Transformation of the Data Center & Building the Private Cloud, Part 2
Jim Frey is Vice President of Research, Network Management for Enterprise Management Associates (EMA) and is a featured guest blogger.
In Part 1 of this blog, we looked at the growing adoption of cloud and Software-Defined Datacenter (SDDC), but also the fact that network virtualization choices are many and varied, creating some uncertainty regarding how to proceed. In part 2, we will look at network virtualization options and how to align for best success.
Unchaining the Boat Anchor: Network Virtualization
Existing methods of network provisioning are too often manual and/or disconnected from those for deploying compute and storage resources in SDDC or cloud environments. So how do we virtualize the network component, so that it too can be fully automated? There are currently three primary approaches to virtualizing the network and each method is at a different stage of maturity and adoption:
SDN Overlays: Network overlays, such as VMware NSX and Alcatel-Lucent Nuage, use encapsulation to establish a virtual network that runs on top of the physical network, abstracting the details of underlying physical connectivity. You still need a physical infrastructure, of course, but most provisioning can be done completely independently of the traditional network. NSX has particular appeal to virtual systems administrators already familiar with VMware solutions since it allows for quicker, more dynamic provisioning of network resources on a tightly integrated basis with VM provisioning.
SDN Underlays: A second set of solutions, which EMA refers to as “underlay SDN” make physical network hardware platforms themselves more directly programmable. Normally using a two-tiered architecture of controllers and network delivery switches, these solutions can leverage a wide variety of southbound control protocols, such as OpenFlow (probably one of the best known), netconf, or even CLI proxies. Such architectures open the possibility of automated provisioning via “applications” that leverage the northbound APIs of the controllers. The OpenDaylight project is a great example attempts to standardize northbound APIs, but other examples include HP’s recently-announced SDN App Store.
NFV: The third approach, which can co-exist with either of the above techniques, is Network Function Virtualization (NFV). Essentially, NFV takes functions that are implemented at the network layer and traditionally deployed in-line via network appliance hardware and packages them in a pure software form factor without sacrificing performance and scale. This allows them to be invoked using virtualization technology and without requiring incremental hardware deployment. NFV is being driven most heavily by telecoms operators seeking value-add services and easier/faster scale-up (and scale-down) of service infrastructure. But it also makes sense for mixed/hybrid cloud infrastructures for enterprises, where similar network functions are desired across both on-prem and off-prem settings.
All three of these approaches are still evolving and most likely there will be no “one size fits all” answer, but rather a gradual adoption and transformation of the networking layer over time to leverage aspects of all three. Despite SDN’s lack of maturity, it is very much top of mind in enterprise IT organizations. In a recent study, “Managing Networks in the Age of Cloud, SDN, and Big Data: Network Management Megatrends 2014,” EMA found that 50% of enterprise networking pros were either actively researching or evaluating SDN technologies. A smaller percentage (just over 10%) claimed to have one or more in production. So interest is high, though actual usage is still in the “early adopter” phase of general market adoption.
Getting Hybrid Deployments Right
Hybrid IT infrastructures are already impacting how enterprise networks are managed and deployed, requiring rethinking/retooling to assure both high availability and high performance. The same EMA research study found that one of the greatest concerns with SDN technology was proper capacity management. Further, network performance and congestion were confirmed as the top causes for application performance. If such issues are not fully addressed during network virtualization, IT risks losing credibility with the organizations they support and even risks fueling the trend towards outsourced and shadow IT.
In the end, perhaps the most important focal point is application performance. EMA’s study found that application performance optimization was second only to server virtualization as one of the top priorities for IT executives and directors in terms of driving their current network management priorities. So how are enterprise IT departments going to address the demands of deploying hybrid infrastructures without sacrificing performance, with SDN and NFV still in their infancy?
Managing Performance Today
There are vendors that can help address and solve that problem today. For instance, Radware, who specializes in application delivery and security solutions, was an earlier adopter of virtualization technology and has been shipping virtual/soft ADC solutions for more than three years. The company just recently added an NFV-compliant version of its virtual solution Alteon VA for NFV for carriers. The company also introduced Alteon VA for AWS – a “pay-per-use” or “bring-your-own-license” solution specifically built for performance optimization of Amazon cloud deployments. These solutions take advantage of emerging network virtualization technologies and cloud friendly design and licensing models. Radware also offers a complete set of hardware, virtual and hybrid solutions that enable both carriers and enterprises alike to smooth the transition between tradition, fixed, on-prem infrastructure to emerging virtualized, hybrid environments.
Major technology shifts such as the virtualization of the networking infrastructure do not happen overnight. It takes years or even decades to come to complete fruition. IT pros need solutions that can embrace emerging technology while remaining compatible with existing infrastructure. Options like those offered by Radware allow testing the new emerging technology waters without sacrificing the stability and performance of core enterprise-grade network optimization.