TL; DR: CoreOS presented a view of cloud computing without vendor lock in to cloud services. There is a huge amount of work necessary to deliver on such a vision.
We had the opportunity to attend CoreOSFest, the main CoreOS conference in San Francisco at the end of May. CoreOSFest is an event for the true believers in what I would loosely define as the CoreOS way. An opening talk about punk, DIY and ownership is appealing on various levels to many of the attendees, but it does not address the needs of the wider enterprise audience that ultimately signs the cheques.
A Vision of Cloud Independence?
This is interesting, @polvi is talking up higher order services (eg Athena) as a problem #coreosfest << value is in the packaging though
— Fintan Ryan (@fintanr) May 31, 2017
In his keynote, CoreOS CEO Alex Polvi built upon the initial themes of DIY and ownership as he spoke about the higher order services that are being provided by cloud providers, and more importantly from the CoreOS perspective the potential for vendor lock in that they create.
By way of example he highlighted Amazon DynamoDB, an AWS NoSQL offering, as a higher order service that locks you explicitly into AWS (as a side note Amazons database services are now one of the fastest growing segments in AWS), before moving onto an example of Presto on Kubernetes as a portable alternative to Amazon Athena.
The vision that CoreOS has is to use Kubernetes as the portability layer, and to use “operators” as the management mechanism for higher level applications. As we stated at the time operators were released “With the introduction of the Operator concept [..] CoreOS are taking complex software and making it easy to work seamlessly on Kubernetes”, and we still believe this to be true. However, it is still extremely early days for operators, and there is a long way to go before they become widely used.
Now we get, and understand, the message CoreOS are trying to put out there – but our view at RedMonk has long been that the best packager wins in each evolution of software. Packaging has multiple dimensions, and removing management overhead is a very large one. For example, we are seeing the rise of DBaaS offerings for a reason – databases are painful to maintain at scale. An operator for Kubernetes may reduce some of the overhead, but not all of it.
If your lens for viewing the world consists, almost exclusively, of open source software and a view focused around the actions of skilled technologists, this approach makes a lot of sense. However, the same question arises as we have previously noted with the current fascination, and almost fanatical obsession, around orchestration within the cloud native community. Engineering wise orchestration is an extremely interesting problem. From the perspective of a P&L owner, if someone else is responsible for all the operating overhead, maintenance and headaches, why would I not pay for it and free my resources to work on something else?
We welcome the approach being taken by CoreOS, competition is good, but a huge amount of work remains to be done before this is a viable approach for any but the very earliest adopters.
Key Partner Announcements
Oracle & Kubernetes
.@mcavage & @tjfontaine onstage at #coreosfest announce container linux support on @oracle cloud<< @oracle #cloudnative story is solidifying
— Fintan Ryan (@fintanr) June 1, 2017
The most significant announcement of the event came from Oracle, who confirmed that they will be participating in the Kubernetes project. From a strategic sense this make sense for Oracle. They are doubling down on multiple aspects of their cloud native approach, from working with Docker, to bringing CoreOS to the Oracle Cloud and now their work with the Kubernetes community.
As my colleague, James Governor has noted before, the technology industry creates strange bed fellows at times. We often find such alliances are mere press releases, but this is not the case in Kubernetes. Oracle is a case in point, this is far more than just lip service, there are a significant number of new hires working on Kubernetes, such as TJ Fontaine, formally the lead for node.js. At this point, rumours aside, Kubernetes has significant resources committed to the project from most of the major vendors, including Google, Microsoft, IBM, VMware and Oracle.
Oracles cloud offering is still limited in a number of areas, and significantly lags AWS, Azure and Google Cloud. However, the core IaaS components are coming together and proper support for tooling such as Kubernetes and Docker will address some of the initial requirements that enterprises have.
Draft from Microsoft
In keeping with the DIY theme of the day one keynotes, Gabe Monroy of Microsoft, who joined with the recent Deis acquisition, introduced Draft.
"the stuff is really hard to use for developers" @gabrtv #coreosfest << packaging, packaging, packaging (again)
— Fintan Ryan (@fintanr) May 31, 2017
Simply described Draft is a packaging mechanism for developers. Think of build packs for Heroku or Cloud Foundry, but with far more of the “magic” of the build environment accessible and customisable by the developer. Draft also creates a very simple deployment process for development and testing.
Draft is currently very much in the alpha software realm, but this is a project we are watching with interest.
Customer Conversations
We had several conversations with CoreOS customers during the event. Among the most interesting to us was with Concur, where security was highlighted as a key aspect for adopting CoreOS Container Linux and Tectonic.
Interesting, Ilya Borisov from @concur highlights the importance of @Quayio from @coreos to his team #CoreOSFest << security is vital
— Fintan Ryan (@fintanr) May 31, 2017
In general, however, conversations followed a relatively similar pattern. Where organisations already have a high degree of digital literacy, adoption of cloud native technologies is proving to be relatively straight forward. For others, many of the normal organisational challenges remain.
Conclusion
The vision of independence from cloud services is appealing to many segments of the core operations community, and there is indeed a large market for such offerings. However, the overall trend towards cloud providers offering managed services for infrastructure such as databases or message queues is only going to accelerate.
Ultimately enterprises will decide if removing the management overhead of complex infrastructure is a significant enough cost saving to offset the vendor lock in risk.
Disclaimers: CoreOS covered my T&E. CoreOS, Microsoft, Oracle, Amazon, Salesforce (Heroku), SAP (Concur) and Google are current RedMonk clients.
No Comments