First things first

In hindsight, taking vacation the week of Google Next probably wasn't the greatest idea, as I spent most of Monday trying to catch up on and digest everything that was announced. There was a lot -- some of which I link to later on, and some of which you can find reported in news stories and on Google's various blogs -- but the biggest deal has to be Google's hard embrace of private/hybrid/multi-cloud computing. As I think anybody would have predicted (especially after the announcement of on-prem Kubernetes applications last week), Google made Kubernetes the focal point of its new on-prem capabilities.

To me, the main items here are GKE On-Prem and Knative, the latter of which is a set of tools designed to let users run serverless applications on Kubernetes wherever it's hosted. Google explains GKE On-Prem more in this blog post, and Knative in this blog post.

I wrote a lot about cloud providers going on-prem last week, so I'll spare you another explanation of why that's such a good idea. But here's a recap of last week's newsletters in case you missed them:

When it comes to Kubernetes specifically, it's considered such an ideal platform for hybrid computing because it marries the build-once, deploy anywhere nature of containers with a standard platform for orchestrating all those containers and the applications they comprise. While cloud providers stand to make a lot of money selling managed services for Kubernetes and even higher-level abstractions such as serverless/Lambda/functions-as-a-service, those types of offerings also come with a degree of platform lock-in and lack of control that makes some customers uncomfortable. That's there's also a huge market for letting customers do these things on infrastructure they manage, be that on bare metal or VMs inside their data center, or on raw cloud-provider IaaS resources.

The last point has always been a given as far as I'm concerned, which might explain why my post-Gigaom software-industry career has been exclusively with companies focused on containers and private clouds. And, it should be noted, my current employer -- Pivotal -- is very much a part of this discussion. Pivotal is a significant contributor to the Knative project Google announced last week, and the two companies have been partnering on Kubernetes since early 2017.

This matters because it gets to the core question of how cloud providers (not just Google, but also AWS and Microsoft) will ultimately choose to run their hybrid cloud businesses. On software they build themselves, they could try to cut out the middlemen and own every part of the sales and installation process or, as Google's Urs Holzle suggests Google will do around GKE On-Prem, they could rely on hardware and integration partners to handle the on-prem part. But what about partners such as Pivotal and Red Hat, which have their own products for doing private/hybrid Kubernetes deployments?

Assuming Google makes a serious investment in GKE On-Prem and that other cloud providers do the same for any hybrid software they release, I predict the answer is that they will keep all options open depending on what any given customer wants and needs. We already see this in the database and data-processing spaces, for example, where each cloud provider has its own proprietary services as well as managed versions of popular open source options such as MySQL and Apache Spark. And, of course, many customers forgo managed services and instead run open source or third-party software on top of IaaS resources -- a deal that retains control for customers, but still results in revenue and case studies for cloud providers.

But, really, it's far too early (for me, at least) to foresee how the cloud market will shape up over the next several years, especially now that it's moving so fast into private/hybrid/multi-cloud/edge architectures. From the technological to the ethical, there are so many forces at play that the only thing that seems predictable is continuous change. The movements we've been following over the past several years are all converging, and it's going to be very exciting to watch what happens.

And if you want some good analysis of the Google Next event overall and Google's enterprise IT chances, both Stephen O'Grady and James Governor at Redmonk wrote good blog posts on it:

On AWS and facial recognition

I'm generally a supporter of the American Civil Liberties Union, but its "study" that claims to demonstrate the inherent bias of AWS's facial recognition service, Rekognition, seems kind of like a publicity stunt designed to capitalize on hype over artificial intelligence.

Facial recognition is a technology that raises lots of ethical concerns, and stories about problems with law-enforcement and judicial applications of certain technologies are all too real. However, as AWS machine learning head Matt Wood points out in a blog post responding to the ACLU piece, its methodology appears flawed along several different fronts -- including the potential of throwing the baby out with the bathwater.

Also, Wood makes a solid point about the differences in confidence levels required for different applications: "There’s a difference between using machine learning to identify a food object and using machine learning to determine whether a face match should warrant considering any law enforcement action. The latter is serious business and requires much higher confidence levels." The ACLU apparently used the default setting of 80 percent, whereas AWS recommends 99 percent for that use case.

If we're going to seriously debate the threats and opportunities that AI poses, it would be best to do so honestly and accurately.

Read and share this issue online here.

AI and machine learning









Cloud and infrastructure










Data and analytics