Total marks for Lab2: 10 (5 for report and 5 for demo)

Overall suggestions:

• Many groups pointed out that IX is a multiaccess network so that route server will modify the next hop when redistributing routes to POPs. This is correct as Ethernet is a broadcast multiaccess network and each POP can reach each other over the IX LAN. Next hop rewriting will be carried on Frame Relay network also as it is a nonbroadcast multiaccess network (NBMA). But a POP may need to go through the route server in order to reach other POP. In this case, a next-hop-self parameter have to be set on route server so that the next hop will be set like peer-peer connections. Check Fig 12-15 of 'Using the BGP for Interdomain Routing'
• In BGP4, the ADVERTISER attribute may not be used. The NEXT_HOP rewriting are done by route server, not by route server client (like POP). I have dumped the route updates and found no ADVERTISER attribute defined in the updates.
• At the end of the lab demo, you are asked to traceroute from core router to internal server on other NSP. Some groups can do and some groups can't. For groups who cannot complete the traceroute, only one group tell me exactly what happen. The problem is that you have not advertised the P-P links where core routers is connected. For some group that haven't advertised the routes, they can do the trace also because other branch office advertise the summary route /16 which cover the whole NSP network. The most elegant answer is there is no need to advertise the internal P-P links to other NSP.
• Routes for IX LAN (e.g. 10.(6|12|28|36).160.0/24 in the lab) will not be advertised in real world IX.

Group	Report	Demo	Total	Comment
A-HK	4.5	5	9.5	BGP path selection is not clearly explained. Link cost is not considered in BGP.
A-SH	4	5	9	Not know clearly how IX works and how it can be scaled (support more NSP in single IX) easily.
A-BJ	3	5	8	If the POP refuses the incoming routes with netmask > 24, why redist connected network in core router? Concept of IX is totally wrong. Multilateral peering does not means POPs are peered with each other. (As you have done to config your POP). BGP config at POP not attached.
A-CD	5	5	10	Mark for version 1 is 0.5 which is just a copy of Cisco menu. The design of IX cannot reduce the size of BGP database as route server and all POPs will carry all routes advertised eventually.
B-HK	4.5	4	8.5	example of shortest AS-PATH selection is wrong. 10.2.128.1 is not in your network.
B-SH	4	5	9	why not peer with other branch in same NSP?
B-BJ	5	5	10	good discussion on relation between ADVERTISER and NEXT_HOP attribute. But it may not be the case in BGP4.
B-CD	5	5	10	in IX, the route updates are advertised from route server to POP. Not 'requested' by the POP explicitly.
C-HK	3	4	7	'redist connected' is not used to link the OSPF route to BGP route. 10.6.160.0/24 is the HKIX, not the whole NSP C HK network. don't know how IX works and how it can be scaled. BGP config in Part II is not the same as in appendix.
C-SH	2.5	1.5	4	"redist connected" is not a bridge OSPF & BGP don't know how IX works no explanation on path selection, just copy the manual definition
C-BJ	3.5	4.5	8	need futher explanation on IX with example esp. the next hop attribute more examples on path selection needed
C-CD	2	4.5	6.5	explain the route you advertised. Why inject such routes? don't know how IX works for path selection, should provide live example found in the lab, not from manual or tutorial
D-HK	5	5	10	Frame relay network is an non-broadcast multiaccess network (NBMA).
D-SH	3	5	8	don't know how IX works examples on BGP route selection needed
D-BJ	4	5	9	don't know how IX works on the discussion of IX, the traceroute provided is OK. But the you should investigate why the route server is not included in the trace. to advertise BGP routes in core router, you can use network command instead of redist static / ospf and with route-map