Autor: Joerg Hertzer (Hertzer_at_rus.uni-stuttgart.de)
Data: Mon 07 Nov 1994 - 19:48:32 MET
Szanowni Panstwo,
bardzo przepraszam, ale jeszcze nie znam dosic jezyku Polskiego
pisac caly list po Polsku, dlatego pisam po Angielsku.
The message from Wojciech A. Myszka, dated Sat Nov 5 22:31:26 1994
was forwarded to me by Maciek Uhlig.
Maciek Uhlig asked for some comment on the observation that the connection to
Germany is rather slow from Poland.
Since the original message from Wojciech A. Myszka was in Polish I did
not understand all details (yes, I should understand much more using my
dictionary, but please accept that I wanted to use some time otherwise.),
but at least I understand that additional he was wondering why he saw
routing from Poland to Germany via Vienna, but from Germany to Poland
via Stockholm.
First I should mention, that we have three nation-wide IP-networks in Germany:
win-ip, Xlink and Eunet.
Scientific and research institutions are usually connected via win-ip,
sure your connection goes that way if you find a node like
'ipgate2.win-ip.dfn.de' (at least something ending 'win-ip.dfn.de')
in your traceroute result.
EVERYTHING I WRITE HERE IS VALID FOR WIN-IP ONLY!
I am myself working with parts of my time in DFN-NOC, responsible as
'Network Operation Centre' for win-ip.
A few days ago I discussed current situation of the Poland-Germany IP connection
with Peter Merdian, the leader of DFN-NOC. So the following is partly based
on things I learnt in this discussion.
All connections from win-ip to other countries (than Germany) are going via
some nodes in Duesseldorf, to all European countries from there to EuropaNET,
further way to Poland passes via some gateways Ebone and/or NorduNet,
reaches NASK while crossing the border to Poland and is distributed
via Warszawa to the different locations in Poland you know better than I.
A few minutes ago (November 17th, 18:11 h CET) traceroute from a
certain host in Duesseldorf to Poland (esp. Katowice) and back looked
as follows:
ws-dus>date; traceroute -g usctoux1.cto.us.edu.pl -m 50 -w 10 ws-dus.dfn.de
Mon Nov 7 18:11:02 MET 1994
traceroute to ws-dus.dfn.de (193.174.74.41), 50 hops max, 40 byte packets
1 Duesseldorf1.win-ip.dfn.de (193.174.74.45) 25 ms 3 ms 7 ms
2 ipgate2.win-ip.dfn.de (193.174.74.46) 8 ms 3 ms 4 ms
3 duesseldorf2.empb.net (193.172.4.12) 9 ms 11 ms 19 ms
4 stockholm1.empb.net (193.172.4.4) 62 ms 48 ms 67 ms
5 empb-gw.nordu.net (193.172.30.6) 62 ms 64 ms 70 ms
6 193.10.80.43 (193.10.80.43) 96 ms 62 ms 65 ms
7 kirdan.nask.org.pl (193.59.201.33) 471 ms 397 ms 367 ms
8 193.59.201.18 (193.59.201.18) 804 ms 389 ms 405 ms
9 waw-gw.nask.waw.pl (193.59.200.10) 433 ms 394 ms 436 ms
10 193.59.200.26 (193.59.200.26) 409 ms * 394 ms
11 kat-cs.nask.katowice.pl (193.59.200.18) 426 ms 554 ms 830 ms
12 * * *
13 * usctoux1.cto.us.edu.pl (155.158.2.8) 461 ms 455 ms
14 * * *
15 192.102.225.29 (192.102.225.29) 471 ms 462 ms 560 ms
16 kat-gw.nask.katowice.pl (193.59.200.17) 470 ms 474 ms 522 ms
17 waw-gw.nask.waw.pl (193.59.200.25) 535 ms 591 ms 623 ms
18 * 193.59.200.9 (193.59.200.9) 542 ms *
19 193.59.201.17 (193.59.201.17) 484 ms 473 ms 460 ms
20 192.121.159.33 (192.121.159.33) 460 ms 473 ms *
21 * Paris-EBS1.Ebone.NET (192.121.156.17) 628 ms 647 ms
22 Cern-EBS1.Ebone.NET (192.121.157.9) 670 ms 560 ms 546 ms
23 Geneva1.Dante.net (194.41.0.49) 526 ms 555 ms 602 ms
24 amsterdam6.empb.net (193.172.4.8) 634 ms 635 ms 505 ms
25 duesseldorf2.empb.net (193.172.4.12) 457 ms 489 ms *
26 dante.win-ip.dfn.de (193.174.74.66) 467 ms 471 ms 559 ms
27 ipgate2.win-ip.dfn.de (193.174.74.94) 482 ms 572 ms 497 ms
28 ws-dus.dfn.de (193.174.74.41) 531 ms 605 ms *
ws-dus>
We see that round trip time was something like 600 ms and we had not
very many packet losses. So far the international connection
from Germany to Poland and further to Katowice looks good.
