The Rise in Popularity of BIG-IP Load Balancers

31 May 2009 by Simon Haslam (in HA)

In my experience the vast majority of high availability Oracle application server installations use hardware load balancers. Whilst there are software alternatives, Web Cache 10g in particular for Oracle environments, several factors make hardware appliances attractive:

• they are self contained and easily sized,
  
• widespread usage in webserver farms makes them a known quantity,
• they tend to be treated as 'fit and forget',
• relatively inexpensive,
• basic load balancing is a mature technology.
  

The most popular load balancers I come across are made by Cisco and F5 Networks, though I do see Alteon (now Nortel) and Juniper from time to time. Recently Netcraft published an article showing their observation of the increase in popularity of F5 BIG-IP:

Note that these are only public-facing web servers - I'm sure there will be many more behind corporate firewalls, for example delivering call centre applications.

Hardware load balancers are usually very easy to install, but much more difficult to set up properly such that a highly available system responds in a timely manner to component failure.

Within most organisations the load balancer is uncommon compared to routers, firewalls and switches, and once installed, not often changed. This usually means that network admininstrators are often not too familiar with their configuration. The Oracle Middleware Administrator needs to understand how load balancers work and talk to her network team to ensure that the networking layer is appropriately configured to meet the application's SLA. Whilst this can be an iterative, and sometimes tedious, process the benefit is that you will end up with a robust middleware platform that should live up to management expectations for availability during both planned system maintenance and unexpected hardware failure.