Barracuda Networks today unveiled a 40G load balancer appliance designed to offload network and security traffic from virtual machines.
Designed to be deployed at the edge of the data center, the Barracuda Load Balancer Fast Distribution Controller (FDC) is based on technologies from Intel that enable IT organizations to better scale virtual data center environments.
Sanjay Ramnath, senior product manager for Barracuda Networks, says the Barracuda Load Balancer FDC is optimized for data center environments that need high throughput to scale. By offloading the load balancing function from the hypervisor, Ramnath says it becomes possible for data centers running hundreds of virtual machines.
Priced starting at $20,000, the Barracuda Load Balancer FDC makes use of a combination of Intel processors and an Intel Data Plane Development Kit (DPDK) to direct requests to application workloads running on a particular server more efficiently.
“Hypervisors on their own just can’t keep up with load balancing,” says Ramnath. “This appliance runs heads and shoulders above any other load balancer we have.”
According to a report from Principled Technologies, a consulting firm contracted by Barracuda Networks, the Barracuda FDC can actually provide up to 60.71 Gb/s of throughput and that the Barracuda FDC handled 9.99 million simultaneous web connections and 1.33 million connections per second.
In addition, because the Barracuda Load Balancer FDC makes use of Intel processors Ramnath says Barracuda has been able to install an instance of its firewall software directly on top of the appliance.
Collectively, Ramnath says the Barracuda Load Balancer FDC is a new element of an overall application delivery network architecture that makes it possible to federate the management of load balancing across appliances, hypervisors and physical servers.
In some ways, relying on a physical load balancer appliance is a step back towards the future. With the rise of virtualization, many IT organizations moved the load balancing function on to a virtual machine. But as data center environments scale Ramnath says the virtual switches inside hypervisors can’t process all the requests being made as efficiently as a dedicated appliance.
As workloads become more distributed Ramnath contends that IT organizations will increasingly be required to deploy load balancing capabilities at the edge of the data center. And given the number of connections that need to be managed Ramnath says that going forward it won’t be too long before 40G Ethernet connections are more the rule than the exception as the number of physical and virtual servers deployed inside and out of the cloud continue to multiply.