Edge Computing 4US
- Edge Computing 4US Str
- Edge Computing Keys, Trends
- Edge Computing 101
- Trend: Pervasive IoT Micro-Controllers with Internet Connections
- Why: Edge Solves Latency Critical Problem
- Why: Bandwidth Consumption reduced by Edge Computing
- Why: Offline Apps enhanced by Edge
- Why: AI Edge Apps - privacy and efficiency
- What is special about Edge - not just a buzzword!
- Edge Computing Architecture and Technical Review
- Security in IOT and Edge - enabler
- Data Privacy and PII in IOT and Edge - transmission and storage
- Admin, DevOps is a weak link - Edge Computing must be easy to Manage
- Edge Device IoT security with Certified Secure Elements
- Evergreen updated versions of Firmware - upto-date security
- Certified Secure and Cheap MCUs Edge Computing ~Azure Sphere
- Shift away from User Settings, Awareness of Smarts hidden and Control of their devices
- 2020 Device/Mobile Apps, Limited AI processing on Chips
- SVR: Lambda Functions IOT and Edge - enabler for Smarter Edge
- Evolution of Edge Computing Architecture
- 1940s? Relays, PLCs
- 1960s Centralized - Mainframe, Minicomputers
- 1970s Minicomputers - Industrial Cell-Controls, 16 bit, 32 bit
- 1981+ PC Era => Thick PC Client-Server
- 1995+ Internet era
- 2007+ Mobile little Apps => MDM/MEAP => AppEconomy with Central Processing
- 2012+ SaaS Cloud: Dropbox, Gmail, Office 365, Collab - Slack
- 2017 PA Centralized With Lambda Processing ~SOA => Higher Latency
- 5G impact
- Applications and Examples
Edge Computing 4US Str
Edge Computing Keys, Trends
NEW Growth in Cloud is at the Edge
Cloud majors realized that even with Public->Private/Hybrid that there isn’t much growth left in the cloud space. Almost everything that can be centralized has been centralized. Most of the new opportunities for the “cloud” lie at the “edge.”
Edge Computing 101
Edge in this context means literal geographic distribution - and opposite of centralized - whether cloud/thick.
Edge computing is computing that’s done at or near 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.
Trend: Pervasive IoT Micro-Controllers with Internet Connections
[In 2020s] nearly every consumer gadget, every household appliance, and every industrial device will be connected to the Internet. These connected devices will also become more intelligent with the ability to predict, talk, listen, and more. The companies who manufacture these devices will have an opportunity to reimagine everything and fundamentally transform their businesses with new product offerings, new customer experiences, and differentiate against competition with new business models.
Every year, over 9 b of these devices are already being shipped with a SOC/CPU/MCU chip(s) with the compute, storage, memory, and an operating system right on the device. Right now few of these are connected.
Increasingly devices like Google/Apple Home/Amazon Echo, etc, Nest thermostats, Ring doorbells, security cameras, Home Area Networks, Wifi Router-Extenders, etc. all work of a Home's Wifi or Bluetooth networks.
Soon many of these 10b+ devices/year will be connected to internet. Or at least MCUs will have the BT/WiFi built-in as it is cheaper to build it in, instead of special purpose dumb parts.
Why: Edge Solves Latency Critical Problem
Almost any technology that’s applicable to the latency problem is applicable to the bandwidth problem.
Multiplayer video games implement numerous elaborate techniques to mitigate true and perceived delay between you shooting at someone and you knowing, for certain, that you missed.
Similarly, when you do a lot of writes to a database, there is a latency, before they appear in the read only version of a site, so much so, it becomes a bottleneck for scalability. By decoupling and pipelining the writes you can cache the read-only parts of the site, so it becomes much worse - more than latency of "notifying" the server, it can take minutes to hours (to days in payment ACI reconciliation systems) for the transactional integrity to be done and then reflected in the caches.
Why: Bandwidth Consumption reduced by Edge Computing
- If you buy one security camera, you can probably stream all of its footage to the cloud. If you buy a dozen security cameras, you have a bandwidth problem. If you have several facilities protected by security cameras, you have to have dozens of people watching (hopefully) and now you have a huge payroll, cost and compliance problem. But if the cameras are smart enough to only save the “important” footage and discard the rest, your internet pipes are saved and you save oodles on human costs.
