Many network virtualization products appear to be aimed at the top 10,000 customers worldwide, accounting for their price as well as their published product direction. While this is a limited and myopic view, many claim it is for the best, their reason being that network virtualization is only really needed by the very large networks. The more I think about this approach, the more I believe it is incorrect. Let us be frank here. Most networking today, within many different organizational sizes, is a hodgepodge of technologies designed to solve the same problem(s) over and over: how to get data quickly from point A to point B with minimum disruption to service.
The hodgepodge of technologies creates confusion and a mess within the network. The technologies at layer 3 on up can cause performance and other issues with the routing of data. Let us look at a VCE Vblock a bit closer. The Vblock is designed to be dropped into the data center as an island of predetermined technology to run predetermined workloads in a predetermined way. Given that as its basic definition, I would also expect its networking to be self-contained, yet it is actually designed to use core-switching technologies to perform all layer-3 routing, bridging, etc. between the various network constructs within the Vblock. This implies that for a Vblock placement, the core-switching becomes a critical part of the overall architecture. This is understandable, as the makers of the Vblock fully understand that there are many ways to route data and that not all networks are clean, concise, or even well designed.
If we were to look at different-sized enterprises, we would find a wide variety of needs and requirements for network virtualization—a need to simplify networking:
- Small enterprises generally have very simple networking layouts managed by one or perhaps two people with minimal use of VLANs and other technologies. Where the cable goes is often how the data travels. In these cases, network virtualization could help by examining the network, determining how traffic moves about, and recommending some simple changes (even automated changes) in how data flows between devices to improve overall performance. Networking in this case is fairly simple, but there could be one or two layer-3 and higher constructs already in place that could be muddying the waters. Most small enterprises cannot afford to have the expensive CCIEs come in and fix their network; they may not even realize the network is a hodgepodge of unconfigured or even misconfigured technologies.
- Medium to large enterprises generally have quite a few networking administrators making the best use they can of VLANs, virtual routing, and other protocols. They try their best to maintain the network at peak efficiency at all times. However, they also have a need for network virtualization, not only to examine the increased complexity of the network and recommend changes, but also for new projects that drop in islands of technology within a data center, such as a Vblock. These organizations may also have two to three different hypervisors in use for various projects that also need to be brought into the network virtualization purview; these should be part of the network, not behind some gateway device because the network virtualization technology does not understand or does not live within the hypervisor virtual switches.
- Large enterprises have a wide variety of knowledge and network components as well as technologies in use at all times. These organizations have a need for network virtualization to simplify their networks, but they cannot just add network virtualization technology to every aspect of their environment, as they have hidden pockets of old technologies that cannot be disrupted: perhaps a mainframe or windows NT box that is still doing critical work. These enterprises have the greatest need to make sense of many years of network mechanisms that already exist within their data centers.
- Extremely large enterprises are the current targets of network virtualization technologies. These technologies are used to break past the normal limits of VLANs and other technologies by approaching the data center build as brand new.
Network virtualization should be not only a protocol, but also a set of architectures that can be used to expand all levels of network understanding and technologies and to apply automation to clean up or at least recommend suggestions for fixing badly designed environments. Network virtualization should not be an island of technology, but instead it should eventually become the normal technology within an organization, one that understands that there are many forms of networking underneath it: layer 2 extending from the wire into each hypervisor, for example. My requirements for network virtualization are therefore a bit different, yet also the same:
- Break past the arbitrary limits on network segmentation methods (aka VLANs)
- Analyze existing networks and recommend how to implement network virtualization to improve overall performance and delivery of data to and from the user as well as between application constructs
- Use automation to implement any recommendations with minimal or no disruption in service
- Extend into the virtual environments, and do not just sit within the physical layers
- Act as a way to unify networking technologies under one management umbrella
- Provide plenty of documentation on use cases and how to implement the technologies, as well as on which networking technologies can be used with the particular network virtualization constructs.
Ultimately, network virtualization is for all sizes of businesses and enterprises. Networking today is a mess and understood by very few. Some have a very narrow view of networking (data goes where the wires connect), others the more complex view that data only travels within VLANs unless bridged by some construct (firewall, bridge, etc.), and still others the view that data is routed within each switch using virtual routing protocols, shortest path bridging, tunnels to enforce data flows, and other networking technologies. In addition, network virtualization is there to simplify complex and extremely difficult-to-view networks. The latter is where VMware, Cisco, Avaya, Big Switch, Midokura, and others are concentrating, yet they should also consider the other sizes of organizations.
However, to implement network virtualization, the technology should act like ink in water: it should start in one place and slowly spread itself out through the network. We should think about adding network virtualization as a migration and integration with minimal disruptions, even if it is considered a disruptive technology.
Share this Article:
Latest posts by Edward Haletky (see all)
- Scale and Engineering - March 23, 2017
- SDS and Docker: The Beginnings of a Beautiful Friendship - March 21, 2017
- Security Operations Center: Not Just Visibility - March 14, 2017