|
|
Fast switching allows higher throughput by switching a packet using a cache created by the initial packet sent to a particular destination. Destination addresses are stored in the high-speed cache to expedite forwarding. Routers offer better packet-transfer performance when fast switching is enabled. Fast switching is enabled by default on all interfaces that support fast switching.
Use the commands in the following sections to configure appropriate fast-switching features:
Fast Switching is not supported for the X.25 encapsulations.
AppleTalk access lists are automatically fast switched. Access list fast switching improves the performance of AppleTalk traffic when access lists are defined on an interface. Refer to the "Configuring AppleTalk" chapter in the Cisco IOS AppleTalk and Novell IPX Configuration Guide for guidelines on creating and using access lists and configuring AppleTalk.
Fast switching involves the use of a high-speed switching cache for IP routing. Destination IP addresses are stored in the high-speed cache to expedite packet forwarding. In some cases, fast switching is inappropriate, such as when slow-speed serial links (64K and below) are being fed from higher-speed media such as T1 or Ethernet. In such a case, disabling fast switching can reduce the packet drop rate to some extent. Fast switching allows outgoing packets to be load balanced on a per-destination basis.
To enable or disable fast switching, use either of the following commands in interface configuration mode:
| Command | Purpose |
|---|---|
ip route-cache | Enables fast switching (use of a high-speed route cache for IP routing). |
no ip route-cache | Disables fast switching and enable load balancing on a per-packet basis. |
You can enable IP fast switching when the input and output interfaces are the same interface. Thisnormally is not recommended, though it is useful when you have partially meshed media such as Frame Relay. You could use this feature on other interfaces, although it is not recommended because it would interfere with redirection.
Figure 6 illustrates a scenario where this is desirable. Router A has a data link connection identifier (DLCI) to Router B, and RouterB has a DLCI to Router C. There is no DLCI between Routers A and C; traffic between them must go in and out of Router B through the same interface.
To allow IP fast switching on the same interface, use the following command in interface configuration mode:
| Command | Purpose |
|---|---|
ip route-cache same-interface | Enables the fast switching of packets out of the same interface on which they arrived. |
By default, Cisco IOS software switches IPX packets that have been directed to the broadcast address. To enable fast switching of these IPX-directed broadcast packets, use the following command in global configuration mode:
| Command | Purpose |
|---|---|
ipx broadcast-fastswitching |
By default, SMDS fast switching is enabled.
To re-enable fast switching if it has been disabled, use the following commands in interface configuration mode:
| Command | Purpose | |
|---|---|---|
Step1 | interface type number | Defines the type and unit number of the interface, and enter interface configuration mode. |
Step2 | encapsulatio n smds | Sets SMDS encapsulation. |
Step3 | ip route-cache | Enables the interface for IP fast switching. |
Step4 | ipx route-cache | Enables the interface for IPX fast switching. |
Step5 | appletalk route-cache | Enables the interface for AppleTalk fast switching. |
Fast switching uses a cache created by previous packets to achieve a higher packet throughput. Packet transfer performance is generally better when fast switching is enabled. Fast switching also provides load sharing on a per-packet basis.
By default, fast switching is enabled on all interfaces that support fast switching. However, you may want to disable fast switching to save memory space on interface cards and to help avoid congestion when high-bandwidth interfaces are writing large amounts of information to low-bandwidth interfaces. This is especially important when using rates slower than T1.
Fast switching is not supported on serial interfaces using encapsulations other than HDLC.
 |
NoteTurning off fast switching increases system overhead. |
For some diagnostics, such as debugging and packet-level tracing, you will need to disable fast switching. If fast switching is running, you will not see packets unless they pass through the route processor. Packets would otherwise be switched on the interface. You might want to turn off fast switching temporarily and bypass the route processor while you are trying to capture information.
This section includes the following topics:
To disable AppleTalk fast-switching on an interface, use the following command in interface configuration mode:
| Command | Purpose |
|---|---|
no appletalk route-cache |
Fast switching is enabled by default on all interfaces on which it is supported.
To disable fast switching on an interface, use the following command in interface configuration mode:
| Command | Purpose |
|---|---|
no vines route-cache |
By default, DECnet routing software implements fast switching of DECnet packets.
