Share this post:
In this article I look at the emerging area of Edge Computing: how the compute power of decentralised Edge devices will be open and not only run apps, store content and data and host virtual networks, but also enable new and novel experiences by being autonomous and able to collaborate with each other.
The ‘Edge’ is the technology space at the edge of the internet, in the hands of users and embedded in physical infrastructure such as buildings, cars and cell towers.
Edge Computing promises to disrupt a number of markets including Products, Electronics, Communications and Cloud, upsetting some and creating opportunity for others. ‘The Economist’ sees that capitalism itself will evolve as a result.
A year or so ago I was talking to some bright young techno-hipsters at a start-up.
Like a lot of start-ups they were passionate and enthusiastic about their plans and for these hipsters the plan to get rich quick was to deploy microprocessors with sensors through the home environment and connect them to everyday appliances, to each other and to the internet and so improve the way we interact with those appliances.
Not a new idea but part of their novelty was to make their microprocessors open to anyone, so that other developers could come in and innovate by deploying further apps and logic. The key words here are “open” – anyone can participate, and “innovate” – create something new and useful.
For the start-up, there was no financial downside to making their device open. They reasoned that while they only needed 16 bit grunt, 64 bit processors cost the same as 16 bit processors so why not spend the same money and make the real estate available for others to come in and use?
The hipsters said their only motivation was curiosity: how would others would use their open devices? I suspect however they were also motivated by the ‘Network Effect’ which makes some technologies (such as telephones and the internet) more useful and more attractive as more people use them: the more developers deploying innovations onto the start-up’s device, the more attractive the start-up’s device will be to the market.
This is Edge Computing. The microprocessor sits close to the edge of the network (hence the name ‘Edge’, get it?) and becomes the open compute infrastructure on which other developers deploy apps or other workload. In this example the start-up is the manufacturer of the Edge Device and the ‘other developers’ are Third Party developers that deploy onto someone else’s Edge Device or infrastructure.
Edge Computing devices are different from earlier classes of devices in that they are built on powerful commodity IT infrastructure, expose device capabilities such as storage and sensors via application programming interfaces, are connected (sometimes intermittently) to the internet, can host third party applications and run autonomously in a decentralised mode of operation.
With computer components becoming smaller, more powerful and commoditised, the internet becoming increasingly ubiquitous, and an increasing amount of capital and skilled people investing in tech research and development, there’s an increasing range of ‘Edge Devices’ premised on a chassis of open, powerful, general purpose IT infrastructure in the market. These include:
– Consumer Electronics devices such as smartphones, tablets, televisions, set top boxes and wearables such as fitness devices. By contrast, feature phones such as 2G handsets from the last decade were not open and beyond limited microjava support were programmable hence were terminals and not Edge Devices.
– Network Equipment, being customer premise equipment (eg home gateways) and telco service provider kit such as radio access equipment for cellular networks. Built in the past as physical appliances comprising hardware and software from the one vendor, Network Equipment is increasingly being deployed as virtual network functions or Software Defined Networks running on commodity IT infrastructure.
– Industrial or OEM equipment such as embedded systems in buildings and vehicle telematics systems
- Edge Computing to support Consumer Electronics Innovation
The best known example of Edge Devices are smartphones and tablets, popular and useful (and more profitable for the vendor) because they are open and enable others to innovate by building apps that we (often but not always) use and value.
These consumer electronics devices typically run iOS or Android and are supported by vendor-run app stores that manage the service lifecycle by approving code and enabling end users to shop, purchase, charge, distribute, provision and upgrade the code.
The smartphone is an innovation in itself, but an equally-important adjacent innovation is the app store that commercially binds third party innovators and users over the web. With the app store and developer ecosystem, the Network Effect takes hold.
- Edge to support and extend Internet of Things
In addition to consumer electronics, Edge is becoming infrastructure for the Internet of Things. Many of us think of IoT only in the traditional model where IoT devices are single-purpose devices comprising sensors, embedded into physical environment, connected to microprocessors that relay data over the internet back to the Cloud where the sensor data is processed and stored.
This is a narrow view of IoT and what Edge adds is the ability for the IoT to be multipurpose as per the example of the start-up’s home automation system. The open compute infrastructure supports multiple applications that can run autonomously and in a disconnected mode of operation and process data very close to the point where the data is generated.
With the home heater providing a programmable interface, one application can provide a remote control app. A second application can monitor the temperature and keep it in a range. A third app can detect the car arriving in the driveway and turn the heater up. With Edge different third party apps run on the same infrastructure and can interact to create new user experiences.
I suspect some reading this section will see IoT and Edge as one and the same, so just to be clear, while you can deploy IoT in a closed model, Edge is an open collaborative platform for IoT and other classes of service.
- Edge, Software-Defined Networks and Media Servers
Edge can of course also run virtual network and media functions. Where once the network function was deployed as a closed physical appliance, it can now be deployed (and far more easily upgraded) as software on IT. For example, a virtual firewall can be remotely deployed to a home gateway to improve internet security, and an improved virtual firewall can be deployed as a replacement as performance or security issues become known and are addressed.