The only strange thing is that routes from here and back went different ways.
The latter is a fact I observed not always, but very often in past already.
I don't remember for sure when I saw this first time, but sure already
a year ago when I worked for three months in Katowice. From our point of view
(DFN-NOC) routing should better be done the same way in both direction;
even more because a year ago during a visit in Warszawa I learnt that the line
from there to Stockholm has higher bandwith than that one to Vienna
(But my information on that of course may be obsolete now).
On the other hand this seems to be based on a
decision by NASK and they may have good reasons we don't know here in Germany.
At least we see that this doesn't cause any problems.
Compared to that the connection from Stuttgart to Duesseldorf is much worse:
A traceroute from Stuttgart to Katowice started:
Asn3>date ; traceroute -g usctoux1.cto.us.edu.pl -m 50 -w 10 awssn3
Mon Nov 7 18:43:13 MET 1994
traceroute to awssn3.rus.uni-stuttgart.de (129.69.14.68) 50 hops max, 40 byte packets
1 cisco4.rus.uni-stuttgart.de (129.69.14.200) 0 ms 0 ms 0 ms
2 cisco3.rus.uni-stuttgart.de (129.69.18.110) 10 ms 0 ms 20 ms
3 Stuttgart1.BelWue.DE (129.143.70.1) 10 ms 10 ms 0 ms
4 Stuttgart4.BelWue.DE (129.143.203.2) 10 ms 10 ms 0 ms
5 * * *
6 * * *
7 duesseldorf2.empb.net (193.172.4.12) 1850 ms * 5751 ms
8 amsterdam6.empb.net (193.172.4.8) 3040 ms * 5861 ms
9 * * *
(And than I decided to cancel instead of spending my time
waiting for further results.)
You see that Duesseldorf was reached with round trip times of several
seconds!!
(and a 'ping -s duesseldorf2.empb.net' from Stuttgart finished with:
819 packets transmitted, 147 packets received, 82% packet loss
round-trip (ms) min/avg/max = 240/4742/20822
!!!! )
This can be explained only as a result of very high load on our national
connections.
Please note that win-ip (different to most other IP networks)
is based on X.25 .
I suppose that because of flow control inside of X.25 in case of high load
a very long time passes trying to deliver an IP-Packet till the packet is
dropped at the end.
At least in older measurements done in April 1994 on the Stuttgart-Duesseldorf
connection during several days I observed that with increasing load
up to a certain level only round trip times are increasing
while packet loss rates are staying low, but with even higher load packet
loss rates are increasing too.
During this days several people are trying to understand better
the situation of the X.25 level in win-ip. So we may hope for some results
and improvements.
The load situation from Duesseldorf to different locations in Germany is
different of course, therefore you may see very different quality of
the connection depending on the location you want to reach.
---------------------------------------------------------------------
Related to this I want to tell you that I plan to give (amoung others)
a lecture "IP Networks in Stuttgart and Germany" during the week from
21st to 25th November 1994 at Politechnika Krakowska and to repeat this one at
28th November at Politechnika Wroclawska.
For further details on the lectures in Krakow please contact
Dr Zbigniew Mrozek (pemrozek_at_cyf-kr.edu.pl)
and for details on the lecture in Wroclaw please contact
Prof. dr. hab. Zbigniew Huzar (zhuzar_at_ci-2.ci.pwr.wroc.pl)
---------------------------------------------------------------------
At the end I want to give some courage to all readers in Poland:
You see from above that not all your problems are caused by the special
situation in Poland, but that they may have their reasons elsewhere.
Last summer I came by train from Frankfurt/Main to Gliwice, where some
Polish friends picked me up. The train arrived there with 1 1/2 hours delay.
My friends could hardly believe that the train had the same delay already
while crossing the German/Polish border at Zgorzelec!
I think that situation of Poland and chances for future are better than
(following my impressions from a lot of travels) most people in Poland believe.
Please note that I am not on POLIP-List,
therefore please send any questions/comments personally to me.
Dr.-Ing. Joerg Hertzer Phone: ++49-711-685-5734
Computer Center University Stuttgart Fax: ++49-711-682357
Allmandring 30 E-Mail: Hertzer_at_rus.uni-stuttgart.de
70550 Stuttgart, Germany
To archiwum zostało wygenerowane przez hypermail 2.1.7 : Wed 19 May 2004 - 15:46:55 MET DST