Basic Concepts - Huawei S9700 Series Configuration Manual

S9700 Core Routing Switch
Configuration Guide - SPU

Basic Concepts

l
l
l
l
l
l
Issue 01 (2012-03-15)
– DNS-based load balancing: binds multiple IP addresses to a domain name. The DNS
server equally distributes user requests to servers.
– VIP-based load balancing: maps the real IP addresses of network devices to virtual IP
addresses. (A server or firewall group uses a virtual IP address.)
NOTE
This document classifies load balancing based on objects.
Load balancing
A technology binding network devices (such as servers and firewalls) or links into a group
and distributing network services to them. It improves service processing capacity and
service reliability.
Load balancing member
An entity in a load balancing group that provides services to users, such as a server and
firewall.
Load balancing group
A set of network devices or links that provide the same service for users. A load balancing
group may be a server group, a firewall group, or a link group.
Load balancing member instance
A load balancing member can join multiple load balancing groups and the mapping between
a load balancing member and a load balancing group is called load balancing member
instance. If the load balancing member is a server, the instance is called server instance; if
the load balancing member is a firewall, the instance is called firewall instance; if the load
balancing member is a link, the instance is called link instance.
VIP
A virtual IP address used for server and firewall load balancing. Multiple servers or
firewalls share a public IP address. Users access the servers or firewalls through the public
IP address, whereas the servers or firewalls use different internal IP addresses. The SPU
distributes the traffic destined for the virtual IP address to each real server according to the
load balancing policy.
Load balancing algorithm
Selects the optimal load balancing member to provide the best services for users. The
SPU supports the following load balancing algorithms:
– Weighted round robin (WRR) algorithm
In the WRR algorithm, the SPU makes load balancing decisions according to the
priorities and weights of load balancing members. The SPU selects load balancing
members with higher priorities for providing services according to the weights. The
load balancing member with a greater weight has a greater possibility of being selected
and is allocated more services. After selecting a load balancing member, the SPU
determines whether to use the member according to the bandwidth limit, connection
quantity limit, and connection rate limit. If no load balancing member with higher
priority is available, the SPU selects a load balancing member among those with lower
priorities according to the WRR algorithm.
If the server performance in the server group or link bandwidth in the link group have
great differences, consider the WRR algorithm.
– Least connection algorithm
The SPU uses the weight and least connection algorithm to perform load balancing.
Huawei Proprietary and Confidential
Copyright © Huawei Technologies Co., Ltd.
6 Load Balancing Configuration
182