To disable fast switching of DECnet packets, use the following command in interface configuration mode:
| Command | Purpose |
|---|---|
no decnet route-cache | Disables fast switching of DECnet packets on a per-interface basis. |
To disable IPX fast switching, use the following command in interface configuration mode:
| Command | Purpose |
|---|---|
no ipx route-cache |
ISO CLNS fast switching through the cache is enabled by default for all supported interfaces. To disable fast switching, use the following command in interface configuration mode:
| Command | Purpose |
|---|---|
no clns route-cache |
|
NoteThe cache still exists and is used after the no clns route-cache interface configuration command is used; the software does not do fast switching through the cache. |
To disable XNS fast switching on an interface, use the following command in interface configuration mode:
| Command | Purpose |
|---|---|
no xns route-cache |
The high-speed route cache used by IP fast switching is invalidated when the IP routing table changes. By default, the invalidation of the cache is delayed slightly to avoid excessive CPU load while the routing table is changing. To control the route cache, use the appropriate commands in the following sections:
To control route cache invalidation, use the following commands in global configuration mode as needed for your network:
| Command | Purpose |
|---|---|
no ip cache-invalidate-delay | Allows immediate invalidation of the cache. |
ip cache-invalidate-del ay [minimum maximum quiet threshold] | Delays invalidation of the cache. |
 |
CautionNormally, this task should not be necessary. It should be performed only under the guidance of technical staff. Incorrect configuration can seriously degrade the performance of your router. |
You can display the contents of IP routing tables and caches. The resulting information can be used to determine resource utilization and to solve network problems.
To display system and network statistics, use the following command in privileged EXEC mode:
| Command | Purpose |
|---|---|
show ip cache [prefix mask] [type number] | Displays the routing table cache used to fast switch IP traffic. |
Adjusting the route cache allows you to control the size of the route cache, reduce memory consumption, and improve router performance. You accomplish these tasks by controlling the route cache size and route cache invalidation. The following sections describe these optional tasks:
You can limit the number of entries stored in the IPX route cache to free up router memory and aid router processing.
Storing too many entries down in the route cache can use a significant amount of router memory, causing router processing to slow. This situation is most common on large networks that run network management applications for NetWare.
For example, if a network management station is responsible for managing all clients and servers in a very large (greater than 50,000 nodes) Novell network, the routers on the local segment can become inundated with route cache entries. You can set a maximum number of route cache entries on these routers to free up router memory and aid router processing.
To set a maximum limit on the number of entries in the IPX route cache, use the following command in global configuration mode:
| Command | Purpose |
|---|---|
ipx route-cache max-size size | Sets a maximum limit on the number of entries in the IPX route cache. |
If the route cache has more entries than the specified limit, the extra entries are not deleted. However, they may be removed if route cache invalidation is in use. See the "Controlling IPX Route Cache Invalidation" section in this chapter for more information on invalidating route cache entries.
You can configure the router to invalidate fast switch cache entries that are inactive. If these entries remain invalidated for one minute, the router purges the entries from the route cache.
Purging invalidated entries reduces the size of the route cache, reduces memory consumption, and improves router performance. Purging entries also helps ensure accurate route cache information.
You specify the period of time that valid fast switch cache entries must be inactive before the router invalidates them. You can also specify the number of cache entries that the router can invalidate per minute.
To configure the router to invalidate fast switch cache entries that are inactive, use the following command in global configuration mode:
| Command | Purpose |
|---|---|
ipx route-cache inactivity-timeout period [rate] | Invalidates fast switch cache entries that are inactive. |
When you use the ipx route-cache inactivity-timeout command with the ipx route-cache max-size command, you can ensure a small route cache with fresh entries.
Some IPX end hosts accept only even-length Ethernet packets. If the length of a packet is odd, the packet must be padded with an extra byte so that end host can receive it. By default, Cisco IOS software pads odd-length Ethernet packets.
However, there are cases in certain topologies where non-padded Ethernet packets are forwarded onto a remote Ethernet network. Under specific conditions, you can enable padding on intermediate media as a temporary workaround for this problem. Note that you should perform this task only under the guidance of a customer engineer or other service representative.
To enable the padding of odd-length packets, use the following commands in interface configuration mode:
| Command | Purpose | |
|---|---|---|
Step1 | no ipx route-cache | Disables fast switching. |
Step2 | ipx pad-process-switched-packets |
Posted: Mon Jul 17 16:48:23 PDT 2000
Copyright 1989-2000©Cisco Systems Inc.