Chapter 2 Content Switching Module with SSL Commands

predictor (serverfarm submode)

Usage Guidelines Use this command to define the load-balancing algorithm used in choosing a real server in the server farm. If you do not specify the predictor command, the default algorithm is roundrobin. Using the no form of this command changes the predictor algorithm to the default algorithm.

Note The nat server command has no effect when predictor forward is configured, because no servers can be configured.

The portion of the URL to hash is based on the expressions configured for the virtual server submode url-hashcommand.

No real servers are needed. The server farm is actually a route forwarding policy with no real servers associated with it.

Cache servers perform better using URL hash. However, the hash methods do not recognize weight for the real servers. The weight assigned to the real servers is used in the round-robin and least connection predictor methods. To create different weights for real servers, you can list multiple IP addresses of the cache server in the server farm. You can also use the same IP address with a different port number.

Note The only time the sequence of servers starts over at the beginning (with the first server) is when there is a configuration or server state change (either a probe or DFP agent).

When the least connection predictor is configured, a slow-start mechanism is implemented to avoid sending a high rate of new connections to the servers that have just been put in service. The real server with the fewest number of active connections will get the next connection request for the server farm with the leastconns predictor. A new environment variable, REAL_SLOW_START_ENABLE, controls the rate at which a real server ramps up when it put into service. The slow start ramping up is only for a server farm configured with the “least-conns” method.

The configurable range for this variable is 0 to 10. The setting of 0 disables the slowstart feature. The value from 1 to 10 specifies how fast the newly activated server should ramp up. The value of 1 is the slowest ramp-up rate. The value of 10 specifies that the CSM would assign more requests to the newly activated server. The value of 3 is the default value.

If the configuration value is N, the CSM assigns 2 ^ N (2 raised to the N power) new requests to the newly active server from the start (assuming no connections were terminated at that time). As this server finishes or terminates more connections, a faster ramping occurs. The ramp up stops when the newly activated server has the same number of current opened connections as the other servers in a server farm.

Examples

This example shows how to specify the load-balancing algorithm for the server farm:

Cat6k-2(config-module-csm)# serverfarm PUBLIC

Cat6k-2(config-slb-sfarm)# predictor leastconns

This example shows how to configure a server farm, named p1_nat, using the least-connections (leastconns) algorithm.

Router(config-module-csm)# serverfarm pl_nat

Router(config-slb-sfarm)# predictor leastconns

Router(config-slb-sfarm)# real 10.1.0.105

Router(config-slb-real)# inservice

Router(config-slb-sfarm)# real 10.1.0.106

Router(config-slb-real)# inservice

 

Catalyst 6500 Series Switch Content Switching Module with SSL Command Reference

2-104

OL-7029-01

Page 126
Image 126
Cisco Systems OL-7029-01 manual 104