------------------



  • 3.  RE: Frame loss test 10G when using using frame size below 512 byte through L2VPN on ACX5448..

    Posted 7 days ago
    Hi E.K.H,


    IGP i'm used ISIS. For MPLS i'm using LDP as signalling. It just basic setup just to simulate from traditional Metro-E using MPLS-TP to use MPLS-IP. All SFP already swap but the result is same.


    Thanks


  • 4.  RE: Frame loss test 10G when using using frame size below 512 byte through L2VPN on ACX5448..

     
    Posted 7 days ago
    Hello kronicklez,

    in general it is not a secret that the less the packet size is, the less bandwidth throughput is possible.
    Although I do not have experience with ACX5448, my personal guess is that the Broadcom ASIC inside is handling packets <= 512 Bytes differently than bigger packets.

    If you can't find someone other within this community who can help you with your questions, the only remaining way is asking JTAC about this.

    ------------------------------
    ------------------------------
    If my answer provides the solution, please mark my post as "Accepted Solution".
    If you think my answer helps, please spend some Kudos
    ------------------------------



  • 5.  RE: Frame loss test 10G when using using frame size below 512 byte through L2VPN on ACX5448..

    Posted 6 days ago
    This is the case for most platforms, they cannot provide line rate forwarding with frame sizes such small. To be honest, I have not seen a lot of platforms out there passing RFC 2544 (Throughput benchmarking RFC) for smaller frame sizes.

    This is expected, you might want to contact Juniper account and ask what was the minimum benchmark tested with this platform by the corresponding business unit, they should share this info accordingly.

    Regards,

    Elvin


  • 6.  RE: Frame loss test 10G when using using frame size below 512 byte through L2VPN on ACX5448..

    Posted 6 days ago
    Hi Elvin,


    The issue happen only when using MPLS. If i'm test direct without using MPLS then no issue on small frame.


    Thanks


  • 7.  RE: Frame loss test 10G when using using frame size below 512 byte through L2VPN on ACX5448..

    Posted 6 days ago
    Can you check and share pps rate on ingress and egress interfaces as well as both interface extensive statistics and configuration during the test? (VLAN tag, manipulation or anything?)
    We can also try to calculate wire pps as explained here

    I assume that the issue might be related to the buffer size per port that is used by device also. As from what I've found for now is: By default a port on an ACX5448 router gets a dedicated buffer of 100 microseconds and shared buffer from DRAM. Delay buffer controls the latency of the queue during congestion and maximum number of packets that can be held in a queue. Default buffer size per port is 100 microseconds.



    ------------------------------
    Regards,
    Elchin
    ------------------------------