(The Russian version is here)
I am often asked these questions, so I decided to write a summary of the answers.
So let’s start in order.
NGN (Next Generation Network) is more a marketing term than a technical one. In fact, what does the “Next” mean? When will this “next generation” come true? The technical basis of the NGN concept is the so-called SoftSwitch, which is often understood as «software switch».
In fact, it is not a switch at all, since it does not switch packets, but only controls the transfer of packets in the underlying IP network. And the word Soft means not “software”, but really soft, or flexible. Only control protocols’ packets are passing through the SoftSwitch.
It was planned that the SoftSwitch will be able to very flexibly manage media streams over the network, assign service policies, and manage the carrier network very quickly and in centralized manner.
In reality this concept has been facing many problems. First of all, ALL carrier networks must be SoftSwitch based, otherwise, they all will work inefficiently.
However, NGN was the first attempt to abstract the Control plane of the network from the Data level. This concept was further developed in IMS, and then emphazised in SDN / NFV.
IMS (IP Multimmedia Subsystem) is the next step of functional separation, or abstraction, when network functions are separated from equipment.
Traditional network (including NGN), were based on the concept of “hardware — function”. In IMS, all of the various network functions (CSCF, PCRF, MGSF, AGCF, etc.) are not necessarily located in the same hardware rack. One functional block, for example, CSCF could be located in different pieces of hardware and even in different locations.
The NGN’s SoftSwitch was a germ of the MGCF (Media Gateway Control Function) and AGCF (Access Gateway Control Fucntion) function blocks in the IMS. As a hardware basis of all these functions, specialized powerful servers began to be used in IMS.
The next logical question arises: is it worth to design a specialized servers for the Telecom industry, or maybe it’s better to use standard IT servers? Maybe, they are not so powerful, but they are standard ones, and thus cheap.
At first, this approach was somewhat inconvenient, since Telecom networks demanded a higher level of reliability and availability than corporate IT networks. But enhancements of reliability and throughput permeameters of standard commercial servers for IT made it possible to solve this problem (so called COTS — Commercial Off The Shelf).
And thus, moving to full virtualization of network functions has become possible.
NFV (Network Fuction Virtualization). I dare to say that NFV is the implementation of the IMS concept on COTS. Concept, but not the technology. In NFV, you will not see IMS functional blocks — all these CSCFs, PCRFs, etc. NFV technology was developed from scratch. But conceptually this is a new incarnation of IMS.
SDN (Software Defined Network) is somewhat separate, and it even appeared before NFV. If we talk about abstraction, which runs the «red thread» through the concept of NGN — IMS — SDN / NFV, then we can say:
- NFV is the separation (abstraction) of functions from hardware.
- SDN is the separation (abstraction) of control from execution (separation of Control Plane and Data Plane, if to use IP routing terminology).
As for Triple Play, this is an access technology, when a user receives three services simultaneously: voice, Internet access and video services. That is, the integration of three networks, three media sources in one access node. But if we talk about mobile communication, then we see that a user receives everything at once and through the same 3G/4G network: the voice, the mobile Internet and the video (also through the mobile Internet). And he also starts to get more voice through the mobile Internet (Skype, Whatsapp, Viber, etc). VoLTE – is the voice service over IP-network. Many telecom carriers are already disconnecting 2G (GSM) networks as a separate network for voice. This means that only smartphones can be used in their network. Therefore, the concept of Triple Play, which was very popular 10-15 years ago, has now become obsolete.
Finally, about abstraction. Why is this necessary? Why distribute functionality at different levels?
- First, in order to quickly identify the problem in the system, and correctly determine the area of responsibility.
- Second, to quickly introduce new functions.
Third, to simplify the protocol architecture, understanding and development capabilities of technical personnel.