In 2015, there were more than 60 million cellular machine-to-machine (M2M) connections in North America alone, and by 2020, that number will reach 175 million, according to a new study from the GSMA. Forecasts vary on the global internet of things (IoT) market, but several put it at more than a trillion dollars in the next few years. Scalar Market Research estimates that the IoT market will be worth $1.47 trillion this year, and reach nearly $3 trillion by 2022. The global telcos sit in the middle of this – they interconnect all of these devices (along with everything else we use in modern communications today) by hosting “edge” locations in every city and town worldwide. Because of this, telcos are perfectly positioned to support the IoT market by bringing intelligence to the edge in the form of mobile edge computing (MEC).
MEC is a highly distributed architecture based on the deployment of small computing units at these edge locations. As a member of the European Telecommunications Standards Institute (ETSI), Red Hat is helping define MEC use cases and architecture details. This is complementary to ETSI’s existing work in network functions virtualization (NFV) and 5G initiatives. As my colleague Marcos Garcia wrote back in February, MEC’s goal is to provide on-premises computing that can run independently from the core network, with proximity and location-aware services that capture and analyze key information from the edge. An MEC architecture is ideal for M2M and IoT scenarios. As networks are getting more distributed and virtualized, it’s possible to run workloads in distributed sites, which opens doors to do new and interesting things beyond traditional networking services.
How does an MEC approach differ from traditional centralized IoT computing approaches? One area that’s very important to consider is security. Recently, in October, there was trouble on the U.S. east coast when many IoT devices, communicating only via centralized cloud services, inadvertently hosted a botnet which launched a massive distributed denial-of-service (DDoS) attack, disrupting dozens of major websites. By contrast, with an MEC model, such an attack would be contained within the local site, since critical functions are hosted at the edge itself, and edge devices do not need direct access to the Internet. This makes it possible to systematically prevent such widespread problems from occurring in the first place. Further, this approach benefits users as well, as it lowers the response time of these services and therefore improves user experience.
Also, by using standard x86 architecture rather than custom hardware, along with a virtualized software stack, it is possible to decouple the software from the hardware. This speeds up time to deployment of new services, as only new software needs to be installed rather than new hardware. Telcos can also exploit economies of scale by consolidating many IoT applications onto the same industry-standard high-volume servers, switches, routers, and storage, transforming these environments into elastic, pooled resources that can scale up or down as needed.
The MEC architecture also introduces interesting new business models for telcos, allowing them to monetize their tremendous capital investment in areas well beyond network services. For instance, in every edge location worldwide, you can still find stacks of landline switching equipment that are hardly used. This valuable space – with redundant power and environmentally hardened – can be reused to build small regional data centers filled with servers, with “vertical virtual software stacks” installed on them. These OpenStack-based environments transform old landline central offices into next-generation central offices (NGCOs), that can host home security services, media delivery services, agricultural monitoring services, and anything else for which locality is important. With these services, telcos can deliver higher value to the user and create far higher revenue opportunities than pure network transport – services can be tied to the value they bring the user rather than the raw cost per bit.
Of course, building this requires a lot more than just OpenStack – middleware, analytics, rules engines, data integration and messaging services are also required. With the MEC approach apps can also be distributed as microservices – some components in the main data centers, some in the NGCO, and some in gateways closer to the actual IoT devices. This modern approach to app development opens up a world of possibilities.