Amazon unveils Monitron, a service that uses ML and low-cost sensors to monitor industrial machinery, as it looks to push into the industrial sector

Sensor and computer vision hardware come as tech giant looks to push into industrial sector — Amazon is rolling out cheap new tools …

Financial Times 2020-12-02 Dave Lee

Context & Ripple Effects

Monitron is the public unveiling of the ML service that internal documents showed Amazon testing in September as AWS Thor — vibration, sound, and other signals used to predict machinery maintenance needs. It stacks on top of AWS IoT SiteWise, the managed IoT monitoring service whose industrial customers already include VW and Bayer, giving AWS a hardware-plus-software entry point into factories.

The launch extends a build-out that runs from Amazon's 2015 machine learning platform through Lookout for Vision for defect detection and, later, the warehouse-scale Sequoia robotics system. With Monitron, the same ML stack is pointed at the machines themselves, with low-cost sensors and computer vision hardware as the wedge.

First-order effects

Industrial operators get a low-cost retrofit path to predictive maintenance without building custom sensor networks, and AWS's existing SiteWise base — VW, Bayer — is the natural first pool of adopters.
AWS gains a physical-hardware foothold inside factories, where each sensor deployment generates recurring data and compute consumption on Amazon's cloud.

Second-order effects

Cloud rivals' industrial IoT offerings face a competitor that bundles sensor, connectivity, and model in one service, pressuring them to partner with sensor vendors or discount their own stacks.
Industrial automation incumbents see machine monitoring — a service long sold at premium prices — undercut by commodity sensors paired with cloud ML.

Third-order effects

If cheap-sensors-plus-cloud-ML becomes the default maintenance model, equipment makers face pressure to open machine data to third-party monitoring, shifting ownership of operational data from the factory floor's OEMs to cloud platforms.

The trend: Cloud providers are moving down the stack into industrial hardware and sensing, using low-cost sensors as an on-ramp to recurring ML and compute revenue.

