Today’s companies, particularly those involved with banking and Internet hosting, require their network to be available 99.999% of the time. When the network isn’t available for employees, clients and business partners, the cost can be thousands of dollars per minute. Diversity or failover is a key component of any high availability strategy. It describes a backup device or link that is available should the primary device or link be unavailable.

Today’s e-commerce web hosting facility are designed with diversity for circuits, routers, firewalls, links, modules and servers. Each number shown corresponds to a single point of failure that is provisioned with some backup or diverse connection. Each server is dual homed to different switches should a Layer 2 switch or server link fail. The Layer 2 switches are connected with a Gigabit trunk. As well the Layer 2 switches are connected to multilayer switches with Gigabit trunks. That allows for link or switch failure. There is module diversity at all campus switches with dual supervisor engines. If the primary supervisor engine fails, the secondary activates and traffic is diverted across trunk lines to the adjacent switch once the spanning tree algorithm is run and the trunk ports are in forwarding mode.

Firewall failover is provided with a link between them that will detect if one of those is unavailable. Traffic is then routed through the active firewall. The routers are connected with a link that will detect when one of the devices is unavailable (HSRP). There are separate telecom demarcations at this facility, which provide diverse local loop circuits to different central offices. There are dual WAN circuits to different Internet Service Providers from those central offices as well.

High availability designs must consider the failover time at each point in their network. If a network application will timeout after 10 seconds then any failover point must be 9 seconds or less. For instance, spanning tree protocol (STP) will run when there is a topology change such as a switch or link failure. Design your network such that the link speeds, spanning tree protocol version and switch topology do not add an excessive failover time.

1. Different Internet Service Providers

2. Central Office Diversity with Diverse Local Loop Circuits

3. Router Diversity with Hot Standby Router Protocol (HSRP)

4. Firewall Diversity

5. Switch to Firewall Link Failover with Gigabit Trunking

6. Switch and Link Diversity with Gigabit Trunking

7. Switch and Link Failover with Gigabit Trunking 

8. Module Diversity with Dual Supervisor Engines

9. Dual Network Interface Cards at Servers and Clustering Software

Shaun Hummel is the author of Network Planning and Design Guide and has a web site focused on information technology job search solutions and certifications.

http://www.networkjobsolutions.com

Technorati Tags: gigabit network