Why: Offline Apps enhanced by Edge
Progressive Web Apps typically have offline-first functionality. That means you can open a “website” on your phone without an internet connection, do some work, save your changes locally, and only sync up with the cloud when it’s convenient.
Why: AI Edge Apps - privacy and efficiency
Companies are combining local computing including AI features for the purpose of privacy and bandwidth savings.
- Google Clips
- iPhone Biometic FaceID
What is special about Edge - not just a buzzword!
- Edge computing - Wikipedia
- d What is edge computing? - The Verge
- Edge computing vs. fog computing: Definitions and enterprise uses - Cisco
- Enterprise Networks - 5G technology needs edge computing architecture - Cisco
- Edge computing – how important is it? | InsightaaS
- What Is Edge Computing? | Cloudflare
- What is Edge Computing? | GE Digital
- What Is Edge Computing? | NVIDIA Blog
Edge Computing Architecture and Technical Review
Security in IOT and Edge - enabler
There has been huge pain and suffering with corporate CEOs being fired and consumers have experienced with insecure IOT.
IOT and edge devices can disrupt and do damage on a larger scale.
Target suffered loss of hundreds of millions, data privacy of customers lost, and CEO fired due to HVAC/IOT insecurity
2016 Mirai botnet attack where roughly 100,000 compromised IoT devices were repurposed by hackers into a botnet that effectively knocked the U.S. East Coast off the Internet for a day.
Data Privacy and PII in IOT and Edge - transmission and storage
The security and privacy features of an iPhone are well accepted as an example of edge computing : by doing encryption and storing biometric information on the device, Apple offloads a ton of security concerns from the centralized cloud to its global users’ devices.
Admin, DevOps is a weak link - Edge Computing must be easy to Manage
The management aspect of edge computing is hugely important for security.
Edge Device IoT security with Certified Secure Elements
It seems an easy guess that most of the hardware you buy a few years from now will have its software updated automatically and security managed centrally. Because otherwise your toaster and dishwasher will join a botnet and ruin your life.
Consumer electronics are just plastic boxes with an abandoned Linux distribution inside.
Evergreen updated versions of Firmware - upto-date security
- Google, Microsoft, and Mozilla have had in moving browsers to an “evergreen” model which are always updating their Chrome, Edge or Firefox browsers automatically without even prompting in vast majority of cases. Nowadays, users don't even know what version of Chrome, etc. browser they have, and they are all updated to latest security patch.
Certified Secure and Cheap MCUs Edge Computing ~Azure Sphere
- Microsoft is working on Azure Sphere for Edge Computing a managed Linux OS, a certified microcontroller, and a cloud service. The idea is that your toaster should be as difficult to hack, and as centrally updated and managed, as your Xbox. Microsoft has a specific solution to the IoT security problem but it may not be adopted universally.
- Secured MCUs: Azure Sphere certified microcontrollers: Combines both real-time and application processors with built-in custom Microsoft security technology and connectivity with learnings from Xbox.
- RTOS: Azure Sphere OS: This OS is purpose-built to offer unequalled security and agility. Unlike the RTOSes common to MCUs today, our defense-in-depth IoT OS offers multiple layers of security. It combines security innovations pioneered in Windows, a security monitor, and a custom Linux kernel to create a highly-secured software environment and a trustworthy platform for new IoT experiences.
Cloud: Azure Sphere Security Service - brokering trust for device-to-device and device-to-cloud communication through certificate-based authentication, detecting emerging security threats across the entire Azure Sphere ecosystem through online failure reporting, and renewing security through software updates. It brings the rigor and scale Microsoft has built over decades protecting our own devices and data in the cloud to MCU powered devices.