Discussion

@mcwm Mike Murphy on x
isn't that a TV brand from the 90s ? https://twitter.com/...
@sub8u Subrahmanyam Kvj on x
Meet Amazon, the supplier of AI-based industrial sensors and analytics services. https://www.ft.com/... https://twitter.com/...
@9thstyle @9thstyle on x
there was a board member meeting and they decided “monitron” was the least dystopian of all the options they had https://twitter.com/...
@emmavj Emma Jacobs on x
https://www.ft.com/... Launched by Amazon's cloud arm AWS, the new machine learning-based services include hardware to monitor the health of heavy machinery, and computer vision capable of detecting whether workers are complying with social distancing. @DaveLeeFT
@rakyll @rakyll on x
Lambda is announcing container support. Base images for Python, Node.js, Java, .NET, Go, and Ruby are available. You can build your custom image by implementing Lambda Runtime API. https://aws.amazon.com/... https://twitter.com/...
@bianchiluca Luca Bianchi on x
The new Container Support for AWS Lambda is a game-changer for Machine Learning! Now you can package your preferred libraries to perform inference into a docker image and deploy them on AWS Lamda: https://aletheia.medium.com/ ...
@mojombo Tom Preston-Werner on x
You can now run up to 10GB of code on AWS Lambda via brand new container support. When we started @RedwoodJS almost 2 years ago, we targeted the serverless infrastructure we thought would be available several years hence. Nailed it! =) https://aws.amazon.com/...
@ericjhonsa Eric Jhonsa on x
Also announced by $AMZN: Network-optimized Graviton2 instances supporting 100G connections, and graphics-optimized instances featuring $AMD Epyc CPUs and Radeon Pro V520 GPUs. https://www.businesswire.com/ ... https://twitter.com/...
@_msw_ Matthew S. Wilson on x
This is a meaningful effort that extends the capabilities of PostgreSQL, and helps more customers and developers break free from closed software with bad licensing terms, and move to #OpenSource databases. #reInvent https://twitter.com/...
@mjasay Matt Asay on x
What I love about this is that it seems like incremental innovation, in a way (faster chips!), but it enables enterprises to do really groundbreaking work. The best innovation should never be about the vendor https://twitter.com/...
@hunter_owens Hunter Owens on x
fairly big news for the user (me) who has several thousand SQL Server licenses and DBs to try and integrate data from into a pydata-y stack https://techcrunch.com/...
@timbray Tim Bray on x
Without any inside knowledge, I'd be willing to bet that the EC2 Minis don't look like the kind you bring home from the Apple store: https://aws.amazon.com/...
@craigkerstiens Craig Kerstiens on x
Migrations to Postgres from both Oracle/SQL are a growing trend in large enterprises, big move from AWS (shots fired at MSFT) on open sourcing babelfish to help with SQL->Postgres migrations - https://aws.amazon.com/...
@brento Brent Ozar on x
Wait, WHAT?!? AWS is open sourcing Babelfish, a tool to let SQL Server apps use a PostgreSQL back end: https://aws.amazon.com/... “adds an endpoint to PostgreSQL that understands the SQL Server wire protocol Tabular Data Stream (TDS), as well as commonly used T-SQL commands” hooo…
@shawnbice Shawn Bice on x
Excited to announce Babelfish for PostgreSQL (open source project). A new end-point in PostgreSQL that understand commands from applications built for Microsoft SQL Server. Inviting all to participate! https://aws.amazon.com/...
@mjasay Matt Asay on x
Want to have more PostgreSQL in your life? We're open sourcing Babelfish to make it easy to move from SQL Server to PostgreSQL without changing your application code https://aws.amazon.com/...
@stevesi Steven Sinofsky on x
New - Use Amazon EC2 Mac Instances to Build & Test macOS, iOS, ipadOS, tvOS, and watchOS Apps https://aws.amazon.com/... // Apple is getting very close to end point and cloud symmetry, across endpoint devices too.
@tapbot_paul Paul Haddad on x
How does the Mac AWS stuff work? Are they running a hyper visor, reimagining Macs or booting from the network/FC?
@jmanooch John Manoochehri on x
In case you wonder what this is, it is the evolution of a remote compound eye & a kind of post-monadic visual intelligence (ie ‘sight’ from multiple objective positions), which is, indeed, new under the sun. Niander Wallace saw like this in Blader Runner 2049. Future be here! htt…
@quinnypig Corey Quinn on x
“These natively run macOS and are powered by Nitro and actual Mac Minis.” Holy crap that means you can mount EBS volumes on them, among other things. #reinvent https://twitter.com/...
@profmatsuoka Satoshi Matsuoka on x
Now it would be interesting if Amazon offers Mac instances on #A64FX; you could develop your favorite HPC apps on your M1 Mac & directly run much faster on EC2 & in massive parallel. A64FX CMG = Mac instance allows 1536 instances per FX1000 rack. https://aws.amazon.com/...
@dinodaizovi Dino A. Dai Zovi on x
This is pretty neat, AWS bare-metal Mac instances: https://aws.amazon.com/... Can't wait until they have M1 instances!
@drunknbass @drunknbass on x
Why Now? https://aws.amazon.com/...
@dcominottim @dcominottim on x
This was quick and it's an extremely interesting development. I wonder whether it'll help Apple's legal argument against Nuvia's Williams that he was setting up a competitor before leaving Apple. Possibly, yes. #SiliconGang https://aws.amazon.com/...
@timbray Tim Bray on x
@DonMacAskill @QuinnyPig @awscloud Of course, the shape of the deal between Apple & AWS is one of those Things We Will Never Know. To be honest, I'm seriously shocked, the whole thing seems out of character for Apple.
@edjgeek @edjgeek on x
AWS Lambda functions now have more room to grow with up to 10 GB memory and up to 6 corresponding vCPUs: Merry #Serverless Christmas. #ServerlessForEveryone #reInvent2020 https://aws.amazon.com/...
@gunnargrosch Gunnar Grosch on x
You can now allocate up to 10 GB of memory to a Lambda function. That means you can now have access to up to 6 vCPUs in each execution environment! Combined with the new billing in 1 millisecond increments, Lambda just got a whole lot better. #reinvent https://aws.amazon.com/...
@matthieunapoli Matthieu Napoli on x
What else is new for now: AWS Lambda now bills at the millisecond instead of 100ms https://aws.amazon.com/... Up to 10GB memory and 6 vCPU Lambda functions 💪 https://aws.amazon.com/... RDS Aurora Serverless v2 announced, extremely exciting but we'll wait for more details!
@hichaelmart Michael Hart on x
10GB of memory and 6 vCPUs on Lambda! Gee, I wonder what sort of workloads could use this? (hello ML and CI 👋) https://aws.amazon.com/...
@3266miles Dan ‘Hot Takes’ Miles on x
I have sunk countless hours into trying to tell people that DevOps isnt something you can “just buy” and then AWS goes and does this to me smdh https://twitter.com/...
@quinnypig Corey Quinn on x
“Is DevOps culture or tools?” “Neither, it's now a line item on your AWS bill.” https://twitter.com/...
@mike_puterbaugh Mike Puterbaugh on x
@Techmeme @Kyle_L_Wiggers “WARNING: DevOps Guru has detected that you have services running in us-east-1”
@mndoci Deepak Singh on x
A few weeks ago we announced that we were going to launch a public registry option for ECR. It's here https://aws.amazon.com/...
@nathankpeck Nathan Peck on x
The Amazon ECR Public Gallery has officially launched! https://aws.amazon.com/... One of my favorite things is that the limits for anonymous users are based on bandwidth usage, not pull rate. This encourages smaller images, and doesn't discourage the restart on crash pattern

Chronicles