 We looked at the big picture where we had operator A and operator B which were connected through the network-to-network interface. Could we allow a mechanism through which an operator could program the infrastructure of another operator through possibly a memorandum of understanding or through a contract? Well it is true. This is known as the network first function virtualization infrastructure as a service. The framework or the rules of engagement or implementation are going to be such that we have let's say two service providers A and B or one and two. Service provider one instantiates virtual network functions on the network function virtualization infrastructure on service provider two. So the since the user or the client is connected to network function virtualization infrastructure of A or one. So it is the responsibility of service provider one to provide end-to-end services across both NFVIs. So the advantage is going to be a multi-fold. First, because of geographical independence in terms of location and proximity, the latency and reliability requirements could be met for real-time services and for mission critical applications. This could also serve as an advantage to achieve compliance to the regulatory constraints by different government entities on location of processing and storage. Let's take an example of how possibly network function virtualization could be achieved in a specific scenario. So we've got two actors. So these two actors both agree to have the joint network function virtualization infrastructure. This is going to be like infrastructure as a service cloud computing environment, which is integrated or coupled with the network as a service infrastructure. So this is going to provide on-demand connectivity with a defined set of service termination and access points. So we can read the description here. We can understand that since all the virtual network functions are location independent, this capability shall improve the reliability and latency. A good comparison would be to have a look at the traditional approach. The traditional approach basically was manual, especially engineered for whole-sale services between two service providers. In this case, the service provider one was required to negotiate individual leased lines cloud computing resources with service provider two. And then at that particular moment, assembling and operating the infrastructure for end-to-end services required intensive coordination between the two service providers, especially if something went wrong and troubleshooting and corrective regime was required to be carried out. So the traditional approach now replaced with programmability approach has some advantages. Let's look at that as well. The network function virtualization infrastructure provides the compute nodes with infrastructure as a service capabilities. And these are provided at runtime for virtual network functions to be executed by service provider one. So it means service provider one gets real-time updates on the infrastructure because it's part of the programming implementation. And if there is something wrong, then without involving service provider two, service provider one could possibly take corrective actions. Now the role of programmability in this particular scenario is going to be more computing intensive. In this case, service provider one is going to be authorized. The network as a service has to be provided on demand by service provider two with authorized termination within the network function virtualization infrastructure of service provider two. Then some kind of separation between data and control planes have to be implemented because the service provider one is now accessing the network function virtualization infrastructure of two, not the data, but the control services of it and the data services, but not the data itself. Now the transition challenges that one could imagine would be that not all network function virtualizations could be virtualized. So it means that the infrastructure that is jointly hosted by the two and programmable by the two must have interfaces for existing network elements and it's not possible in all the situations. Similarly, once we talk about deploying, how is it going to impact the operations and the business support systems because traditionally the OSS and BSS do not have the virtualization part as their inherent feature. So it means that altogether a new set of functionality and operations have to be orchestrated at the OSS and BSS. Similarly, the network operations which are mostly based on APIs, these have to be taken as applications which are running on certain platform. It means that the more traditional hardware which would run only a limited set of applications, this would require now software defined virtual network functions. So it means that we are now going to look at special network elements, special hardware which would be based on cloud technology which is programmable as the computer nodes for the virtualization. We need to incorporate a skilled resource operator staff because we need now much more computer intensive operations. And then last but not the least, the business practices now need to support the network function virtualization infrastructure as a wholesale service or offer between operators with the detailed requirements of the SLA, the service level agreement and other terms and conditions. The advantages are numerous but in short we can talk about the sharing and pooling of resources that could possibly take place which makes these localized infrastructures more location independent. Then full control of computing resources is now going to have more efficiency and flexibility. Likewise the service providers would now be able to respond to an increasingly large number of customers so the customer base could vary significantly and then the infrastructure is now not going to be depleted or exhausted. The reference is essentially the same that is the ATS report back in 2013.