Information Technology Reference
In-Depth Information
It is important to note that packets directed to the router, such as BGP packets, are process-
switched. When a packet is destined for the router, the IP Input process queues the packet
for the next-higher layer for processing; in the case of BGP, that layer is TCP. The efficiency
of this process can directly affect the BGP performance. During convergence, for example,
TCP might receive a large number of ACK packets. If these packets are not delivered
to TCP in time, sessions might not be established. Chapter 3 goes into detail on how to
tune various parameters to avoid this type of situation.
Cache-Based Switching
Cache-based switching
is a more efficient switching mechanism that takes advantage of the
information gained from the first packet switched by a scheduled process. In this type of
switching, the IOS process currently running on the processor is interrupted to switch the
packet. Packets are switched on demand, rather than being switched only when the IP Input
process can be scheduled, as in the case of process switching.
The processor switches the first packet at the process level and creates an entry in the route
cache so that subsequent packets with the same destination address are switched based on
the cache entry. Switching packets based on the route cache requires less processing, which
allows the packet to be switched at the interrupt level. This is why cache-based switching
is also called
interrupt context switching
.
Compared to process switching, cache-based switching has the following advantages:
•
It switches packets as they arrive without the need to wait for the forwarding process
to be scheduled, which reduces delay.
•
Only the first packet to a destination needs to be process-switched to populate the
route cache, minimizing the amount of time spent performing CPU-intensive tasks.
•
Subsequent packets are switched based on the information in the route cache.
Several forms of cache-based switching are currently available in Cisco routers:
•
Fast switching
•
Optimum switching
•
Distributed optimum switching
•
NetFlow switching
The cache-based switching paths differ in how the information is stored in the cache. The
following sections briefly review these switching paths and their shortcomings.
Fast Switching
Fast switching
stores the forwarding information and MAC header rewrite string (the new
MAC header) using a binary tree for quick lookup and reference. You can display the
content of the fast cache using
show ip cache verbose
.
