🚀 go-pugleaf

dave@ecrc.de (Dave Morton) writes:

>In article <Bt88s8.Cov@immd4.informatik.uni-erlangen.de>, husemann@camelot.informatik.uni-erlangen.de (Dirk Husemann) writes:
>> schoett@informatik.tu-muenchen.de (Oliver Schoett) writes:
>> 
>> >In article <1992Aug18.082075.pell@lysator.liu.se> pell@lysator.liu.se writes:
>> 
>> > > I wonder if it might be time to question the future of the Internet
>> > > (TCP/IP) in Germany. It's not my purpose to tell anyone how to do
>> > > things, but my fear is that not everybody is aware of what's going on.
>> > > Maybe it's not too late for those who want IP to do something.
>> 
>The time has long passed. We (ECRC) did something about a year ago, we have
>our own line to Dortmund and we'll shortly have a direct one to Amsterdam
>as well. 

Fine. Just out of curiosity --- How much do the lines to Dortmund/Amsterdam
cost you and what capacities are available on that lines? Also, does the ECRC 
have the need to connect to German sites at all? I.e. do you have a larger
amount of international (EC & US & et al.) traffic than German?

>Basically we got sick and tired of X.25 and the lack of European
>connectivity from Germany. Except for EUnet I do not know of other service
>providers here providing good connectivity to the rest of Europe. 

Yeah, sure, switching your own line to EUnet will probably give some kind of
reasonable connectivity --- but that's a little of the mark, as that requires
*money* for own lines. Service --- as pointed out by other posters --- via the
EUnet to European sites --- frankly --- sucks. Yes, I'm aware that this is
due to an overloaded WIN point-of-attachment, still it sucks from *my* point
of view as a customer of EUnet (we are doing --- for historical reasons our
European traffic with EUnet and the rest with DFN-IP).

>Having said that - yes - many, many people in Germany are aware of the
>problem and are attempting to do something about it, let's wish them success.

I heard rumors that things will probably change to the better, let's hope
things will proceed as necessary.

>> >I noticed an even crazier routing last night: from Munich to France
>> >over London - Stockholm - Amsterdam!
>> 
>> >   [lida] traceroute -m 20 -w 40 cnam.cnam.fr
>> >   traceroute to cnam.cnam.fr (192.33.159.6), 20 hops max, 40 byte packets
>> >    1  * * *
>> >    2  bro4cz.lrz-muenchen.de (129.187.220.254)  9 ms  8 ms  8 ms
>> >    3  bro1c1.lrz-muenchen.de (129.187.10.251)  16 ms  10 ms  10 ms
>> >    4  dfnbigate_s2.GMD.DE (188.1.236.124)  12255 ms  7205 ms  11573 ms
>> >    5  int-gw.ulcc.ac.uk (192.16.192.185)  13256 ms  15098 ms  8323 ms
>> >    6  London-EBS1.Ebone.NET (128.86.1.6)  14269 ms *  7120 ms
>> >    7  Stockholm-London.Ebone.NET (192.121.155.18)  12593 ms  8802 ms  11991 ms
>> >    8  Stockholm-EBS1.Ebone.NET (192.121.154.33)  5711 ms  4533 ms  5486 ms
>> >    9  Stockholm-EBS2.Ebone.NET (192.121.154.18)  10077 ms  13970 ms *
>> >   10  * Stockholm-EBS4.Ebone.NET (192.121.154.50)  12680 ms  13180 ms
>> >   11  Amsterdam.NL.EU.net (192.87.4.20)  9131 ms  6456 ms  7259 ms
>> >   12  Paris.FR.EU.NET (134.222.2.2)  6293 ms *  18449 ms
>> >   13  fnet-gw.inria.fr (192.93.1.108)  11685 ms *  22512 ms
>> >   14  cisco.cnam.fr (192.70.71.6)  16838 ms  14640 ms  21646 ms
>> >   15  * * cnam.cnam.fr (192.33.159.6)  11884 ms
>> 
>This is perfectly normal, as seen from a DFN/WiN customers site, 
>and it is part of the current EBONE routing setup. Note IXI is being used
>between Germany and the UK *before* traffic enters EBONE.

