Hello, I'd like your opinion on the following concept. It is non-specific to any single MAE or continent but my original idea is of MAE-EAST and Europe. An overseas operator could get a circuit from one of its POPs directly to the MAE ATM switch (matrix) and set up peering PVCs between its peers and the router(s) in its POP overseas. This would eliminate the need for colo space and remote hands service for an overseas router from the operator's point of view. I can immediately think of four drawbacks to this: 1) Routing Protocol Latency This shouldn't be an issue with BGP. 2) Routing Protocol Bandwidth Waste This would only be an issue if the operator would receive lots of same routes from two or more peers. To be a constant issue those routes would need to flap a lot too. 3) Router Not On Shared Medium (Ethernet, FDDI) This could be overcome by some cooperation with a local provider that has a switch between its shared medium connection and its router. The switch should be capable of ATM LANE though and MAE ATM switches would need more complex configuration as well. Ideally the MAE ATM switches would do it directly, but I'm not sure thay are capable of it? 4) Bandwidth Waste From ATM Encapsulation A POS circuit would use the long distance bandwidth more efficiently. -- Aleksi Suhonen Network Development Saunalahden Serveri Oy AXU-RIPE Please keep me in the Cc:-field as I'm not subscribed to the list.
Been there, Done that. The IEPG web page (currently not visable) has the earliest version I know of, the dGIX (circa 1991-1992). This idea was implemented in the NSF ATM NAPs (AADS and PB) in 1994-1995. These continue to operate. The largest problem is where there is apparent layer 3 reachability but the NLRI (Network Layer Reachability) is fragmented. The easiest way to "fix" this is to ensure that there is a full mesh in the ATM cloud. Doing so is not always trivial. --bill
participants (2)
-
Aleksi Suhonen
-
bmanning@vacation.karoshi.com