Technical data

6
Using DNS to Balance Work Load
This chapter describes how to use DNS to balance the network traffic on a
multihomed host or on network servers when you have multiple systems
providing the same network service.
TCP/IP Services provides two methods for balancing work load using DNS:
Load sharing using the default DNS method of round-robin scheduling.
Load balancing using the TCP/IP Services load broker. Load broker is a
configurable, calculated, load-balancing mechanism for distributing the work
load among DNS cluster members.
This chapter discusses how to use DNS to balance server work load and includes
the following topics:
DNS clusters (Section 6.1)
Round-robin scheduling (Section 6.2)
Load broker concepts (Section 6.3)
Load broker startup and shutdown (Section 6.4)
Configuring the load broker (Section 6.5)
Metric server startup and shutdown (Section 6.6)
Solving load broker problems (Section 6.7)
6.1 DNS Clusters
TCP/IP Services defines the term DNS cluster to refer to several A resource
records for a single host name. This could be the A resource records for a
multihomed host or the A resource records for one or more servers which are to
share a work load.
6.2 Round-Robin Scheduling
Round-robin scheduling is the default load-sharing method used by a DNS server.
If multiple resource records satisfy a query, the BIND server returns them each
time in a round-robin order. The round-robin scheme is a simple rotation where
client requests are passed from one cluster member to the next. The round-
robin scheme is also useful for MX records to share mail loads among multiple
equivalent gateways of the same MX preference.
Unlike the traditional load-balancing method, round-robin does not take into
account the current work load on the DNS cluster members and does not know
whether these hosts are up or down.
The following example demonstrates how round-robin load sharing works.
Using DNS to Balance Work Load 6–1