526 Chapter 14: SNA Topologies
When designing for high availability, keep these rules in mind:
·
TCP encapsulation provides nondisruptive recovery from link failures between peers.
·
Multiple active peers or ports allow for fast dynamic recovery from loss of peer, TIC,
or FEP.
·
Use backup peers for disaster recovery sites or where dual paths are cost-prohibitive.
When designing for very large networks, try to implement "promiscuous" peers where possible.
The following are the benefits of promiscuous peers:
·
Configurations at the central site are reduced because fewer commands are needed with
promiscuous peers.
·
NVRAM requirements are minimized because the only command needed is the
promiscuous command.
·
When adding remote sites, there is no need to configure or change the configuration of the
central site.
TIP
Each page is worth about 2 K of NVRAM when you do a show run command. For example, if
the router has 32 K of NVRAM, it will display roughly 16 pages at the most of configuration
information. CCDPs need to know their NVRAM limitations!
Designing Peer Group Size (50 or More)
When deciding how large peer group sizes should be, remember that a single broadcast by a
peer router must be replicated to every peer in the group and to all border peers. For example,
in Figure 14-14, Border Peer Doug must replicate any broadcast by Router Greg to Laurie and
Jeff. Border Peer Doug must in turn replicate to Border Peer Connie, who replicates to Holly,
Melissa, and Mike. Router Connie replies that it can reach Resource Mike. That makes for three
broadcasts (one each to Connie, Greg, and Jeff) by Router Doug if the resource is not found in
cache. As a rule of thumb, a router with 50 or fewer peers can probably be all in one group. It
is when scaling higher--say, 1000--that a single peer group is not a good idea. Each broadcast
would have to be replicated 999 times.
87200333.book Page 526 Wednesday, August 22, 2001 1:41 PM