Network function performance in a virtualized world
Over the course of the last 15 years, both enterprises and service providers have come to embrace the benefits of server virtualization. The benefits are multi-fold – ranging from increased efficiency and return on investment, and greater business agility, to streamlined management and automation – and have driven market adoption to the point where greater than 75% of all x86 servers are now virtualized.
With the era of virtualization in full swing, IT is doubling down and quickly moving to find additional use cases for virtual machines. In addition to application workloads, businesses are now looking to use VMs to run networking functions. The benefits are apparent. Rather than a multi-vendor mix of hardware appliances, a homogenous array of general-purpose servers may be deployed to save on space, power, cooling and hardware costs. In addition, networking functions can be deployed when and where they are needed, and automated within a centralized management framework for operational efficiency.
However, there are some drawbacks. General-purpose servers were not designed run compute-intensive networking functions. In addition, running a hypervisor on a general-purpose server further eats into available resources. When dealing in smaller workloads or highly distributed workloads, the performance offered by virtualized servers may be sufficient. But, for high-traffic enterprise applications and cloud services with large numbers of end users, using general-purpose servers to run networking functions can be prohibitively expensive and result in performance bottlenecks.
Take the example of the load balancer or application delivery controller, an essential networking component that provides high-availability, performance optimization and security for Web applications and services. It terminates requests, offloads SSL encryption and performs application-level inspection for thousands and often millions of end users. While the promise of business agility and operational efficiency is compelling, neither outweighs the cost of achieving sufficient levels of performance.
To bridge this gap and deliver the agility of virtualization with the performance of dedicated appliances, new approaches will be required. One such approach is the virtualized networking appliance. Virtualized networking appliances use the same purpose-built components as dedicated networking appliances; however, they run a hypervisor that supports multiple virtual machines and multiple network functions.
For instance, the virtualized networking appliance could support in excess of 50 virtual machines, could support different size virtual machines and could support different networking functions including load balancing, Web application firewall, SSL VPN and WAN optimization. In addition, the appliance capacity could be purchased incrementally on a pay-as-you basis in line with cloud utility principles. What’s more, the virtualized appliance would feature strong integration with common automation and orchestration frameworks including VMware VCenter, Microsoft System Center and OpenStack to retain the agility of virtualized environments.
On the flip side, the virtualized appliance allocates system resources in a manner that not only provides superior performance, but also ensures that each virtual machine provides guaranteed performance. Every virtual machine is allocated dedicated CPU, SSL cores, memory and I/O and system resources are reserved for hypervisor management to fully eliminate the possibility of VM contention. In this manner, a consistent environment for virtualized network functions may be created, one that successfully combines the agility of cloud and virtualization with the performance of dedicated appliances.
For enterprises and service providers, the benefits are apparent. In an enterprise environment, applications that require guaranteed performance and service level agreements can now be deployed on a platform that delivers a far greater degree of business agility, operational efficiency and return-on-investment. In a service provider environment, particularly today’s new breed of infrastructure-as-a-service cloud providers, networking services – such as load balancing-as-a-service – may be offered from a highly-efficient, highly automated platform capable of maintaining and demonstrating customer SLAs.
As businesses move towards increasingly shared environments and the demand for cloud applications and services grows, it will be critical to deploy networking solutions that are scalable and adaptable and are capable of bridging the gap between the agility of cloud and virtualization and the performance of dedicated appliances. Virtualized networking appliances are an ideal solution to meet these emerging requirements. With the ability to support guaranteed performance in shared environments, virtualized networking appliances will see increased adoption in the coming years – allowing enterprises and service providers to efficiently and cost-effectively grow in proportion to customer demand and gain a competitive edge.
Paul Andersen is the director of marketing at Array Networks (www.arraynetworks.com). He has over 15 years’ experience in networking, and has served in various marketing capacities for Cisco Systems, Tasman Networks and Sun Microsystems. Andersen holds a Bachelor’s Degree in Marketing from San Jose State University.
Editor’s Note: In an attempt to broaden our interaction with our readers we have created this Reader Forum for those with something meaningful to say to the wireless industry. We want to keep this as open as possible, but we maintain some editorial control to keep it free of commercials or attacks. Please send along submissions for this section to our editors at: [email protected].