>If I, from Munich do the same trace I get:

>ecrc# traceroute -m 20 -w 40 cnam.cnam.fr   
>traceroute to cnam.cnam.fr (192.33.159.6), 20 hops max, 40 byte packets
> 1  Munich.DE.EU.net (141.1.1.2)  6 ms  5 ms  5 ms
> 2  Dortmund.DE.EU.net (139.4.4.1)  42 ms  38 ms  47 ms
> 3  Amsterdam.NL.EU.NET (134.222.1.1)  1046 ms  671 ms  751 ms
> 4  Paris.FR.EU.NET (134.222.2.2)  1443 ms  798 ms  1079 ms
> 5  fnet-gw.inria.fr (192.93.1.108)  1229 ms  2729 ms *
> 6  cisco.cnam.fr (192.70.71.6)  2349 ms  1762 ms  1240 ms
> 7  cnam.cnam.fr (192.33.159.6)  1600 ms *  1227 ms

>and this I did a few minutes ago, 15:10, not exactly the best
>time of the day...... Even taking into account an overloaded 
>Dortmund - Amsterdam line, this is pretty good compared to the
>above........

Come on, this is not exactly fair comparision, right? You read my posting ---
at least you included my followup to the above --- and should know that
the GMD (dfnbigate_s2) does/did suffer from severe CRC error on their X.25
p-o-a. Comparing a line that's used exclusively by the ECRC (and presumably in
good shape) with a faulty one (knowing that it is faulty) is kind of 
preposterous, IMHO that is.

> 
>> I cannot speak for neither the DFN nor the GMD doing the actual service of 
>> providing IP/X.25. I think a couple of circumstances happened to coincide here:
>Neither can I nor do I wish to do so.

>> 
>> 	(1)	The GMD (dfnbigate_s2) has been experiencing severe line
>> 		problems with their X.25 point-of-attachment. The German
>> 		Telekom seems to be busy trying to fix that situation. 
>> 		(Basically they suffer from a huge amount of CRC errors on
>> 		that line).
>> 
>> 	(2)	The DFN-IP (what all the people complaining so far have been
>> 		using) is using a new line between Germany and the UK. This
>> 		line will be run at 2 Mbit/s in the next future. The UK seems
>> 		to have a >= 512 Kbit/s line to Stockholm. Thus the net 
>> 		effect will be better connectivity.
>The "new line" is IXI - nothing more, nothing less and IXI as a pilot
>has been around for a while, nothing new here. Also IXI was really
>overloaded at the access point in Amsterdam the last time I looked.
>The JANET line to Stockholm is only 256, not 512, and it's part of the
>EBONE core. The so-called 2 Mbit/s line from Germany to London that you
>mention is IXI production service, i.e. again X.25 and I have not seen any firm
>dates for it's going into operation. Besides, even if it did go into
>operation tomorrow morning, the COSINE figures I have seen indicate
>16% X.25 throughtput in the first instance for 128 octet packets, not exactly
>overwhelming.

Resulting in a capacity of ~ 330 KBit/s (maximum, admittedly) --- not too bad.
What kind of a line does the EUnet entertain? 128 Kbit/s?

>From my WiN connection I basically gave up even trying to connect to
>the rest of the world. Pity I missed the date to cancel the connection,
>for us it has proven almost useless, still there's always next year.....

``Useless'' because you have no communication needs within Germany (i.e. no
IP sites to do business with)? 8=) In that case ... I mean it's easy to
bash the WIN if you're not in need of connecting to any German site (IP,
X.25 or otherwise) ...

>As I see it the WiN is the problem, that why you're seeing the lousy times
>above. Again, watch these figures over our line from Munich to Newcastle Uni.
>Note JANET is also and X.25 network......if you see what I'm getting at.....

I can imagine where you would like to get to, yet, the JANET does have a
different structure, different equipment (which has been shown to run X.25 at
speeds of 2 Mbit/s and more even under load *and* in real life situtations ---
the computing center people here did it, too) and a different environment:
British Telecom does offer lines at far lowere prices than German Telekom does.
The people of the WIN department at the German Telekom need to rent the lines
themselves from the department of the German Telekom that's doing in lines.
Apparently even those prices inside German Telekom are pretty stiff (don't
ask me why, I really don't know). JANET can --- also --- rent lines that run
thru property of privately owned utilities companies (British Rail [yeah, I
know, not currently privately owned, but it seems to have an equivalent status],
electricity companies, water works, etc) which it seems difficult to do in
Germany.

Sigh, before the CRC problems at the GMD we had pretty good RTTs and 
throughput (matching if not exceeding your examples) ...

