Reinvent

Tag: Reinvent

Parsing Amazon’s FreeRTOS announcement: Interesting, “but kind of strange”

roy murdock VDC researchThe December 4 edition of PPR talked about some of Amazon’s IoT and cloud announcements at the AWS Re:Invent conference in Vegas. Amazon’s move to take over development of the FreeRTOS embedded operating system was a surprise to many. PPR reached out to Roy Murdock, Analyst, IoT & Embedded Technology at VDC Research, to get some context around Amazon’s FreeRTOS announcement. As Murdock notes, it’s not just the vendors for sensors and microcontrollers that will be impacted.

PPR: What’s Amazon’s big play here with FreeRTOS?

Roy Murdock: One of the reasons we saw this as a really interesting (but kind of strange) move is that the IoT & Embedded Operating System (OS) market is a mature market that’s in decline right now in terms of revenues.

One of the numbers that we cite in terms of more recent revenue figures is that from 2015 through 2020, we expect the compound annual growth rate on average for that market is really only around 2 percent. It’s not a high-growth market, and it’s really not a market that a lot of companies are looking to get into right now.

In fact, a lot of acquisitions and mergers are happening right now are talked about in the IoT & Embedded OS market. So there’s a lot of consolidation going on. It’s an uncertain time for the companies in that marketplace.

It’s interesting that Amazon has decided to get into this marketplace given the low growth expectations. But in terms of your original question why would they be doing that, while revenue growth is pretty slow, is muted, is slowing down, expected to slow down, unit growth is actually growing very strongly.

So given IoT, given the ubiquity of sensors and MCUs and small chips and whatnot going into devices these days, which all require operating systems (usually an RTOS for a lot of these really small MCU/DSP-based systems) the unit growth is actually very strong.

We see in the future Amazon taking a lead in RTOS by unit shipments such as FreeRTOS which you know doesn’t generate much revenue for the company given that it’s free, but does capture a large share of the unit shipments, and taking those unit shipments and trying to get some of those devices onto AWS to generate services revenue.

So that’s what we think Amazon’s long-term play is. That’s what they’ve said their long term play is, it is really not too much of a secret. They’re providing a free onramp for a lot of those small devices to get hooked up and start generating data and generating services revenues through AWS.

PPR: One thing that was missing from the announcement page on the Amazon Web site is there weren’t any other vendors that were mentioned.  

Murdock: They do have a few vendors in place, one off the top of my head would be NXP. So they do have a few hardware vendors in place that are set up to kind of support this announcement.

And one of my coworkers was joking. He is more on the hardware side, but he was joking, ‘Everybody is going to want to slap an AWS FreeRTOS sticker on their hardware in the days to come.’ Especially some of these guys like NXP who are making MCUs and some of the smaller-resource hardware.

PPR: The other part of this announcement concerns AWS Greengrass, which puts a local instance of AWS, so it doesn’t have to be connected to a public cloud. Is that what it’s all about?

Murdock: That’s exactly what it is. FreeRTOS would be running on an MCU, or Greengrass would be running on a gateway, a Linux-based system usually with 128 megabytes of memory or more. So, a pretty high-resource system, but it would be running a local instance of AWS IoT which is the data ingestion/data sync portion of AWS’ service.

So the idea there was mostly that there are a lot of companies that would want gateways running locally that might have that might be in environments where they’d have spotty connectivity. So devices at the edge could sync up to Greengrass which wouldn’t have to Always be connected to AWS, the cloud portion of the IoT deployment.

So when the device communicates into Greengrass, it could store some of this data locally. And when that connection to AWS IoT in the cloud is available, then push that data to a cloud and sync things up.

And it could also run some edge compute locally instead of having to send everything to the cloud. You could run an algorithm to say ‘hey, this is the data I want to send up to the cloud.’ You can run edge analytics with Greengrass.

PPR: There’s a lot of vendors out there, and a lot of choice that these vendors have. Is the Amazon FreeRTOS announcement a real game changer, or vendors can already go to somebody else if they want to have this type of MCU/operating system/software stack behind it?

Murdock: I think it’s a real game changer for two different types of competitors, in that these two groups of competitors should be scared given this announcement.

Number one would be your traditional kind of MCU/RTOS vendors. In the past that would have been Micrium and Express Logic, vendors that ship hundreds of millions of RTOS systems into MCUs every year. Those guys have already been under attack by FreeRTOS for a long time. Of course, they had some protection in the types of systems they’re getting into that required safety certification or required some type of commercialization effort.

And then Micrium had actually gotten acquired by Silicon Labs last year. So they were they’re no longer really challenged by this announcement because they’re out of the game now, they’re under a hardware makers wing, so to speak. Express Logic is definitely feeling the heat right now, as are a few other smaller MCU vendors. That would be the first class of vendor that’s looking to be acquired or looking to really get out of that space, because Amazon is going to make it extremely hard for them to compete against (the already very-popular) FreeRTOS.

The second class of vendors that should feel the heat, and should kind of re-evaluate their OS strategy in light of this, would be Google and Microsoft. So the the cloud competitors on the Amazon side. And obviously both of them have OS ecosystems. Microsoft has Windows, and has Windows 10 IoT core, which is it’s not exactly an MCU-class operating system ecosystem, but it is a kind of smaller-scale Windows. And Google has Android.

So they both have operating system ecosystems to play with. But neither of them has just the sheer reach in the MCU engineering space that Amazon now has with AWS FreeRTOS. So they should both be kind of concerned about how many devices Amazon is really going to be able to drive onto AWS through this acquisition.

