15 May 2018
Edge computing is where the processing of data is performed at the network edge (i.e. closer to sensors or devices collecting data) rather than data being transferred back to a centralised datacentre for processing and analysis.
An edge device can be any device or ‘thing’ that can be connected to a network – it’s what we mean when we talk about the Internet of Things. Essentially, a kettle can be connected to a network through sensors that relay information, but in order to quickly process data about the kettle or water temperature, that information might need to be analysed closer to the device; e.g. to tell the device to turn off if the water has reached a certain temperature. This is an example of edge computing.
Mobile devices such as our phones or tablets are enabling more processing to be done closer to sensors. For example, if you’re wearing a FitBit or activity tracker then you can sync it with your phone to review your statistics and data locally. The processing of the information is done on your phone rather than on the ‘dumb’ sensor (the FitBit).
Think about if you were in a self-driving car that is connected to the internet and using GPS info and analytics to drive you around, stop at red lights and avoid hazards. If you suddenly reach a stop sign, the sensors in the car need to quickly take in the information about the stop sign, process the data and make a decision about what action to take next. You can’t afford to wait a few seconds for data to be sent across a network to a datacentre where it can be analysed then sent back to the car if you are heading for a stop sign. The datacentre could be located in a different country or continent, slowing down the time to send and receive the data across the network. Processing in scenarios such as this needs to occur closer to the device – i.e. the edge, so edge computing is about moving server, storage and networking resources closer to the edge where data is being collected so that decisions can be made quicker and data from sensors can be quickly processed.
As we develop more sophisticated edge devices and sensors, we will need to process much more data, which will require more bandwidth and network speed to handle the increase in sensors and data production.
Moving processing to the edge and closer to devices means that network latency is reduced and information doesn’t need to be relayed over different networks to provide an answer. Coming back to the self-driving car example, without edge computing, data would be recorded by a car’s sensors, transmitted to a central datacentre to be processed, then sent back to the car. Edge computing is about delivering that processing activity closer to the car than a centralised datacentre, speeding up the process and enabling the sensors within the car to process and analyse information rapidly.
This is important as we rely more heavily on artificial intelligence technologies that need to analyse data and come up with decisions or answers very quickly. And, as more of our ‘things’ are connected, we will rely more heavily on connected devices to provide important data and take autonomous action. Sensors are now being built into aeroplanes, manufacturing assembly lines, machinery and home devices.
This doesn’t mean that datacentres are not needed, it may be that data is sent less regularly to centralised datacentres, or datacentre resources are delivered across a grid-like architecture, with lots of mini-datacentres in locations that are closer to edge devices.
When talking about edge computing, you might hear the term ‘fog computing’. Cloud computing was the centralisation of resources delivered from fluid pools of computing resource in a central datacentre. Fog computing is extending traditional cloud computing to the edge of the network. Rather than devices just connecting back to one central datacentre, fog computing refers to the complex web of devices that are all interconnected and connecting to compute resource close to the edge, which is then connected back to datacentres.
In fog computing, data is transferred across the Local Area Network (LAN) and processed at ‘intermediate nodes’, perhaps a small server located close to the sensors. In edge computing, more of the processing is done at the actual device location; i.e. a mobile phone or on the sensor itself.
Some believe that edge computing is more secure than cloud computing because data is not having to travel long distances over networks and because data is not being centralised in one datacentre so the risk is spread. However, there is also the issue that edge devices, such as sensors, mobiles and devices such as connected fridges or cars can be more easily compromised compared to enterprise, secure datacentre resources. There have been worries in the industry about hackers gaining access to self-driving cars or connected home solutions, in a bid to control the environment or extract sensitive customer information.
Find out more about how you can deploy your own edge computing solution with our whitepaper on ‘Hybrid Cloud grows up: Introducing Azure Stack’: https://www.pulsant.com/knowledge-hub/whitepaper/hybrid-cloud-grows-introducing-azure-stack/
Explore our world using our interactive map