Hello,
we have sometimes heavy performance problems and we are now at the
end of our knowledge 
The facts: we have a 3 Mbit Part of the OC3 going from
Frankfurt (ECRC/Ebone) to Sprint.
We never had longer peaks, there is enough space in the line.
Sometimes more than one MB/sec. Ping times are quite good, so this
cant be the reason for our problems.
Some ftp Downloads from US-Sites (cisco.com, sun.com etc...) are as
fast as a ISDN line. (6-7 kb/sec :-)) othes have 60-80kb/sec...
Several Traceroutes now have shown us, that almost all routes in the
usa are asymetric:
Just one example:
Von nitrous.digex.net:
Type escape sequence to abort.
Tracing the route to anify.farside.net (195.110.11.2)
  1 icm-mae-e-f0/0.icp.net (192.41.177.240) 4 msec 4 msec 4 msec
  2 icm-bb1-dc-1-0-T3.icp.net (198.67.131.10) [AS 1800] 4 msec 4 msec
4
msec
  3 icm-pen-1-1-0-T3.icp.net (198.67.131.18) [AS 1800] 8 msec 4 msec
4
msec
  4 icm-pen-14-P0/0-OC3C.icp.net (198.67.142.70) [AS 1800] 8 msec 4
msec
4 msec
  5 icm-pen-13-P1/0-OC3C.icp.net (198.67.142.97) [AS 1800] 8 msec 8
msec
4 msec
  6 198.67.142.86 [AS 1800] 88 msec 88 msec
    165.117.55.2 64 msec
  7 NCP.Frankfurt2.ECRC.NET (192.121.158.110) [AS 1755] 92 msec
    NCP.Frankfurt2.ECRC.NET (192.121.158.110) [AS 1755] 88 msec
    NCP.Frankfurt2.ECRC.NET (192.121.158.110) [AS 1755] 92 msec
  8 165.117.50.26 76 msec
    seicom-gw-FFM2.ecrc.net (195.27.83.90) [AS 1273] 112 msec
    seicom-gw-FFM2.ecrc.net (195.27.83.90) [AS 1273] 136 msec
  9 US-gw.Frankfurt.seicom.NET (194.97.193.18) [AS 5549] 128 msec
    US-gw.Frankfurt.seicom.NET (194.97.193.18) [AS 5549] 168 msec
    US-gw.Frankfurt.seicom.NET (194.97.193.18) [AS 5549] 128 msec
10 Frankfurt1-s0-0.bw-bone.net (195.254.83.33) [AS 5549] 352 msec
140
msec 156 msec
11 Heidelberg1-s3-7.bw-bone.net (195.110.0.85) [AS 8254] 156 msec
156
msec 168 msec
12 portal.farside.net (195.110.7.34) [AS 8254] 196 msec 288 msec 336
msec
13 anify.farside.net (195.110.11.2) [AS 8254] 128 msec 148 msec 160
msec
and the way back:
traceroute to nitrous.digex.net (205.197.249.3), 30 hops max, 40 byte
packets
1 195.110.11.1 (195.110.11.1) 3.218 ms 2.171 ms 2.065 ms
2 core1 (195.110.7.33) 7.054 ms 4.865 ms 5.615 ms
3 Frankfurt1-s0-0.bw-bone.net (195.110.0.86) 8.134 ms 7.691 ms
8.097 ms
4 195.254.83.34 (195.254.83.34) 8.886 ms 7.593 ms 7.070 ms
5 Ebone-gw.I-bone.NET (194.97.193.17) 14.484 ms 16.114 ms 16.668
ms
6 s1-0-r1-FFM2.ecrc.net (195.27.83.89) 38.792 ms 34.726 ms
65.364
ms
7 frankfurt-ebs2-h2-0-1.ebone.net (192.121.158.109) 59.266 ms
47.429
ms 39.759 ms
8 198.67.142.85 (198.67.142.85) 116.419 ms 107.364 ms 140.257 ms
9 sprint-nap.digex.net (192.157.69.42) 127.001 ms 133.519 ms
122.598 ms
10 phl2-core2-f4-0.atlas.digex.net (165.117.51.37) 276.222 ms
144.815
ms 134.420 ms
11 dca1-cpe2-h10-1-0.atlas.digex.net (165.117.51.33) 124.355 ms
127.911 ms 139.501 ms
12 dca1-cpe8-fe4-0.atlas.digex.net (165.117.224.2) 148.073 ms
144.636
ms 131.157 ms
13 nitrous.digex.net (205.197.249.3) 259.808 ms 170.213 ms
196.965
ms
Currently we dont have a second path to the us, so it must have other
reasons.
The route object in the RIPE Datebase contains the correct AS and
Maintainer.
We dont have a
advisory: AS690 1:1800 2:1239(147) 3:1239(218) 4:1239(144)
which some ISPs have even it. Is this still necessary ? Does
it make sense ?
The Source IPs belong either to several /19 Blocks in 195.X.X.X or
one /17 in 195.X.X.X or one /18 in 194.X.X.X.According to the filter-
policies,
theses ip-blocks are good enough :-=)
Are there other sources for gernerating filter-lists on the core
router in the us than the
ripe datebase ?
To avoid wrong results due to the server load: can anybody provide my
a ftp-sute, which is connected very good (OC3 or more) and has no
local load problems caused by too much user on the system...
Winfried