4.5. Advanced Traffic Managers
While the server nodes are important to the operation of the cluster, the ATMs are critical. The primary ATM does all the work of forwarding incoming traffic to the cluster nodes. Backup ATMs stand ready to take over if the primary ATM is unable to fulfill its duties.
The `Advanced Traffic Managers' menu allows you to configure the systems that will act as primary and backup ATMs. Note that you cannot necessarily determine which one will be the primary ATM and which will be the backups. The first ATM system to come up will generally become the primary, and all the other ATMs will become backups. If two ATMs come up at the same time, the one listed first in the configuration file will become the primary ATM.
There are two parts to configuring ATMs: configuring the systems that are to be ATMs, and configuring the settings.
4.5.1. Advanced Traffic Manager Systems
The `Advanced Traffic Manager Systems' menu lets you define the computers that you want to act as ATMs. The list of traffic managers works like the other lists in the configuration program. There are `Add', `Edit', and `Remove' buttons. To add a system as an ATM, click `Add', then enter the IP address or full hostname of the system.
Don't forget that one system can work as both an ATM and a cluster node at the same time. The list of ATMs and list of server nodes are completely independent of each other.
4.5.2. Advanced Traffic Manager Settings
Selecting the `Advanced Traffic Manager Settings' menu allows you to set some parameters for the ATMs in the cluster. We'll touch on these settings in this section, but you can generally accept the defaults when configuring your cluster. For more information on tuning these settings, consult Chapter 7.
At the top of the screen, you will see all the ATMs listed. This ATM pool is similar to the server pools, except that there is currently no way to create more than one ATM pool -- all ATMs are automatically added to a single pool.
Following the list of ATMs you will see the ATM settings. Here is what each of these setting mean:
The `Send ARP delay (frequency)' setting tells the primary ATM how often to send out an ARP broadcast. The ARP broadcast lets other systems on the network know that the primary ATM is associated with the virtual IP address of the cluster.
The `HeartBeat Frequency' is the number of seconds between heartbeat broadcasts. The primary ATM generates heartbeats to let the backup ATMs know that it is working.
The backup ATMs listen for the heartbeat broadcasts from the primary ATM. If they miss `Max. missed heartbeats' in a row, the backup ATMs assume that the primary is down.
The `Number of services', `Number of servers', and `Number of connections' values specify the size of the kernel tables that should be used. The defaults should be okay, unless you are expecting a lot of traffic or have a large number of servers in your cluster. Make sure that these values are larger than the maximum that will be needed for the server to run.
The `Number of services' setting is just like it sounds -- how many different services are defined, as listed in the `Service Settings' list in the configuration program.
The `Number of servers' is actually the number of server nodes times the number of services that each server handles.
The `Number of connections' value is the most critical. If the number of concurrent active connections ever exceeds this number, all further connection attempts will fail until the number of connections falls below this value.
In most instances, the default settings should work for you.
The `Connection timeout' setting tells the ATM how long to leave a connection in the connection table after the connection has closed or has not had any packets sent. It is given in seconds. The default value of 300 should be okay for almost all services. You may want to fine tune it if you are only running one or two services that are expected to have shorter connection times.
 | For help in determining the optimum values for these parameters, see Chapter 7. |