If AWS is serious about Kubernetes, here’s what it must do
- 15 August, 2017 20:00
Amazon Web Services has joined the “Anyone-but-AWS” club, pledging its support to the Cloud Native Computing Foundation to better align with the Kubernetes crowd. It’s not as if the cloud giant had much of a choice: As much as AWS wanted to ignore Kubernetes into obsolescence, the gravitational pull around Kubernetes is simply too strong. Although most people view the CNCF announcement as a big endorsement for Kubernetes, AWS has been far cagier, offering precious little information on what it plans to do.
The big question is just how far Amazon will go to support the open source project voted most likely to “take down AWS,” as WS02 CEO Sanjiva Weerawarana said. Here is what AWS must do: Build a Kubernetes service.
Rumors have been swirling that AWS intends to do just that, but the easiest way to determine how serious the company is comes down to its level of Kubernetes code contributions.
So far, it’s not enough.
Kubernetes upset AWS’s serverless game plan
Kubernetes wasn’t the script AWS had written for containers. AWS has been aggressively advancing the serverless revolution through things like AWS Lambda. In the world of serverless, container orchestration becomes less of a thing. As Expedia’s head of technology, Subbu Allamaraju, posits, “Serverless patterns are pulling the rug from underneath container cluster managers faster than the latter [are] becoming industrial grade.”
Even so, far more enterprises are digging into containers than serverless at present, and for those that do, their default container manager is Kubernetes. As containers became hot, AWS built out its own Elastic Container Service (ECS), no doubt believing that its tight integration with other AWS services would be more than enough to see off the Google-spawned Kubernetes challenge. Indeed, Deepak Singh, general manager of AWS Container Services, told me that customers didn’t see the need to use something like Kubernetes to avoid AWS lock-in. If anything, they wanted more integration with AWS services, not less.
Reality, however, bites, and the industry has rallied around Kubernetes. Despite Apache Mesos being early to market, and Docker having its name stamped all over containers, Google’s Kubernetes has amassed the most momentum, with close to 75 percent of all enterprises indicating that their container orchestration strategy includes Kubernetes.
Kubernetes remains a long-term threat to AWS
Ironically, all that Kubernetes love hasn’t led to any sort of an AWS exodus, despite Kubernetes’s propensity to lower switching costs to other clouds, as Box engineer Demetri Mouratis said. Far from it. In fact, the majority (63 percent) of those running Kubernetes do so on AWS, according to CNCF survey data.
Except for the fact that Kubernetes could eventually uproot enterprise dependence on AWS, surely AWS could just sit back and collect a paycheck on all those workloads. At some point, however, Kubernetes simply makes it too easy to leave AWS, with vendors like Red Hat building its entire strategy on multicloud realities, facilitating exits from the AWS cloud. So, AWS has to get involved with Kubernetes, to protect its own interests.
According to Joseph Jacks, an active member of the CNCF, “Kubernetes increasingly represents an industry-standard API for running complex distributed apps. It is in AWS’s best interests to support this API.” In similar manner, analyst Krish Subramanian said, “AWS has tons of Kubernetes workloads and it makes sense for them to have a say in CNCF.”
AWS needs to puts is developers where its mouth is
It’s one thing for AWS to pay Kubernetes lip service via the CNCF. It’s quite another to pay down AWS’s Kubernetes interest in code. So far, however, AWS has not been active enough in Kubernetes to really have influence.
With Google and Red Hat doing the bulk of Kubernetes contributions—one with a rival cloud and the other with a strategy of eliminating lock-in to any particular cloud—AWS needs to ensure it doesn’t get locked out of Kubernetes product management, even as it has been MIA on Kubernetes (and container orchestration) product marketing.
You won’t, for example, find AWS in the Top 20 contributors to Kubernetes, except perhaps as part of the “Unknown” (Other) category. Heck, if you scroll down through the Top 40 contributors, you still won’t find AWS.
But with the hiring of Adrian Cockcroft, the former Netflix cloud chief who turned that company into an open source powerhouse, presumably AWS is planning to get serious about open source. AWS isn’t there yet, however, and it’s going to find it hard to build a service on a platform that it neither controls nor influences. Sure, AWS has built successful services on MySQL, PostgreSQL, and other open source projects, but none of these others is a good pattern match for Kubernetes—especially because it comes with built-in animus toward AWS in the form of Google and other competitors.
Of course, the infusion of AWS into CNCF will put significant pressure on the foundation, as Redmonk analyst James Governor stresses: “Will competitive engineering dynamics begin to hobble the project? Certainly we’ll now see the mettle of the foundation and its ability to govern. It’s easy to be ‘the lightweight foundation’ alternative when you have fewer technologies to manage and fewer competing interests.”
A strong AWS role may not be in Kubernetes’s interest
As WS02 CEO Weerawarana declares, “Kubernetes is the one thing that can take down AWS. They can’t afford not to join.” By the same token, the Kubernetes community can’t afford to have AWS hijack, sidetrack, or otherwise slow its momentum.
So far, as Redmonk’s Governor points out, AWS’s CNCF announcement gave no clear indication of its plans for Kubernetes: “The real question is how and when Amazon chooses to offer Kubernetes as dialtone. Those details are no clearer from the announcement.” That AWS plans to invest heavily in Container Networking Initiative (CNI) and Containerd comes across loud and clear. What it will do with Kubernetes, by contrast, is murky.
But AWS really has no choice but to get serious about Kubernetes, building out a Kubernetes service. If AWS did that, AWS would be at the center of the container market. If. More is needed, and much faster.