A key benefit is that product upgrades and improvements are done without involvement of on-site support or end user intervention and certainly without the need for a replacement part to be shipped.
Here’s an example of the old-world way of doing things. After a Denial of Service attack which saw my plan’s monthly download limit consumed in a few hours, my Internet Service Provider sent me a replacement home gateway at no cost to me, as they were not able to remotely upgrade the firmware in my only-recently-acquired and possibly-compromised gateway.
Within the telco network there is already a move towards Software Defined Networks and Network Function Virtualisation. From what I can see there is no clear trend yet amongst the traditional Network Equipment Providers towards their new IT-based, software-defined Network Equipment being open to third party developers. The argument is that only the focus of a single vendor can deliver the latency and availability requirements required by Telcos.
A market approach of a ‘virtualised but closed’ single-vendor model is likely to be flawed. The internet is now larger than the aggregate telco networks in the world and multiple parties contributed to the development of the internet. Rather than harness the internet and its community, traditional Network Equipment Providers are avoiding being open to the internet community, with the result that traditional Network Equipment Providers are at great risk of being displaced from the market by innovators from the internet community with a strategy and track record of open development and execution.
To be successful, Network Cloud needs to be an extension of mainstream internet and cloud and not a separate market that fails to harness ongoing internet and cloud developments.
Moving on to media, Edge is also a great platform for media and content cacheing. Tired of buffering delays when you rewind your Over The Top video streaming service? Cache the content on the set of Edge devices in your home. Want to provide the best possible quality sports or movie streaming service? Cache the high quality content at cellular network base stations on someone else’s hardware.
Opening up Edge compute capacity as infrastructure for innovation looks a good idea, especially when you consider the predication that in a few short years the aggregate computing and storage capacity of smartphones will surpass all worldwide servers.
The implication is Edge itself could become an alternate hosting environment to today’s data centre-based Clouds. In the future if you want to run your workload on cheap available compute capacity, you could consider Edge in addition to or as an alternative to Cloud.
Deploying the innovation at the Edge rather than the Cloud also has performance advantages:
- improved latency – the application, content and data is closer to the field where the environmental data is generated and where an immediate response is needed
- improved network capacity and reduced network cost – why should the data travel all the way to the Cloud to be processed? Someone pays for this bandwidth. As an example, analysing security video at the Edge rather than back in the Cloud removes the need for tens of thousands of bulky video streams to traverse telco access and backhaul networks.
The good news for many telcos is that they already have physical real estate such as exchanges and base stations spread throughout the community; this is valuable real estate which provides the telcos with a natural advantage as they already have property into which they could deploy and monetise as an ‘Edge Cloud’. This real estate is already serviced by a tech-savvy field workforce who could readily put the Edge infrastructure in place.
Telcos also have significant fulfillment, assurance and billing experience and capability, delivering communications services on third party network equipment and more recently IT services on third party cloud. Extending this model to also fulfil, assure and charge for services on Edge is a natural extension of the telco business model.
It’s also likely we’ll see the emergence of an Uber-like infrastructure broker who will help connect developers with available infrastructure, and facilitate the contracting and realtime movement of workload into and between any of a number of cloud and Edge platforms.
- The Decentralised and Collaborative Edge
Samsung and IBM built the ADEPT proof of concept showing how decentralised devices can interact and collaborate (see the youtube video here). Using blockchain as a means to publish capabilities, negotiate and partner, common household appliances become autonomous Edge devices that interact with each other.
With technology increasingly being embedded in everyday appliances and products and in the physical infrastructure such as buildings, and the embedded software able to be improved and automatically upgraded, we see a change in the economic mode of production.
Citing the example of Tesla, who resolved a problem with hill starts in their electronic vehicles by automatically provisioning an over-the-air software fix to the cars (ie upgrading the car’s capabilities without bringing it in for a service), The Economist points out, “The old form of capitalism based on built-in obsolescence is giving way to a new one in which products get better after they are bought. This robs firms of the ability to make a quick profit by selling new models, but may bind them much closer to their customers.”
Like a lot of good ideas, it doesn’t take long for someone like me, conditioned by years of addressing clients’ ofttimes (but not always) reasonable concerns, to start to raise some questions about open, programmable Edge Devices. Questions like:
- How does the third party developer know there is compute resource available to them at the Edge?
- How do they find out what performance characteristics exist? ie will the available Edge Computing infrastructure meet their needs for processor, storage, connectivity and sensors?
- How do they negotiate a price (or, if this is free, how does providing free resource for others make the Edge Device provider any money?)
- How do the developers deploy and upgrade their code? ie how is the service provisioned and assured in an environment here there is no end user clicking on buttons on the device?
- How does the developer ensure their code can be readily ported to a wider angle of Edge Devices? (There’s an increasing number of heterogeneous Edge Devices out there and no developer wants to (or should have to) commit to one type of infrastructure or spend time and money porting to different types of infrastructure and maintaining different code bases).
