It’s been just over a year since I left Netflix and joined Battery Ventures. So it seemed appropriate (if a couple of weeks late) to take a look back at some technology and cloud themes that bubbled up in 2014 and offer a few predictions for the coming year.
In 2015 I expect more hubbub over everything from the Docker/containerization craze to Netflix’s open-source cloud platform to — dare I say it? — the end of enterprise computing. Here are some thoughts about the recently ended year in tech, in no particular order.
The Netflix open-source platform got traction
The Netflix team continues to release projects (about ten new repos on GitHub during 2014) and get more traction.
Notable external use cases for the Netflix platform include growth in interest in the Reactive programming model using Hystrix; the Spring Microservices architecture including Netflix; IBM’s Watson services, built using NetflixOSS; and Nike’s online services using NetflixOSS described at the AWS Re:Invent conference. Some aspects of the NetflixOSS architecture have been more widely influential, as seen in the growth of interest in microservices and the immutable service model encouraged by Docker.
Docker wasn’t on anyone’s 2014 roadmap, but is on everyone’s 2015 roadmap. (There was even a New York Times story about it earlier this year.) The Docker open-source project—which automates the deployment of new applications inside software “containers” — is an excellent example of how to drive viral adoption of a developer product, and it combines four useful things in one. It’s portable, speeds up development, defines the configuration and is shared via Docker hub. It’s become a key ecosystem and will undoubtedly continue to grow in 2015.
The concept of “anti-fragility” took off
The ideas behind the Netflix Chaos Monkey, a Netflix service that helps test the automation that helps systems recover from problems, are that you have to prove you are resilient and exercise failure-recovery mechanisms by creating your own failures. This is now so prevalent that it’s being mentioned in unexpected places, such as a business discussion with Workday, and a talk by the CIO of the Department of Homeland Security Citizenship and Immigration Services at the DevOps Enterprise Summit. As enterprises re-architect their systems using principles from DevOps, micro-services and cloud native architectures, the trend is to bake in and automate recovery and resilience.
Amazon Web Services continues to dominate cloud computing, and the service doubled its IP address range again this year, to about 10 million. The IP address range sets an approximate upper limit on the number of instances that AWS could run at the same time, since by default most instances get assigned one address. It is one of the few available metrics that shows the growth rate.
An interesting reversal occurred during 2014: Previously, clouds were seen as having missing features compared to data centers, but now there are many startups that are building products for data centers to provide features that already exist for AWS. It appears that the most sophisticated operations architectures are now on AWS, not on premise.
Microsoft Azure is getting a lot of enterprise-cloud signups but still doesn’t represent a significant proportion of the overall cloud market. There were several large and embarrassing Azure outages in 2014, and relatively few non-Microsoft services were impacted enough for the public to notice.
AWS had a few zone-level, partial outages or network partitions, but nothing significant enough to cause widespread impacts. Notably, it’s now two years since the last big AWS outage. (Remember all the press those used to get?). While AWS has matured and made its services and operating practices more resilient, Azure has some work to do.
Google spent 2014 getting enterprise features in place and hiring lots of ex-AWS people. But it still has a lot to prove as a cloud vendor. Google is an interesting alternative to AWS for startups, but the company is going to have trouble getting the enterprise market adoption that Microsoft and AWS have already figured out. Startup cloud vendor Digital Ocean, meanwhile, is growing fast and has carved out a space for itself as the simple, developer oriented, cloud solution. AWS has added so many features that it’s a full-time job just trying to keep up with them, so I think there is a place in the market for something easy to understand and use.
Bottom line: The big, traditional enterprise-computing vendors are failing to grow their customer bases. You can watch their revenue from new-product sales fade.
Services and support revenue will increase to compensate in the short term, but even that will eventually collapse as customers move on to low-cost, open-source solutions or outsource to cloud-based services. This is one of those times in which replacement technology revenue is an order of magnitude cheaper than the incumbent revenue.
For example, we could see market segments that currently generate $10 billion of revenue for traditional enterprise-computing vendors be entirely replaced by $1 billion of revenue for cloud vendors and open-source based startups. My friends Peter Magnusson and Marten Mickos joined Oracle Cloud and HP Cloud in 2014. I wish them well, but I’m not optimistic that they will be able to generate enough revenue to offset the losses elsewhere.
Adrian Cockcroft is technology fellow at Battery Ventures. Prior to that he was cloud architect at Netflix, but also spent time as distinguished engineer at eBay and Sun Microsystems.