Network Virtualization Key to IaaS Clouds
Here's why network virtualization is a better fit than traditional networking for IaaS cloud deployments.
Network virtualization delivers network services in a more scalable, manageable, and fault-tolerant manner than traditional networking can deliver. For this reason it is gaining traction among cloud architects, rapidly becoming a “must have” component of IaaS clouds.
So, what is network virtualization? When cloud architects speak of network virtualization they are typically referring to overlay-based network virtualization. With OBNV, network functionality is abstracted away from the underlying physical network and into the software. What this means is that network services like virtual switching, routing, and firewalling can happen in software at the “edge” of the network rather than relying on the physical core of the network to have this functionality available. What this means in practice is that a cloud architect can design the network with basic low-cost hardware, while providing higher layer functions in the software, where advanced switching, routing and firewalling can be be managed with greater ease and precision.
Why wouldn't I just use traditional networking in the cloud? When building an IaaS cloud, traditional networking faces two serious limitations: scalability, and management and automation.
In an IaaS cloud, it's very difficult to properly plan for the size of your tenant's deployment. Tenant Alice might initially have five VMs in the first month of operation but could grow her VM count to 1,000 if the initial deployment is successful.
Take the example of a router. Whether you use a physical device or a virtual machine to give Alice her router, the size of the router will need to change based on the number of VMs that are being serviced. Upgrading the router will invariably introduce service interruptions for the tenant. Furthermore these traditional devices aren't designed to deal with the micro-granularity and high churn that cloud environments require of them.
In contrast, OBNV simulates the desired network topology at the very edge of the network (ingress point) and delivers the traffic to the destination (egress point). Because all higher layer network services are handled in software at the edge, the only thing needed from the physical network is basic IP connectivity. The result is smooth, incremental scaling without network interruption.
Management and Automation
Using traditional networking within an IaaS cloud results in multi-vendor heterogeneous network devices that are difficult to provision and manage. The overhead involved in manually configuring network resources increases substantially as the size of the network increases. This is further exacerbated by the fact that traditional devices are not designed for deep integration with cloud management orchestrators such as OpenStack and CloudStack.
To solve this pain, OBNV provides an abstraction layer between the physical infrastructure and the network, permitting the underlying hardware to be disparate, multi-vendor and heterogeneous. Whatever the equipment, it only needs to be configured once, on installation of the cloud.
The best OBNV solutions are architected to play well with the popular cloud orchestrators, enabling cloud users to manage and configure virtual networks centrally. And because OBNW is software, it will evolve in tandem with all the other software components, with ever-increasing benefits. This high level of integration, and the automation that comes with it, substantially reduces operating costs.
As your organization looks to build an IaaS cloud, we recommend you place OBNV offerings high on your list of components to evaluate. OBNV is no longer in the future. It is here to help, and it is here to stay.
Ben Cherian is the Chief Strategy Officer of Midokura, a network virtualization company focused on the IaaS market. Previously, he ran the cloud business unit at a cloud services provider DreamHost. Reach Ben on Twitter at bencherian or LinkedIn at bcherian.