>ecrc# traceroute cragsted.ncl.ac.uk
>traceroute to cragsted.ncl.ac.uk (128.240.2.80), 30 hops max, 40 byte packets
> 1  Munich.DE.EU.net (141.1.1.2)  6 ms  5 ms  5 ms
> 2  Dortmund.DE.EU.net (139.4.4.1)  37 ms  37 ms  37 ms
> 3  Amsterdam.NL.EU.NET (134.222.1.1)  930 ms  1268 ms  369 ms
> 4  Stockholm-EBS4.Ebone.NET (192.87.4.22)  427 ms  359 ms  108 ms
> 5  Stockholm-EBS2.Ebone.NET (192.121.154.49)  359 ms  440 ms  208 ms
> 6  Stockholm-EBS1.Ebone.NET (192.121.154.17)  259 ms  675 ms  494 ms
> 7  * Stockholm-EBS5.Ebone.NET (192.121.154.34)  458 ms  302 ms
> 8  London-Stockholm.Ebone.NET (192.121.155.17)  337 ms  536 ms  435 ms
> 9  jips-gw.ulcc.ac.uk (128.86.1.4)  459 ms  688 ms  575 ms
>10  gw.ed.JA.NET (146.97.112.7)  755 ms  856 ms  649 ms
>11  gw.ncl.JA.NET (146.97.224.10)  985 ms  596 ms  821 ms
>12  128.240.2.80 (128.240.2.80)  782 ms  493 ms  740 ms
>ecrc# 
>This is at 15:20. I can live with this.....
> 
>> Bashing the DFN-IP service w/o trying to find out why things go wrong/don't
>> work that well is IMHO not exactly what one would consider fair, right? A call
>> to your network administrator could have reveiled that (the GMD people did
>> post a mail to the sites that use DFN-IP telling about the CRC problems).
>> 
>> 	Dirk
>Your right Dirk, and I'm quite sure GMD cannot be in any way held responsible
>for the existing problems. I can very well imagine that they are probably 
>as disgusted as everyone else in Germany, but what can they do......

>The only way to get anything done is beat up DFN at the Mitgliederversammlung.
>The only way this will happen is if everyone actually running the service
>at the sites beat up their DFN representatives, and I'd guess knowing the
>way things work in Germany this will not happen in my lifetime. What a pity
>and a waste of energy, talent and money when things could be so different.

Just beating them up won't help. Quite the contrary, I fear. I think there
have always been two alternatives:

	(a) Try to improve the current situation, accepting the frameworks, the
 	    current set of rules --- a couple of people are doing that right
	    now. One possibility might --- just hypothetically spoken --- be
	    to get rid of equipment and a layout that has been proven to be
	    corrupt/unsuited to the task --- without condeming the technology
	    used. 

	    I think we should bear in mind that the real world is not always
	    as objectively thinking and decision-making as we assume/wish/
	    think it is. There are people out there making decision about
	    networking in Germany (and to a certain extent --- in Europe) 
	    which we all know are in favor of a certain technology. Now if 
	    you start bashing them (which quite a lot of people have done
	    for quite some time) and tell them that it is all ``bullshit''
	    what they are doing/thinking you're not going to raise a lot
	    of sympathy for your own concerns and ideas, right? If one keeps
	    insisting on the ``bullshit'' attitude that's sure to create
	    a we-and-they situation, each party regarding the other as not
	    quite sane in their heads and in dire need of a sanity check.

	(b) Try to change the frameworks, the current set of rules. Also,
	    a couple of people are trying to do that. This of course involves
	    a lot of talking, a *lot* of patience, the ability to suffer
	    set backs (and possibly quite a lot of them) --- and to act
	    diplomatically, as you want to convince people that what they have
	    been thinking/doing is not quite what you think it is. 

	    If at the same time part of your party is becoming inpatient and
	    tries to push forward too fast you end up with the aforementioned
	    we-and-they attitude (see above).

I'm seeing progress with the (a) method, while (b) already has resulted in
the mentioned we-and-they attitude (making life more difficult for the 
(a)-people). 

I can see progress with respect to other projects where (b) is better suited
than (a) --- take 386BSD for example.

Oh well, this sure is an area for ``religious wars'', isn't it?

I hope I've not offended someone out there,
	Regards,
		Dirk

Dirk Husemann						 Tel.: +49 9131 85 8029
University of Erlangen-Nuremberg			 FAX: +49 9131 39388
Department of Computer Science IV
Martenstrasse 1 	   
D-W 8520 Erlangen, Germany

Dirk.Husemann@informatik.uni-erlangen.de
/C=de/A=dbp/P=uni-erlangen/OU=informatik/S=husemann/G=dirk/

---