Edge Computing

Everything you need to know about edge computing and our micro data centre solution.

what is edge computing?

Edge computing is a distributed computing model in which computation is largely or completely performed on smart devices or edge devices as opposed to primarily taking place in a centralised cloud environment. The "edge" refers to the geographic distribution of computing nodes in the network as Internet of Things devices. These are at the "edge" of an enterprise, metropolitan or other network.

This enables analytics and data gathering to occur at the source of the data instead of relying on the cloud at one of a dozen data centers to do all the work. It doesn’t mean the cloud will disappear. It means the cloud is coming to you.

why are we getting closer to the edge?

The motivation is to provide server resources, data analysis and artificial intelligence (“ambient intelligence") closer to data collection sources such as smart sensors and actuators. Edge computing is seen as important in the realisation of physical computing, smart cities, ubiquitous computing and the Internet of Things.

Devices like Amazon Echo, Google Chromecast, and the Apple TV are powered by content and intelligence that’s in the cloud. As centralised as this all sounds, the truly amazing thing about cloud computing is that a large percentage of all companies in the world now rely on the infrastructure, hosting, machine learning, and compute power of a very select few cloud providers: Amazon, Microsoft, Google, and IBM.

The advent of edge computing has been pushed by the realisation by these companies that there isn’t much growth left in the cloud space. Almost everything that can be centralised has been centralised. Most of the new opportunities for the cloud lie at the “edge.” And as products and services rely on speed and higher bandwidth for increasingly larger data transfers, a geographical solution as well as increased compute power tied to smart devices mean that edge computing is an increasingly important part of a digital future

Latency

One great driver for edge computing is the speed of light. If a Computer A needs to ask Computer B, half a globe away, before it can do anything, the user of Computer A perceives this delay as latency. The brief moments after you click a link before your web browser starts to actually show anything is in large part due to the speed of light.

Voice assistants typically need to resolve your requests in the cloud, and the roundtrip time can be very noticeable. Your Echo assistant (Alexa) has to process your speech, send a compressed representation of it to the cloud, the cloud has to uncompress that representation and process it — which might involve pinging another API somewhere, maybe to figure out the weather, and adding more speed of light-bound delay — and then the cloud sends your Echo the answer, and finally you can learn that today you should expect a high of 85 and a low of 42, so definitely give up on dressing appropriately for the weather.

With Edge Computing, the latency is massively reduced and there are rumours that Amazon are producing their own AI chips for the Echo, meaning more processing can be done on the device, relying less on the cloud and the requirements for high density data equipment at centralised locations.

PRIVACY AND SECURITY

Simply by doing encryption and storing biometric information on their products, Apple offloads many security concerns from the centralised cloud to its users’ devices. Other companies will follow suit as devices become smarter and Internet of Things products around smart homes demand it. The idea is that smart home devices should be very difficult to hack, with software updated automatically via Edge servers and security managed centrally.

Future applications requiring edge technology use

Self-driving cars

• Autonomous vehicles are on the horizon and driving into view faster than most would realise. Due to latency, privacy, and bandwidth, you can’t feed all the numerous sensors of a self-driving car up to the cloud and wait for a response. Your trip can’t survive that kind of latency, and even if it could, the cellular network is too inconsistent to rely on it for this kind of work. But cars also represent a full shift away from user responsibility for the software they run on their devices. A self-driving car almost has to be managed centrally. It needs to get updates from the manufacturer automatically, it needs to send processed data back to the cloud to improve the algorithm.

5G

• As with its predecessor 4G, 5G was developed in a direct response to the growing number of mobile devices seeking an internet connection. However the rise of 5G coincides with the explosion of connected devices and systems associated with the Internet of Things. Processing these high volumes of data, at a faster speed, will require new antennas, new devices, and new applications for wireless data. Optimised network speed from the adoption of 5G will result in even higher user expectations when it comes to low latency and always-on connectivity which Edge computing will be able to provide.

Industry and enterprise

• Edge servers will be added on site next to warehouses, distribution centres and factories, as opposed to corporate headquarters. The main goal could be to ingest, aggregate, and send data from multiple operational data points (think vehicles, equipment, products in storage etc) to proactively monitor company assets. This in turn helps maintain optimal performance of company assets and products at a lower cost and higher safety.  For some industries, the need for rapid processing extends beyond customer satisfaction. In the case of medical wearables, for instance, a lag in data processing could create a life or death emergency. This type of computing will require closer data centers as well. In fact, the International Data Corporation (IDC) predicts that internet providers will add micro-data centres into or nearby 5G towers as a means of creating this access.

New Ideas - new equipment

This age of digital innovation requires new, more powerful types of server hardware, such as GPUs, ASICS and FPGAs.  This new equipment has been designed to process data more efficiently and quicker, while also taking up far less space than traditional CPU equipment. These changes in size and power mean that we have to draw on a different set of engineering principles to achieve optimum efficiency while maintaining reliable environmental conditions. We are dealing with higher densities of equipment, which means we need much more air!