PPR: What about other cloud vendors out there on the enterprise side? Off the top my head I’m thinking IBM, Oracle, and there’s some other enterprise players. Do they have any kind of operating system partnership or assets that they could bring to the fray?

Murdock: Good question. Nothing that would call to mind the direct control that Microsoft and Google have over operating system ecosystems. So I would say that is kind of a secondary thought. I mean it is a good point. IBM, Oracle some of those other kind of enterprise cloud guys could have also benefited from an OS partnership like this one.

PPR: What does this mean for the actual companies that will be using these sensors and MCUs. The OT staff or the IT staff. Does it really make much of a difference to them about about this type of announcement from Amazon?

Murdock: I think it will make the migration path a lot easier. Especially in the OT side who are just looking to get sensors hooked up to their machinery on the work floor or shop floor. I think it will create a much easier path to get the devices onto the cloud, and that obviously continues to be a huge friction point. Especially in safety-critical or time-sensitive networking environments.

So I think you’re going to see Amazon pouring a lot more more investment into that path of getting those types of engineers, getting their sensitive data or running analytics into the cloud. Otherwise it wouldn’t make sense financially for them to do this acquisition.

Amazon’s leap into industrial IoT at #Reinvent17

Amazon made a slew of announcements at Amazon Web Services’ Re:invent conference in Las Vegas relating to new and expanded cloud offerings. You can see a Reinvent roundup here. I can’t go through them all, but will highlight those that have relevance for PPR readers in industrial tech. (Note: PPR parent i30 Media derives a large proportion of its revenue from Amazon).

One-click IoT?

AWS IoT One-Click is a service that can trigger basic AWS Lambda functions such as opening a door or reordering an item. It’s the same idea that makes Amazon’s “Dash” buttons so popular for instantly reordering food and sundries from the retailer without having to log in to the website or app.

But we’re not talking about ordering dog food and trash bags. We’re talking about industry, where the stakes are much higher. For many industrial firms, it’s not possible to leverage such tools, no matter how easy they are to set up and run.

The main issue, of course, is any process that depends on the public cloud has to accept that latency and availability may suffer. What if an AWS IoT One-Click function to reorder a critical part can’t be launched, because AWS is down? It happens, sometimes for hours at a time.

A secondary issue: IoT cloud security. Some firms will never accept data or applications launched from a public cloud, no matter how secure the service provider claims it is.

But there are industrial use cases for the new Amazon services. If a firm depends on data outside of its own facilities, it very may well have to fetch it from an application connected to the cloud. An automated irrigation system that uses weather data to determine water or fertilizer coverage on a plot of land will naturally have to pull the data off of a feed somewhere … and that somewhere is very likely hosted on an Amazon, Google, or Microsoft cloud service.

There are other situations in which it’s just not practical to connect to central IT or have edge processing, either because there is no local power available or there’s no space to put the box. The German railway giant Deutsche Bahn is rolling out an IIoT-driven monitoring and predictive maintenance system built by KONUX. Low-power sensors are placed between or near the tracks to monitor the weight, speed, and other metrics of rolling stock and 70,000 switches. The sensors are small, and for most of them, there’s no place to put dedicated edge processing or storage. Instead, the data has to be uploaded to a cloud service (either public or privately managed away from the edge) to generate analytics, alerts, and other machine language-driven services.

There is also an argument to use public cloud services when they are not time-sensitive or a point of failure for a critical system. Certain types of backups or archives, after-the-fact analysis, and specialized applications can use cloud data or compute resources without impacting operations.

Amazon moves into embedded operating systems

Of the other Amazon announcements last week, the one that caught my attention was the  Amazon FreeRTOS, an embedded OS for sensors and other small IoT devices. FreeRTOS has been around since 2003 as a quasi-open-source project; Amazon last year hired the lead engineer Richard Barry and makes the code available via an MIT license.   

According to the announcement, devices running FreeRTOS can now be more easily integrated with Amazon cloud services, or be used at the edge via Amazon Greengrass, which also includes some ML platform capabilities.  

Endpoint

The slew of cloud and IoT-related announcements at Reinvent are classic Amazon. Jeff Bezos’ flywheel strategy involves linking powerful platforms in ways that lead to positive network effects for both (think Amazon’s online retail store and the Amazon Kindle). Encouraging vendors to use FreeRTOS over another emerging embedded OS (LiteOS, TinyOS, VxWorks, etc.) is not just about grabbing a piece of the embedded OS pie, it’s about expanding demand for various AWS services and laying the foundation for long-term control over the enterprise cloud market.

If anyone from Microsoft Azure, IBM, Oracle, Google, or other players in this space is reading this: If you’re not looking over your shoulder already, you need to start doing so right now. Spin up your anti-Amazon countermeasures, partnerships, and customer bonding strategies ASAP. If you don’t, you may very well see Amazon eating your lunch, faster than you can say “Barnes & Noble.”   

As for OT and IT teams, the new easier-to-manage Amazon cloud services such as AWS Fargate and Amazon ECS for Kubernetes (EKS) will be a boon to groups managing back-end processes and applications that can live with cloud latency or the potential for service outages. They promise to make applications easier to write and manage, and potentially bring down costs if they compete favorably with offerings from other providers. But other types of applications that manage industrial automation or mission-critical services cannot be placed on public clouds–the risk to operations or conflict with security need are too great.

Amazon Greengrass may provide some comfort if Lambda instances can be hosted at the edge, but the one thing that I don’t see happening (yet) are the vendor, application, and hardware partnerships needed to make that happen in a plug-and-play way. But you can bet that startups in this space–and maybe some established vendors–will be taking a hard look at the new goodies announced at #Reinvent17.

Stay tuned for #Reinvent18 to see what Amazon does next …