NIS+ Servers and Clients

When client machines, users, applications, or processes need NIS+ information, they seek an active NIS+ server (master or replica) from which to get the needed data. On large networks, networks with many subnets, and networks that span wide-area links, you may be able to improve NIS+ performance by customizing server usage.

Default NIS+ Client Search Behavior

By default, if no server preferences have been set with the nisprefadm command, a client will first try to obtain the information it needs from an NIS+ server on the client's local subnet. If the client finds an active server on the local subnet, it obtains the information it needs from the first local server that responds. If no server is available on the local subnet, the client searches outside the local subnet, and obtains the NIS+ information it needs from the first remote server that responds.

On large, busy networks, this default search behavior may reduce NIS+ performance for one of two reasons:

• When multiple servers on a subnet are serving a large number of clients, the random nature of the client's default search pattern may result in some servers being over worked while others are under used.

• When a client has to seek an NIS+ server beyond the local subnet, it will obtain its information from the first server that responds even if that server is overworked, or linked to the client's subnet by a slower Wide Area Network connection such as a modem or a dedicated line that is already carrying heavy traffic.

Designating NIS+ Preferred Servers

The Solaris release contains a new feature – server-use customization – that allows you to control the order in which clients search for NIS+ servers.

With this new feature you can balance and customize server usage by:

• Specifying that clients prefer (search for) certain servers over others.

• Specify whether or not clients are permitted to use remote servers if no local servers are available.

The search criteria that you specify can be applied to all clients within a domain, all clients on a subnet, or to individual clients on a machine-by-machine basis.