When you configure RTMON to work with application servers, you must
When you configure RTMON to work with application servers, you need to configure one or more logical application server groups in the appliance RTMON configuration.
Application server groups have the following primary purposes:
Each group is identified by a logical group name of your choosing. You could, for example, assign a group of billing servers a name that would identify them as billing servers. Mappings of the group name to IP addresses is handled by DNS.
You might have different types of application servers that need the same data. Alternatively, you might have two different types of application servers that require different data. You define, by logical group, the data that the appliance is to send to servers in the group. See Planning for data that RTMON can send to application servers.
When multiple application servers are members of the same logical application server group, each server receives the same user events and user attributes. However, NetCache always sends a specific user's data to the same application server, which balances the load over the server as described for load balancing. See also What occurs during failover.
To balance the load of data across application servers in the same application server group, NetCache uses a weighting technique based on a history of connection round trips. Server CPU load and other server resources are not considered, nor are differences such as bandwidth used by the streams that specific clients are watching.
Initially, all servers have the same chance of being selected. If a server is consistently slower in responding to the appliance over time, the chance decreases that the appliance will send data to the application server.
In practice, this scheme has resulted in a fairly even distribution of data across application servers in an application server group. However, you should monitor the load on your application servers resulting from data pushed by the appliance and make adjustments to your server size or number of servers as necessary.
When you define an application group, you are instructing the appliance to distribute the data over all servers in the group. If one server becomes unavailable, NetCache continues to automatically distribute the data over the remaining available servers in the group.
The following occurs if an application server in an application server group becomes unavailable:
The NetCache appliance retains the data until the logs are deleted. By default, NetCache buffers data until it runs out of disk space. A rough estimate of when the appliance would run out of disk space is when the remaining disk space falls below 10 percent of the total NetCache volume size.
Network Appliance recommends that you set up HTTP with buffering so that the appliance can buffer data for resending when the server becomes available.
You can add application servers to application server groups as necessary to increase reliability and scalability. Network Appliance recommends that you do not exceed 10 application servers communicating with a NetCache appliance.
Consider the following when defining your application server groups:
Copyright © 1994-2002, Network Appliance, Inc. All rights
reserved.
Part No. 210-00356
File: dplyrtm2.htm
Last updated: 11/14/02 12:32:47