- How is data managed and privacy ensured?
These are for the most part Service Lifecycle questions (or Provisioning and Assurance in telco-speak) which, for smartphones are largely addressed by the vendor-run app stores. App stores review the code and provide shopping, charging distribution and upgrade functions.
If you’re not using an app store because you’re building a non-consumer application (such as a zero-touch secure home gateway or industrial controller) or you’re an Edge Device developer standardising on Linux as open source operating system, then you need to have an approach to managing the Service Lifecycle, ideally without having to replicate the significant investment needed to design, build and run the equivalent of an app store.
To quote The Economist again, “the biggest winners of all may be those that control a ‘platform’, a layer of software that combines different kinds of devices, data and services, on top of which other firms can build their own offerings.”
This sentiment is echoed by Streaming Media who predict the winner of the long running war for control of the Living Room will be won by a device manufacturer who also has a platform.
Despite the fact industry understands the importance of the ‘platform’, in my view very few people active in industry seem to appreciate the difficulty of establishing a platform or what a platform actually does. There is a lot of excitement of functional use cases where the toaster sends a fax to the smoke detector to signal toast is cooking (and there is no need to call the fire brigade) but little focus on the less glamorous but foundational Service Lifecycle use cases such as ‘register a device’ and ‘apply a patch’.
Beyond Service Lifecycle management, lack of security is the greatest impediment to rapid growth of Edge Computing.
How can a developer trust unknown compute infrastructure, and how can an infrastructure owner trust an unknown developer?
How does the developer know their code or the data is not being spied on, and how does the Edge Device owner know their device is not being used to illegally share pirated movies?
We trust credit cards because the risk is underwritten by the bank or card provider, and in trusting credit cards an entire new payment instrument becomes available to merchants and customers in a large market. We need an equivalent trust model in the market for the industries that will harness Edge.
Some colourful real life examples of security breaches include:
- Thieves in South Africa stealing SIM cards from traffic lights and using the stolen cards to make thousands of dollars of calls
- A hacker in Texas changing a traffic sign to read “CAUTION! ZOMBIES AHEAD!!!”
- Water Pumps destroyed by SCADA hackers in New Jersey in 2011, one of “the first known (attacks) to have damaged one of the systems that supply Americans with water, electricity and other essentials of modern life”
- The Stuxnet attack on Iranian nuclear production circa 2010
- An Australian found guilty in 2001 of hacking “computerized waste management system and caused millions of litres of raw sewage to spill“
It is encouraging that existing security models for mainstream IT also apply to the Edge, and new capabilities required to secure the Edge are available and understood. As a simple example, a compiled language is more difficult to reverse engineer than an interpreted one, and going one step further to use a ‘crypto-compiler’ from a company such as strong.codes makes the compiled code extremely difficult to crack.
The security model differs based on whether you are a device manufacturer, app developer or operator of an Edge platform. It also differs based on the nature of the application. Remote control of the home garden hose does not need the same security engineering approach as the national electricity grid.
Importantly, security needs to be a part of the design process, not an after-thought. Like personal hygiene, security is appreciated and noticed more when it is not there.
There is real risk that, without a cross-industry approach to Service Lifecycle and security, we will see the marketed fractured or slow to grow, with many organisations spending money on duplicating background service management and security capabilities rather than creating innovative new products and services that may actually help someone.
While the risk is real, the pragmatic technologists and tech investors are encouraged both by a number of promising technologies that point the way forward, as well as the internet itself as a precedent for multi-party collaboration. Examples of promising technologies likely to be relevant to Edge include:
- Containers provide ease of application porting and deployment in the cloud and data centre world and there is increasing chatter on the web about containers for low power and low end microprocessors. Containers will help the developer migrate their code from one platform or Edge device to another, and they will make it easier to apply patches.
- Known as the technology that underpins Bitcoin, blockchain is a decentralised database where each node is a peer of the other. Smart Contract negotiation and recording of activities are blockchain capabilities likely to be useful with the decentralised Edge.
- Bitorrent as a distributed file sharing system to send out updates.
- The device at the Edge can plug into the smarts in the cloud. Cognitoys are early into this field, “CogniToys are the next generation of internet connected smart toys that learn and grow with your child.”
So when will Edge Computing become open and decentralised, as opposed to being a set of closed systems that do not interact? Walter Issacson’s wonderful book ‘The Innovators’ gives a number of examples of why technologies succeed and cross the threshold into widespread adoption, while others do not make it, a large part of success being both collaboration and timing.
We’re seeing people collaborate on Edge in the market. Getting back to the earlier example of the start-up techno hipsters with the home automation system, they were open to working with others to iteratively improve the platform so they ticked the collaboration box.
Timing relates to market-readiness: are other factors that will contribute to success also present? Like the start-up, I’m optimistic the timing is right for Edge Computing as a new frontier.
I’m listing here some of the material that helped shaped my thinking in this space:
If you are interested in seeing more of IBM Telecommunication industry solutions, visit ibm.com/communications .