Basically, I am not fully technical expert in this field, because it is the consolidation from multiple technical areas. But clearly, network function virtualization can be treated as network service component transformation, not the hot topic of network virtualization. The concept model is from telecom industry and design for long term efficiency/economy business operation purpose. ( All the request what I received is almost from mobile service company). Some aggressively telecom company has started the experimental trail. Recently, I saw the news that AT&T has started trail and expect to reduce the rely on proprietary devices in future. （http://www.bloomberg.com/news/2013-09-23/at-t-shifting-to-lower-cost-software-defined-networking.html )
Basically, Telecom service company hopes to virtualize the service component from vendor proprietary hardware to run on open/customizable hardware. We can refer the experience from open compute project in the global industry, new X86 processor can offer more and more features and network application enhancement. These experience may take NFV vision more practical. Besides, the cost of ASIC design is increased and less features support on network application. From these industry change, it may be not the dream of NFV in Telecommunication.
.jpg)
Recently, SDN/Network Virtualization is becoming the hot topic in the network market, and people think that is fundamental of NFV. But my personally understanding on SDN/Network Virtualization is not the core part of NFV, it cloud be just option. NFV story has its own background purpose, not the new technology. I personally don't like to have that in my technical product pitch, because if you know the NFV story, then you can understand NFV spirit, that is not the offer what I can give today.
NFV can't virtualize full telecom services, people in the NFV design also understand this, so the reasonable use case is more important, if they want to get it success.
So if you understand above the use cases, you can understand the functions can be run into X86 system. It shouldn't be difficult to be virtualized into virtual appliance. Why does telecom company want to virtualize these function components? Nothing more than life cycle of proprietary hardware is too short, they need to do re-investment, redesign for same function in certain period, power consumption, space allocation is also the considered on business cost. It is all the business critical topic for the low profit telecom services. Innovation is the way to make business more efficiency. Until here, you may understand what " Network Function Virtualization" is, not related to Network Platform Virtualization.
Who will be benefited after NFV ? Major beneficial is probably telecom service company, and next is commercial hypervisor company. But I believe there is no free lunch, because this kind critical system must be pass series function, performance and compatibility testing to measure the competence of virtual appliance.
Once telecom service company get the function virtualization, the next consideration is about the virtual environment operation. The cloud dc operation and traditional function support system ( OSS/BSS) must be considered at the same time, because NFV is just to virtualize the service component, no change the function operate model. Some points may be aware
1) In each NFV topic, the network is always discussed. That's why I mention SDN/Network virtualization is the option. I don't know the the network platform virtualization is suitable to be leveraged in NFV environment, because network platform virtualization originally is designed for virtualized/cloud datacenter networking. We don't know the NFV will meet the same issue what network platform virtualization address, so that is open here for further real environment discussion.
2) Performance is also the concern in virtualization environment, but fortunately the vswitch technology is quite flexible to leverage various offload technologies in hardware level, for example: Linux NAPI or Intel DPDK
3) Automation is the key, if the NFV environment wants to be scaled.
4) OSS/BSS support will be kept as same, because NFV is to virtualize service component, not change the original telecom service profoliio.
5) Security level shouldn't be decreased, but after virtualization, this topic is becoming difficult to manage. so network platform vitalization/SDN may be the solution to keep security level and visibility.
Last, I think the NFV is probably the 1st step to optimize the service infrastructure component, because less and less system is not running on x86 system. Cost/Application on CPU is becoming more colorful than ASIC design. so what is the butterfly effort after NFV, I think we may see in not longer future, I hope to see that in the decade.
