The Network Functions Virtualization, a different kind of animal

If you are involved somehow in the telecom’s industry, you should know the buzz that Network Functions Virtualization (NFV), and its parent trend Software Defined Networks (SDN), are having in both vendors and operators. This in part because the NFV trend was originally encouraged by the main operators of the world themselves, with October 2012’s whitepaper (here), on an exercise where they seemed to explain what they wanted to see in their networks in the future. Which is in other words like commenting what type of technology they wanted to buy, or wanted to be sell on, for having that well-known list of NFV/SDN advantages in the future (i.e. reduced CAPEX/OPEX, simpler and standard networks, shared resources, faster services, etc.).

dilbert-virtualization-cartoon3

The excitement and kind of “shark behaviour” of all the involved can be seen by the fact that more than 100 companies or organizations are directly involved in the NFV standardization, via the ETSI (here), without mentioning those involved in the SDN standardization, via the ONF. Being realistic I believe these trends will not be so great in the short term, but most likely not so terrible in the long term neither, on a classical Amaras’ law fashion. The operators are simply pushing for securing the future of their business, and we know the current traditional networks will not be migrated or replaced to virtual networks in the short nor medium term, but the new ones being installed in the mid-term might be a different story depending on the NFV/SDN evolution pace.

Now what are the NFV and SDN trends after all? I remember an interview I especially liked with one of the Technical Experts for Networks Virtualization initiatives from Telefonica I+D, during the SDN World congress in Barcelona last June 2013. He made a simple explanation by saying, “Current traditional networks are like having a group of dogs and cats for pets, you must feed them and take care of them independently, each one has its own different needs, and sometimes you cannot have them together. What we (the operators) are looking to have with NFV are networks like a herd of horses or cows, where we can maintain them and work with them as a group, despite having faster horses or some dedicated to specific jobs, but sharing the same needs and therefore requiring around the same resources and skills for maintaining those”. This is an analogy that explains by itself with the fact the operators spend huge amounts of money in OPEX for the current networks. This without mentioning that any upgrade, integration, or evolution, requires huge CAPEX with its consequent increased OPEX during the nodes life cycles. Now you can imagine having a whole IMS network hosted in highly efficient and cost effective COTS hardware in a rack, with separate and automatically orchestrated virtual domains in software, dedicated for each function. Note I say function instead of node, since the idea is again simplifying the architectures, so if a SBC and a CSCF share a functionality it would only appear once in this new approach. That would represent multiple advantages compared with the traditional approach of multiple different nodes, with different hardware and software for each stage of the IMS flow, including: power consumption reduction, less signalling messages required, lower latencies, flexible and automatic scalability, and a lot less resources for maintaining the operation… among many others. Now imagine that in the core network, or in the Value Added Services (VAS), etc. So that is the operators’ panacea, and some vendors are already having some solutions thinking on that, starting with the nodes more easy to handle in the virtual world without affecting performance and reliability, like: policy & charging, firewalls, signalling routers, or some IMS network elements.

Therefore if we were to define the NFV and SDN trends in simple terms, the SDN would then be as splitting the control plane from the data plane in the networks, while NFV would be like splitting the software from the hardware; both could be implemented together, or independently, depending on the path followed and the specific needs. A third and final flavour is added to the mix with Cloud computing. Most of the industry experts believe the real gain of NFV/SDN applied to the carriers should be the same already proven in the enterprises’ data centres world, and that is having most of these architectures in the cloud. In the previous example of the IMS network, you can imagine that rack hosted by some cloud provider taking care of its capacity and maintenance on an outsourced model of SaaS, and opening a set of API’s with the operator’s services. If you are concerned about security in that approach you can even consider having that cloud hosted in your own data centre, having then a private cloud, etc.

NFV_vs_SDN

Even when some vendors are already claiming to have carrier grade NFV/SDN solutions for the operators, I think no one has yet offered a full and solid product for this year. According to a survey from Heavy Reading this year, 32% of the respondents think the SDN will be widely adopted by 2015, and 36% think by 2016 or later (here). I would say the vendors closer to have these solutions ready for the operators are Alcatel-Lucent with the CloudBand, and Cisco with the Cisco ONE. Some groups of small vendors are also working together to offer interesting things in the near future, like the CloudNFV group (here). And some specific services oriented vendors are just offering a virtual version of the same nodes previously owned, like the Policy and Charging vendors, including: F5, Amdocs, Openet, Tekelec, etc. A report from Informa Telecoms & Media, with Juniper Networks sponsorship, states: “…operators consider mobile SDN to be a critical technology for the future of networks: Ninety-three percent of operator respondents expect SDN to be implemented in mobile within five years, and half expect it to be implemented in the next one to two years” (here).

