Carrying the Load
A load balancer distributes traffic among a number of servers or connections, either to increase the overall capability of a system to handle heavy traffic or to increase fault tolerance so that no single failure of a server or network connection brings down a critical website or application.
F5's BIG-IP Local Traffic Manager Virtual Edition (VE) has the same capabilities as other F5 appliances but runs on a virtual machine rather than requiring separate hardware. It offers a sophisticated set of features that include support for maintaining database sessions; automatic rewriting of HTTP and HTTPS headers; multiple load-balancing algorithms to ensure traffic is spread out equally across available servers; and the ability to direct traffic based on content.
Setting up the BIG-IP VE requires a VM with more resources than a typical physical server requires, including two virtual CPUs, 2 gigabytes of RAM, one virtual Flexible network adapter (such as PCnet32 LANCE) for management, at least two virtual VMXNET3 network adapters for traffic balancing and one 40GB SCSI disk.
The software comes as an .ova file. For setup, deploy an Open Virtualization Format template, choose "deploy from file" and browse to the .ova file. The network administrator can then perform the basic system setup and power on the virtual appliance.
Advantages
Installation couldn't be simpler; the system is a snap to configure and deploy for load balancing across multiple servers.
Using vCenter and some scripting, it's even possible to automatically deploy additional web servers or application Âservers on multiple VMware hosts as the load increases, and remove them or power them off as the load diminishes. By doing so, the BIG-IP maintains optimum utilization of resources regardless of spikes in traffic from the web or internal users.
Because the load balancer is running on a hypervisor, it may be necessary to spread web or application servers across multiple hypervisors to ensure that loads don't become excessive. If the sole purpose is to increase fault tolerance and the website is relatively lightly loaded, this won't be an issue.
Why It Works for IT
The F5 load balancer can be purchased and deployed quickly and tailored to required loads with licenses based on throughput. It also lets an administrator set up a test bed with a number of virtual web or app servers to test things such as application awareness or handoff of apps from one server to another without involving a large number of systems.
A significant benefit is the ability to obtain and test a load balancer without investing in hardware. If loads increase, a network admin can move to an F5 appliance without any change in configuration and avoid the learning curve that typically accompanies rollout of a new product.
Additionally, license pricing is tied to required throughput levels, which means it's possible to upgrade to higher bandwidth if demands increase.
Disadvantages
The load balancers run on a VM that resides on the same physical server as one or more of the application Âservers, resulting in somewhat less redundancy than with a separate appliance. If the server fails, the cluster nodes and the load balancer fail as well.
In addition, if overall load to the nodes is an issue, then increasing the load on the clustered website will increase loads on both the node and the virtual appliance hosted on the same physical server.
In a rapid-turn or prototyping environment, however, a virtual appliance is an excellent way to evaluate both the functionality of a load balancer in general and the specific features of the BIG-IP in particular because the feature set on the virtual appliance is the same as on F5 physical appliances. Best of all, the virtual appliance is less expensive than a physical one.
