Lessons Learned
Like Sisyphus, who never did manage to roll the huge bolder to the top of the hill, we frequently push for promising new technologies and while we enthusiastically engage, we need to recognize that other technologies start compromising our effort. Examples? Think about all the brilliant brain power that went into ATM! It’s specification and development of respective products kept half a generation of data experts busy. Today, it has almost disappeared with some remaining legacy installations creating headaches for the operational staff.
Did Sisyphus Communicaticus identify with Network Function Virtualization (NFV) a new technology that needs to be pushed to the top? Will he make it or will we talk about the useless attempt sometime in the future?
NFV Promises
Network Function Virtualization aims to replace dedicated communication and IT network elements by software applications running on standardized multi-purpose servers. These servers could be operated centrally in the cloud or could also be hosted in an aggregation or demarcation device as part of the communication infrastructure. Central hosting is considered an operational and economic advantage while decentralized hosting improves scalability and reduces latency.
Haven’t We Seen This Before?
Didn’t we apply such centralized “virtual” architecture before and eventually move away from this concept? Large scale mainframes in combination with thin terminals connecting to it have been the client-server architecture with medium and large enterprises as well as with the public sector until the ‘80s, when the personal computer challenged this “virtualized and centralized” architecture.
The decentralized multi-purpose PC proved to be very attractive for small businesses. The relative openness of the architecture triggered application software innovation making use of the local processing power and graphical capabilities. This approach could better cope with limited connection bandwidth between the central server and the local client, as PC processing mainly happened at the network edge. Hence, this architecture became mainstream and the IT and communication networks aligned with it.
NFV is Not Just Replacing an Existing Technology
Today, many enterprises operate private data centers and interconnect their sites through a (WAN) based on their own private network or public L2/L3 connectivity services. Some connections to the corporate network run over the Internet. Web pages might be hosted on company servers requiring Internet access into the corporate network. Security becomes an increasing challenge, which results in various protective measures such as implementation of firewalls, intrusion detection systems and network address translation among others.
The increasing complexity of the IT infrastructure opens up opportunities for creating synergies by moving functions to a common location and handling complexity once rather than having the required competence replicated across various enterprise companies. The cloud becomes the entity which provides data storage, processing and application hosting.
NFV suggests – among other things - replacing dedicated network elements operated at an enterprise premises with software appliances operated in the cloud on standard servers. Such architecture creates considerable synergies, which results in significant CAPEX but also OPEX savings.
The virtualization of network functions impact not only the architecture of a network but also the operational processes and the organizational responsibility. It is a technical innovation which needs to go hand in hand with an operational process re-engineering. The NFV-initiated network transformation is a comprehensive system integration and consultancy project that requires sophisticated services skills.
This is good news for operators who have such SI and consultancy skills in-house but also for major services companies which will guide and implement the network transition with operators who don’t have this required skill set.
It is likely that the introduction will happen in steps guided by the most promising applications such as CG-NAT, firewalls, IDS as well as caching. There is, however, some risk that such an approach will result in application silos especially if solutions are designed monolithically and provided by a single company.
Open architectures with standardized interfaces will stimulate the creation of eco-systems which combine the skills, competence and expertise of a multitude of fast moving technology companies. And large scale system integrators will become instrumental in moving eco-system solutions from demo to practical implementation.
NFV Impacts the Communication Network
The connectivity network is already “virtualized” today as common physical instances are shared among many users by means of Virtual Private Lines in Ethernet, LSPs in MPLS or virtual containers in good ol’ SDH. However, the NFV discussion will also impact the technology applied in the connectivity network.
As outlined above, the enterprise company so far connects their sites by means of L2/L3 VPN services. Operators offering these connectivity services need to handle a high number of sites and are eager to implement automated provisioning processes. MPLS, which today is mainly applied in the core, is pushed to the edge of the network as routing and signaling protocols can provide such automation.
However, NFV impacts the connectivity requirements as communication is now between connected sites and the centralized data center rather than among the sites. The any-to-any paradigm which guided the development of connectionless IP becomes replaced by an any-to-few paradigm, which is much more static in nature and favors connection oriented connectivity technologies. Hence, there is potential for simplification of the connectivity network. Centrally provisioned, connection-oriented Carrier Ethernet technology allows combining cost advantages with a higher level of security due to less complexity.
For communication networks, the NFV transition will happen in two steps. The initial phase is guided by capitalizing on the advantages of network simplification by moving to centralized provisioning with open interfaces for seamless integration and lower cost connectivity based on L2 rather than L3 technologies. The demarcation devices will provide network demarcation functionality with a focus on OAM, resilience, minimizing size and power consumption and in some cases, security features.
In the second step, the NFV architecture will be further optimized by distributing those functions from the core of a network to the edge, which improves scalability of the network and improves performance via faster response times. Most likely this will be caching functions and functions which reduce latency among devices that connect to a common demarcation device.
In Summary
The title of this blog brought up the question whether NFV will re-define the enterprise and communication network architecture or whether this technology will soon disappear….and might come back in a few years creating astonishment, interest and enthusiasm similar to Halley’s Comet.
NFV has some key ingredients of being a real game changer: it is not a technology innovation, which substitutes a legacy technology with a superior one but allows operators to extend their service offering substantially and as such has the potential to answer the pending question of how operators shall defend against the OTT driven threat which depletes their value chains. Hence, NFV is an evolutionary necessity and will come inevitably, as it allows operators to escape a less promising bitpipe destiny.