Information Technology Reference
In-Depth Information
Remote Site Aggregation
In the enterprise, it is very common to have a significant number of hub-and-spoke connections,
often used to connect retail locations, manufacturing locations, and remote offices. When a
BGP core is not employed, there is typically a single IGP for the entire network, which
makes handling remote site aggregation with geographically diverse backup connectivity rel-
atively simple. The addition of the BGP core results in breaking the network into multiple
regional IGP processes. If the remote site's primary connectivity and backup connectivity
terminate in different regions, the situation becomes more complicated.
It is common for remote site routers to be resource-constrained. This is because they are
responsible for handling low traffic rates. It does not make sense to run multiple routing
processes on the remote site routers.
The easiest and cleanest solution is to make the remote sites a separate region. It can operate
in its own IGP, which is responsible only for maintaining the hub-and-spoke topology. The
hub routers connect directly to the core routers and run eBGP with them. This separation
of the remote sites into their own IGP helps isolate the impact of link failure to remote sites.
The core routers should originate the default route via BGP to the hub routers, which in this
case are R13 and R14. The hub routers should redistribute this default into the IGP. This
redistribution point should be filtered to allow only the default route to be injected into
the IGP.
The method used to inject routes into BGP on the hub routers is a bit different than that used
for the standard regions. The standard regions often can summarize their prefixes and then
inject the summary into BGP using a network statement. The hub routers have a large num-
ber of prefixes, possibly numbering in the thousands, depending on the number of remote
sites. The hub routers should redistribute the IGP into BGP, making use of route filters to
prevent any unwanted routes from entering BGP, such as the default route, which was
injected into the IGP from BGP.
The hub routers should issue a default route to the remote site routers via the IGP instead
of allowing full routing information to propagate. This results in a significant increase in
scalability in the remote site's region by reducing the route propagation load in the hub
routers. It also prevents remote site router resources from bounding the size of the remote
site's region. If only the resources available at the hub sites bound the size of the remote
site's region, upgrading only the hub routers can increase the region's size.
Another benefit of allowing only the default route to the remote sites is that remote sites do
not act as transit for each other in a failure situation. The amount of bandwidth to a remote
site is often tailored to that site's needs by sizing the PVC. If that remote site acts as a transit
device to reach another remote site, this could overload the PVC and degrade performance
for the transiting remote site.
