[j-nsp] transfer between 2 ns2000's is slow

[Leslie]

I got some good advice that my limits are suspiciously close to some tcp 
limits -- I have tried some tuning in my kernel but I am going to try 
hacking around a bit more and see if it helps with anything as well.

Leslie wrote:
> I'm having a strange problem that I haven't been able to fix after much 
> studying --
> 
> Basically my setup is host 1 - fw1 - dedicated 1gige link (~25ms lag) - 
> router2 - fw2 - host 2
> 
> I can blast udp across this pipe without a problem, but tcp traffic 
> seems to be limited to about 3 mbyte/s -- I can make multiple sessions 
> that are all transferring at this speed, but no individual session will 
> go over that "limit"
> 
> Another thing that makes me extremely suspicious is occasionally when I 
> start a transfer I'll see a brief cpu spike -- like shown below
> 
>  get perf cpu detail
> Average System Utilization: 21%
> Last 60 seconds:
> 59: 37    58: 34    57: 38    56: 27    55: 39    54: 38
> 53: 81**  52: 76**  51: 81**  50: 77**  49: 82**  48: 62*
> 47: 36    46: 37    45: 35    44: 36    43: 36    42: 35
> 41: 32    40: 37    39: 33    38: 37    37: 34    36: 39
> 35: 33    34: 38    33: 33    32: 39    31: 33    30: 39
> 29: 31    28: 40    27: 29    26: 42    25: 35    24: 41
> 23: 35    22: 38    21: 31    20: 35    19: 32    18: 41
> 17: 35    16: 38    15: 34    14: 39    13: 35    12: 40
> 11: 33    10: 40     9: 32     8: 39     7: 45     6: 39
>  5: 34     4: 40     3: 35     2: 42     1: 36     0: 39
> 
> 
> I've obviously spent hours and hours on the phone/email with tac without 
> much help.  Does anyone have any ideas of what could be doing this?  Any 
> troubleshooting tips?
> 
> Thank you
> 
> Leslie
> _______________________________________________
> juniper-nsp mailing list juniper-nsp at puck.nether.net
> https://puck.nether.net/mailman/listinfo/juniper-nsp