- Behind the scenes with Azure Sphere
- SEE: The Seven Properties of Highly Secure Devices - Microsoft Research
Shift away from User Settings, Awareness of Smarts hidden and Control of their devices
There is already a rapid shift away from user responsibility for the software they run on their devices. As mentioned for browsers, this starts from the need to get updates from the manufacturer automatically. - Macs, Windows have had auto-update for 10+ years (even without subscriptions) - Browsers (as mentioned above) - Nearly all iPhones and Androids download Gigabyte updates of firmware in background and users have some control when they actually update. - Larger apps like WhatsApp,FB,etc - updates may be very rapid. - Now Google Instant Apps is allowing very rapid incremental APK downloads
A self-driving car is evolving rapidly, eg Teslas keep calling back home to report hands-off-wheels, connect to EV charging networks. For that and technology reasons, inevitably has to be managed centrally. , it needs to send processed data back to the cloud to improve the algorithm, and the nightmare scenario of a self-driving car botnet makes the toaster and dishwasher botnet we’ve been worried about look like a Disney movie.
2020 Device/Mobile Apps, Limited AI processing on Chips
The more processing Amazon can do on your local Echo device, the less your Echo has to rely on the cloud. It means you get quicker replies, Amazon's server costs are less expensive, and conceivably, if enough of the work is done locally you could end up with more privacy.
- Amazon is rumored to be working on its own AI chips for Alexa.
- Facebook is heavily interested in AI accelerators for inferencing
Apple has already added and is enhancing AI accelerators in its X, XR, iPhone 11,etc series
SVR: Lambda Functions IOT and Edge - enabler for Smarter Edge
Evolution of Edge Computing Architecture
1940s? Relays, PLCs
1960s Centralized - Mainframe, Minicomputers
Mainframe Computers had distributed controls and processing
The switching networks, etc collected streams of data from async coupled devices like tape drives, terminals, etc.
32-bit Block mode terminals in IBM 360 - The character mode IO just could not be handled with nation-centralized mainframes. The motivation for doing this, however, was in most cases not response time or bandwidth, but rather to reduce I/O load on the mainframe.
1970s Minicomputers - Industrial Cell-Controls, 16 bit, 32 bit
Mini-computer smart terminals
Instead, the dumb terminals transitioned to smart terminals (still before personal computers or the widespread use of UNIX) those smart terminals amounted to an early example of edge computing according to the definitions proposed in this article. Important for Minicomputers like the VT100, HP "workstations" with function keys were programmable to send templates to the central server
1981+ PC Era => Thick PC Client-Server
1995+ Internet era
2007+ Mobile little Apps => MDM/MEAP => AppEconomy with Central Processing
2012+ SaaS Cloud: Dropbox, Gmail, Office 365, Collab - Slack
2017 PA Centralized With Lambda Processing ~SOA => Higher Latency
Your Amazon Echo has to do all of following in <5 seconds
- Recognize "Alexa" command to "wake up"
- Process your speech
- Send an compressed representation and encrypted packets of speech to the cloud
- The cloud has to decrypt and uncompress that representation
- Speech and MT to text is done - as most AI processing is done to NLP format - not speech based
- PA NLP Central Server (has to be scalable) processes it Rules Server/Logic => Figures out WHAT to do! (Alexa skills)
- Invoking eg Lambda services - often involves pinging another API somewhere, maybe to figure out the weather, and adding more speed of light-bound delay
- The central sends your Echo the answer as a speech again compressed and encrypted
- Finally you can learn that today you should expect a high of 85 and a low of 42 and dressing recco
- 2020+ 5G in production
Applications and Examples
USE: Smart Chip Cards, Secured Trusted Elements
Biometric Phone ID is edge
When your iPhone is trained it to recognize your face - it is AI, CV and Edge Computing - it is all done at cell phone level, with minimal personal data (hopefully) or compute going out on the network.
AI processing at edge
Google Clips keeps all your data local by default and does its magical AI inference locally. It doesn’t work very well at its stated purpose of capturing cool moments from your life.
PAs and Voice assistants are Edge devices, processing Centralized
PAs typically need to resolve your requests in the cloud, and the roundtrip time can be very noticeable.
2025 Self-driving cars and other Autonomous Vehicles
An important example of edge computing as due to latency, privacy, and bandwidth, you can’t feed HUGE amounts of data, cameras from all the numerous sensors of a self-driving car up to the cloud and wait for a response. Even with 5G, the cellular network is too inconsistent to rely on it for this kind of work.
Amazon IoT is very secure by default and Greengrass will push more to the edge.
Microsoft - Azure
Microsoft is chasing this shiny object - could be dangerous gamble in fear of its usual catch up, too late, modus operandi.
- Google Chromecast