Name: Bonn-EBSx.Ebone.net
  (although the EBS will be located some miles away from Bonn, we thought
  that StAugustin-EBS1.Ebone.net wouldn't give too much information to
  someone who is tracerouting across Ebone.)

Connectivity:
  - national: to be discussed with german IP providers and ECRC during RIPE.
    (the initial number of RBSses should be defined in Paris)
  - international: planned 256+ to Stockholm, 256+ to CERN, ? to US (if yes,
    it needs to be prepared for any backup announcements.)

Access:
  - TACACS with Ebone-passwd be installed as soon as possible.
  - Out of band access via X.25 will be provided.

Implementation:
  Since we have to order at least one new box, we are free to implement in
  various ways:
    We could implement the "generic EBS model" (ie only one box for all
  international lines and dedicated boxes for each RBS), but this shows
  some problems and we would like to propose a new "generic EBS model" (ie
  to have one box for each international line):

  
  Stockholm
    |                     ---+---
    |                        |    +------+
    |      +-----------+     |----| RBS1 |
    +------| Bonn-EBS1 |-----|    +------+
           +-----------+     |
                             |     . . .
           +-----------+     |
    +------| Bonn-EBS2 |-----|    +------+
  IDNX     +-----------+     |----| RBSn |
    |                        |    +------+
    |                     ---+---
  IDNX                      GMD
    |                   Interconnect
  CERN


  - Bonn-EBS1 and Bonn-EBS2 could be new Cisco 3000s (Model 3102, available
    from November, one serial, one Ethernet) with Flash EPROM, 16 MB and 68030.
    (So, memory limitations or processor performance problems are not expected
    and cost/performance relation is much better than running a AGS4 using only
    two serial and one Ethernet ports)

  - Bonn-EBS1 and Bonn-EBS2 would both participate in internal IGRP and I-BGP.

  - dual E-BGP peering to all RBSses.


  Pros:

    + Improved availability of international connectivity by redundancy.
      (two machines have to fail to disconnect from Ebone.)

    + Maintenance without connectivity interrupts
      (by always keeping one machine alive during upgrades or reconfigs)

  Cons:

    - Little more transit traffic delays.
      (but not as much as in and around Stockholm ;-)

    - One more internal BGP and IGRP speaker.
      (but not as many as in and around Stockholm ;-)

  We think that the redundancy of international connectivity is the great
  advantage, especiallly where we (like others) don't run a 24x7 hour NOC
  to solve problems during nights and weekends, when users typically want
  to complain.

    In worst case a single box failure out of operating hours would just
  disconnect a single RBS from the network. All other RBS-EBS and RBS-RBS
  traffic flows would not be affected.

  Any opinions? Any bugs/problems/experiences/show stoppers with the new 3000s?
  Is the tli-code expected to run immediate with the 3000s? Any information from
  inside Cisco? Could this be interesting for other EBSses?