[c-nsp] mlppp performance

[Adam Greene]

Hi,

I'm bonding (4) aDSL lines at a customer location and am only seeing about 66 - 75% of the performance I was expecting. Is this normal? I wonder if an IOS upgrade will help things.

I actually have two customer locations experiencing the same issue. The client routers are 2811's with 512MB RAM running IOS 12.3(8)T6. They are plain vanilla configs, running at ~2% CPU with lots of memory to spare. The head end is a 7205 / NPE200 w/ 128MB RAM and IOS 12.3(15b), terminating about 100 ATM aDSL lines. CPU is at about 14% and memory utilization is low.

The head end reports:

Multilink3, 
  Bundle up for 11:29:07, 1/255 load
  Receive buffer limit 48768 bytes, frag timeout 1000 ms
    0/0 fragments/bytes in reassembly list
    5 lost fragments, 1046793 reordered
    0/0 discarded fragments/bytes, 0 lost received
    0x30FA03 received sequence, 0x4C98A7 sent sequence
  Member links: 4 active, 1 inactive (max not set, min not set)
    Vi7, since 11:29:07
    Vi8, since 11:29:05
    Vi4, since 11:28:59
    Vi9, since 11:27:50
    Vt3 (inactive)

Customer end: 

Multilink1, 
  Endpoint discriminator is xxx
  Bundle up for 11:28:50, 7/255 load
  Receive buffer limit 48768 bytes, frag timeout 1000 ms
    0/0 fragments/bytes in reassembly list
    137 lost fragments, 1453838 reordered
    86/57363 discarded fragments/bytes, 0 lost received
    0x4C7B86 received sequence, 0x30F120 sent sequence
  Member links: 4 active, 1 inactive (max not set, min not set)
    Vi4, since 11:28:48
    PPPoATM link, ATM PVC 0/35 on ATM0/3/0
    Packets in ATM PVC Holdq: 0 , Particles in ATM PVC Tx Ring: 0
    Vi5, since 11:28:42
    PPPoATM link, ATM PVC 0/35 on ATM0/0/0
    Packets in ATM PVC Holdq: 0 , Particles in ATM PVC Tx Ring: 0
    Vi6, since 11:27:33
    PPPoATM link, ATM PVC 0/35 on ATM0/2/0
    Packets in ATM PVC Holdq: 0 , Particles in ATM PVC Tx Ring: 0
    Vi3, since 11:28:50
    PPPoATM link, ATM PVC 0/35 on ATM0/1/0
    Packets in ATM PVC Holdq: 0 , Particles in ATM PVC Tx Ring: 0
    Vt1 (inactive)

Thanks for any insight.
Adam