Olivier,

The use of NNStat for statistics collection on the T1 backbone has not been disabled at any point of entry. Ultimately, drastically differing in and out counts are not unusual..................... it may be, but in the case of CERN they are, as CERN is also a data

Merit/NSFNET.Information.Services-Interested.Parties@um.cc.umich.edu, Elise Gerich <epg@merit.edu>, Jessica Yu <jyy@merit.edu>, nsfnet-info@merit.edu, nsfnet-reports@merit.edu, smh@merit.edu, "Stefan Fassbender (GMD/EASInet)" <stf@easi.net> In-Reply-To: Message of Wed, 1 Apr 92 13:36:53 EST from <jaw@merit.edu> On Wed, 1 Apr 92 13:36:53 EST you said: source.

................................................................ Using the net primarily for ftp will appear as a small amount of traffic into the backbone and a significantly larger traffic volume out. ................................................................ We hope this information helps explain your concerns. Please let us know if we might provide further clarification.

The reason I am very suspicious, is that we are collecting the cisco accounting frequently enough that we do not lose too much of it, and we are therefore in a good position to cross-check our results with those of MERIT. Whereas, our figures agree with the NSFNet ones regarding the traffic out of the T1 backbone, there is a complete mismatch regarding the traffic into the T1 backbone. For example, MERIT reports 358,200 bytes coming from 128.141, the main CERN network, whereas we measure 5,239 Gigabytes (factor 14,000). However, we agree for the traffic to CERN from the US, 5,212 Gigabytes (MERIT), 5,239 Gigabytes (CERN). Back to EASINet as a whole, I do not think it is realistic to believe that the real traffic patterns are as distorted as I reported yesterday, namely:

From EASINet--->NSFNet 78.73 Megabytes (which would be less than the CERN originated traffic, according to us)

**** YES, this is Megabytes, NOT in Gigabytes ************

From NSFNet---->EASINet 82.14 Gigabytes (which is probably all right, but we can only crosscheck part of it because some of the routers at CERN are not controlled by us, what we could measure is: 43.73 GBytes).

To put these figures in perspective, the International Connection Manager (ICM) which is handling the lines from Cornell to Stokholm (Nordunet) and INRIA (France) would have, according to MERIT: 68.29 Gigabytes (out) and 43.42 Gigabytes (in) which is more like the 1/3 2/3 rule of thumb which we observed in 1991, and which shows that even though the USA are exporting more data than they receive, the balance is not that bad, especially considering the small number of FTP servers in Europe, Finland excepted. Olivier