First things first


Once again, there was a remarkably high number of interesting news items and other content late this week, so make sure to read the whole newsletter. But here are my personal highlights:

Cloud providers spread their wings. Again.

Every time Amazon Web Services expands into an entirely new line of business (i.e., not just rolling out another IaaS product), we get a better view into how that company sees its place in the world. And with rumors this week that AWS is set to enter the learning-management space, I think it's safe to say -- assuming the story is accurate -- that AWS wants to be much, much more than a "cloud computing provider."

Not that this should come as any surprise. AWS has steadily expanded its suite of products since launching with S3 and EC2 more than a decade ago. From email to virtual desktops, and from vertically focused gaming products to a broadly focused call-center automation service. From a business perspective, the latter service, Amazon Connect, seems like the most direct relation to the rumored learning-management service.

Connect is a service that brings the Amazon approach to customer service to the world, and any sort of learning-management (or, possibly more accurately, employee-training) service could very well be Amazon's way of bringing some of its HR and corporate culture to the world. It seems safe to assume that "learning" doesn't mean challenging Coursera or other MOOCs, but rather training employees on necessary skills and tracking their progress. Whether that's good or bad for employees is debatable, but it's definitely a growing market (as the CNBC story linked to above points out) and one AWS might see itself as particularly well positioned to own.

(I will also be curious to see how much of this learning/training has to do with AWS technologies, and what effect that will have on AWS-training specialists such as A Cloud Guru. I spoke the founders of that company, which also runs the popular Serverless conference, on the podcast back in October.)

At this point, I tend to view AWS, more so than Google Cloud or Microsoft Azure, as a whole-hog enterprise IT vendor, on par perhaps with IBM during its heyday or, really, what a company like Oracle has grown into. IaaS is the core of the AWS business, which the company is using as a foundation for entrenching itself ever deeper and broader into users' business processes. It doesn't hurt that Amazon itself has internal processes and expertise to commercialize across such a broad range of areas, from consumer retail to online marketplaces to shipping and logistics.

Compared with Google or Microsoft, one big difference is that AWS was very distinct from Amazon.com from the start, and it entered into areas like collaboration (email, video conferencing, etc.) as extensions from IaaS. There was no large existing email or productivity business that then had to be integrated with a new IaaS business. AWS got started on IaaS early, established its dominance, and now is using that dominance to become a one-stop IT shop not so different from Amazon.com in the consumer world.

And in related news, Microsoft is launching a cloud-based gaming division to bring game developers onto the Azure platform and, it sounds like, to deliver a subscription-based game-streaming service. Gaming is another area where AWS got an early headstart with services like GameLift, Lumberyard and its Twitch acquisition, but Microsoft does have its own advantage here with Xbox. It knows the gaming industry and the people in it; it understands engineering for gaming systems; and it could build itself a nice data-engineering-business flywheel with a collection of developers on Azure and its own popular streaming service.

Kafka gets more competition?

The Cloud Native Computing Foundation has accepted NATS as in incubation-level project, giving its growing ecosystem (centered around Kubernetes) its own messaging platform. I'll be honest and cop to now knowing a lot about NATS, but my first question upon seeing this news was to ask how NATS compares with Apache Kafka -- the very popular messaging platform originally built at LinkedIn. (Recently, I wrote about another competitive project, called Apache Pulsar).

Apparently, I'm not alone in asking this. At the risk of vastly over-simplifying things, the consensus seems to be that Kafka is more mature and expansive in terms of capabilities (and probably has a much larger community), but NATS is more "cloud-native."

But the bigger-picture question that I always come back to is the influence of open source foundations and projects going forward. Despite being on the upswing in terms of adoption, Kafka is very much a product of the "big data" era where Apache projects (thanks to Hadoop) reigned supreme. The CNCF, on the other hand, started around Kubernetes and has been adding projects around it to fill out a more complete platform for building cloud-native applications.

While there have been some project-level integrations between these two worlds (especially around Kubernetes), they're still quite distinct worlds. Between the various open source options for certain capabilities and the options developed by cloud providers, one has to wonder what will emerge as the de facto set of tools for building next-generations applications. It seems like some sort of convergence of the big data world and the CNCF world might need to be on the horizon, especially if developers are looking for the strongest choices at each layer of their application architectures.

For more on the evolution of Kafka, open source and its connection with the cloud-native world, check out my separate podcast interviews with Kafka creators, and Confluent co-founders, Jay Kreps and Neha Narkhede. (I'm sure we've spoken about it on other podcasts, too. Just check them all out ;-) )

Google Ventures invested in a new AI/data-processing startup

GV led a $56 million series A round in a company called SambaNova Systems, which came out of stealth mode this week. What's most notable to me about SambaNova is the pedigree of its founders, which include Christopher Re and Kunle Olukotun of Stanford. They're both helping lead the DAWN Project there which, like the RISELab at UC-Berkeley, is trying to advance AI at the systems and data level, as well as at the model level.

It seems unlikely that AI efforts in most companies will be able to live entirely separate from other data systems and data stores, so the integrated approach of a company like SambaNova, which is building a hardware+software platform, could go a long way toward bridging those worlds. Or at least helping make them all perform a lot better.

For more on how research projects like DAWN and RISELab are thinking about AI, machine learning and data systems, check out my podcast interviews with DAWN's other two leaders, Matei Zaharia and Peter Bailis, and with RISELab director Ion Stoica.

ARCHITECHT

AI and machine learning













Cloud and infrastructure
















Sponsor: Replicated

Data and analytics