How much latency is enough to support the most stringent real time application? What is the least interruption time to maintain the service integrity for most consistent application during transfer of end user from one point of connect to other point of connect, to access the serving network, or during the fault recovery from transient faults with in the network itself.
These could be some of the citation for the requirements for the highly available network. So what I was taking here as magical networks… and that is what I want to discuss as magical networks paradigm i.e. networks which are so highly available that the application is completely transparent of their internal behaviour for environmental changes like, mobility, fault & recovery.
We know that even currently available networks provide very high availability and the consistency of the service is maintained during hand off, connection failure & recovery, reselection of access points etc. There are quite a procedural executions to maintain such availability with fair transparency to the application.
The magical networks paradigm is to make such lengthy procedural execution to almost disappear or reduce them to the negligibility of computation.
And this is very much feasible if your latency of access is coming down and down and also with some of architectural evolution. The core of the network, where the context of the application requirement for the delivery of services is retained, is dissolving to totally flat architecture with all IP paradigm. There is no rigid or strict architectural requirements in terms of core network architecture as the NFV and cloud is blurring the boundaries of computational requirement of network specific functionalities.
That means networks are such that the end users are able to access service with minimum set up time due to less procedural computation and very low latency. The mobility is such that the procedural executions are negligible to provide the hands off but rather like a fresh connection from one point of connect to other point of connect so fast and consistent that the magic has occurred below the line completely transparent for the things above the line.
This magical paradigm could be aligned with Application defined networking which is about a complete separation of control and data plane and control plane is more defined from application plane. This also see the network access and core completely decoupled. This is rather the arrangement which can give a realization of the magical fidelity to applications.
5G is coming out with certain high expectation like very low latency perhaps of one millisecond only with radio advancement in radio access technology, convergence at access, network function virtualization at radio access network and for core network functions, Network slicing for effective network utilization and off-course software defined networking.
These are some of the advancement to felicitate a flat architecture so flat that only a ‘connect’ is required to change the point of access i.e. no hand offs or handovers. Also so highly available with fault detection and recovery like a 100 percent resilient.
Probably this will be the next optimization objectives for 5G initiatives.