As the momentum for NFV/SDN is building rapidly, make sure you are prepared for the new and different kind of animal in the farm.

From M2M to IoT, or the buy cheap-buy lots evolution

Machine to machine type communications (M2M) have existed since almost 20 years ago, probably starting with Siemens innovations for that point in time around 1995 when the first machines were provided with communication capabilities for interacting with remote servers automatically. The M2M technology has widely evolved the machines communication capabilities, its services, and its applications during these years, getting to complex and sophisticated systems available today for a wide range of applications including: automated health systems, industry appliances, telemetry, transport communication, vending machines, e-commerce and retail, among many others. In addition, the networks carrying the information from these client devices or smart machines to the application servers have evolved in parallel, to the point where most of the M2M devices today uses the cellular networks, most of these on 2G and some of these in 3G and more recently 4G/LTE. Thanks to this it is possible to efficiently have the vending machines filled from the storages in time, the medical reports sent in real time to the hospitals, or the transport trucks float located around the world at all times, for giving a few obvious examples.

Today, as the telecom evolution continues, we have a major trend in the industry known as the Internet of Things (IoT), or Internet of Everything (IoE) depending on who tells the story. The IoT intends to propose an important evolution from the classical M2M communication to some kind of utopia where every object in the world can and will be connected in the future. Current innovative IoT products allows controlling the lights in the house, or the doors, or even the kitchen appliances remotely, with a single click or touch of our tablets and smartphones.

The applications available for encouraging the IoT today range from a wide offer of home appliances connected with the cloud (e.g. from toasters to TV’s, and from the house’s light systems to the AC’s), to devices designed to be wear-on during our day-to-day activity (e.g. connected shoes or t-shirts). This of course includes famous and innovative devices like the smart watches from big companies investing on these e.g. Apple or Sony, or the optic devices like the Google Glass. What all of these devices have in common however, is the fact that the previously industrial purposed M2M technology is being extended to the common user for personal or business usage. Which is mainly happening thanks to the low cost and low power sensors and actuators being available just now in the market, as the electronics for these is getting more popular, among other factors. As such, today we can find in example movement sensors with cellular connectivity on 3G networks interacting with Android applications from USD$10 or so, when a few years back this implied hundreds or thousands of dollars, if at all available for retail.

Cisco counter

The projections in the number of devices and objects potentially connected for IoT, and being shared by the incumbent analysts and vendors, are simply dazzling. Cisco Systems in example estimates we had 8.7 billion connected objects in the world for 2012, and we will have around 14.4 billion and 18.2 billion for 2014 and 2015 respectively. There is even an online counter or estimator of connected objects in real time available in this link. We should keep in mind Amara’s law however where especially in innovations “we tend to overestimate the effect of a technology in the short run and underestimate the effect in the long run”; but even after a reality check, the number of connected objects in the future should still be amazingly high.

These figures justifies the amount of people wanting to participate in the evolution, and this also includes the network operators. Traditionally the M2M technology had historically represented very low ARPU for the operators. In example some reports from Vodafone indicates the average revenue per month per account for M2M during 2012 was around USD$5, with projections to lower this value during the coming months in 2013; while the cost of delivering services in 2G/3G is much higher than this. Therefore, the M2M business is only profitable for an operator when having big amounts of accounts and devices, and as the nature of the M2M traffic is producing very low throughput usage in the network with lots of transactions this is also technically feasible to handle with current networks. The situation for the IoT is therefore only desirable from a massive implementation point of view, just then allowing a credible business case for the operator.

Cisco graph

This year we are starting to see the operators decommissioning their old 2G networks (where most of the current M2M devices are operating today) or planning to do so in the coming years, for allowing this spectrum being reused on future LTE and LTE-advanced deployment. Most likely spectrum is and will be the most desirable good for the operators as technologies evolve these years. This forces the M2M communications to be deployed on 3G or 4G/LTE networks, and triggers issues like seen in the USA where point of sales only operating on 2G networks were suddenly left out of business with no alternative option. The business case for M2M on 3G and 4G/LTE networks however is still not clear, as even when the cost for the operator to deliver services on these technologies is cheaper, not all the M2M devices support these technologies yet, and not all the operators have recovered the capital investments on these access technologies neither. Considering all of this in my opinion, if we are planning to have credible and massive implementations of IoT in the future, we will have to see these adapted to the most recent communication technologies, for making sure these survive… and the operators keep their profitability. In other words if a smart toaster gives little gain to the operator, he would only support to communicate it if you have millions of smart toasters connected in a signed contract… buy cheap, buy lots. Keep the profitability levels for having everybody happy in the evolution towards an intelligent objects world.