• Example: kW/sqm data centre (Traditional v HPC)

While conventional, CPU based data centres will typically have 2kW of technical load p/sqm, EcoCooling is now working with data centres using the latest higher density HPC equipment installing up to 16kW of technical load sqm.  This means that you will need to supply 8 times the amount of air to the same space in order to keep the equipment cool.

This requires a different approach to airflow management as the volumes of air through the same space are so much higher than what we are used to conventionally. In our latest HPC installations we are seeing air velocities of up to 10m/s through cold aisles, that’s over 20mph!

deploying fresh air technology

New types of HPC equipment are increasing their inlet temperature guidelines significantly.  Whereas 5 years ago data centre operators were aiming to achieve between 15-21C inlet temperatures, the new equipment will allow temperatures up to 40C.  One result of this increase is that we can look at using fresh air systems in a larger majority of locations, where external temperatures do not increase above 40C (basically all of Europe). 

In hotter climates, evaporative cooling can be used to lower the temperature, Standard EcoCooling WetBox units can be simply fitted downstream of the supply ventilation if required.

our HPC DATA CONTAINER as the obvious Edge solution

The EcoCooling HPC Container is a flexible, transportable housing solution for the latest in HPC (high performance computing) equipment. The 20ft tunnel shipping container has been designed with rapid deployment strategies in mind and is easily stacked to enable quick scale ups. The container solution allows the operator to scale up as and when is required.  The power draw is designed to fit with standard 250kW module transformers to ensure no resource is wasted.  It provides a low total cost of ownership solution for the low latency infrastructure critical to Edge computing.

• It can be deployed at the edge of sites as a micro data centre dealing with low latency tasks for industry applications and on the outskirts of smart cities. It is the perfect solution for 5G sites and can be connected to antenna/aerial equipment with remote monitoring.

• The plug-and-play design reduces the amount of skilled labour required for installation while the simple cooling system significantly reduces maintenance requirements.  These factors, coupled with the low cost means that 1MW of HPC equipment can be deployed for under £300,000.

• The patented rack design future proofs the container for changes in server design.  Racking can be redesigned and deployed cost effectively without changing the basic layout of the container.  Additionally, the integrity of the container is not affected by the modifications for the IT equipment, meaning it retains its ISO shipping credentials and can be redeployed to alterative locations depending on the business case at the time.

• While containerised systems are the obvious answer to the rapidly increasing demand for edge and HPC processing power, it should be recognised that they could be considered an eyesore if not installed subtly. This can be important on sites at the edge of cities. EcoCooling offer a number of noise reduction and camouflaging options for equipment which can be specified by the client. Containers can be specified to any colour from the RAL colour system for absolute colour matching and logos can be applied if desired. Perfect for Telecoms applications; providing advertising for mobile communications providers.

What makes our container solution special?

Our extreme approach to data centre design and engineering has been refined over the last 10 years working with innovative clients such as BT, Hydro66, Talk Talk and ETIX Blockchain. Developing bespoke products for large UK telco projects increased our manufacturing capabilities by allowing development of our own products and control systems, resulting in what we now call the CloudCooler range.

Developing solutions which are optimised for clients operating across the cryptocurrency space brought a whole new set of innovations based around ASIC, GPU, FPGA and HPC equipment.  The ECV CloudCooler was reflects the requirements of these new sever technologies which do not require supplementary cooling in such cool climates. Working towards the ‘lowest total cost of ownership’ options with data centre operators soon led us to the development of the Data Container, with many of our end users deployment strategies depending on flexible data processing which can be physically moved to different countries/regions, where it is lower cost, closer or more secure.

Total Cost of Ownership

When EcoCooling looks at lowest total cost of ownership we look to factor in a number of different figures:

·      CapEx + Opex (Installation maintenance costs + ongoing running costs)

·      Depreciation of equipment (Modern equipment is usually upgraded within 2 years)

·      Skilled labour requirement for upkeep

·      Reusability of the container

World class performance is often considered to be less 0.10USD per kW/hr total cost of ownership. This includes all facilities, staffing, energy, equipment depreciation etc. This simply cannot be achieved with conventional data centres, EcoCooling’s solutions can be deployed for a facility/capex cost of under 0.02USD/hr, this a major contribution to achieving world class performance.

Deployment - Low cost results in low complexity

Emerging technologies could result in a stratospheric growth in the need for processing. EcoCooling’s low cost and simple solutions are designed to facilitate the rapid deployment of infrastructure to support this new age of data processing.

For more information on our Edge Solutions or any other products:

email: sales@ecocooling.co.uk

call: +44 (0)1284 910 586

Not sure about your requirements?  Our team of application engineers can provide analysis for your location to advise on the right equipment for your supply temperature requirements.  Contact us at sales@ecocooling.co.uk

sources: ge.com, theverge.com