A global land grab for dominance in the public cloud. The rise of OpenStack as a datacenter-automation tool. SaaS investment soaring as enterprises look for business analytics services. Docker making an end-run around Cloud Foundry.
These are a few of the key takeaways from my talk, “Cloud Trends”, delivered at the GigaOm Structure conference in San Francisco this week. (For the full presentation, click here; for the video, click here.)
The purpose of the presentation was to highlight that while cloud computing has been evolving rapidly over the last few years—and is now transitioning to become a mainstream component of enterprise infrastructure—it’s still hard to get an overall picture of the state of cloud broadly: the public cloud, platform-as-a-service, and SaaS applications. I’m putting out my own opinions and predictions, and hope to be back next year to see how well I did and come up with some new ones.
In my new role at Battery Ventures, I’m no longer a direct cloud customer or vendor. But I leverage my past experience and connections to advise small companies and enterprises in transition. Without the constraints of vendor hype, I’ve developed a somewhat independent viewpoint on current trends in cloud computing.
Starting with the leading public-cloud (or “infrastructure-as-a-service”) companies, it’s clear that there has been a significant shift over the last year. Enterprise use of cloud has moved from evaluation projects and unsanctioned, “shadow-IT” activities to the mainstream. There remains a lot of work to be done, but specifically, most enterprises have now added Amazon’s AWS and Microsoft’s Azure as strategic vendors, and are working through any compliance and regulatory approval processes as their cloud use-cases progress from mobile and marketing applications toward core business functions.
Enterprises are looking to adopt a safe-bet, two-vendor strategy with clear market leaders that have scale, a stable platform and a supportive ecosystem. Right now those safe vendors are AWS and Azure. In a back-to-back scale comparison at Cloud Expo earlier this month, Azure touted a customer using tens of thousands of cores. AWS trumped that with an example of a customer using 150,000 cores.
The rest of the IaaS suppliers are operating at much smaller scale and have significant issues to overcome. The contender with the most to prove is currently Google Compute Engine (GCE). This provider knows how to deploy very sophisticated, low-cost infrastructure and is investing at huge scale for Google itself. Yet the current scale of GCE is relatively small and is only available in three global regions, compared to ten for AWS and 15 for Azure. It’s surprising that Azure has data centers in more global regions than AWS, of course, but AWS’s size remains much larger. In-country datacenters are very important for government and financial customers with regulatory and security concerns.
There are very few enterprises using GCE, which may not be surprising considering GCE doesn’t have an effective, enterprise-oriented sales and support model in place. Part of this may link back to Google’s broader consumer-tech orientation. Still, we can expect the most aggressive moves in IaaS to come from GCE over the next year; the provider is driving prices down and forcing everyone else to respond, while GCE continues an aggressive product roll-out to fill their technology gaps.
Startups are interested in the comparison between AWS and GCE, so we took the public list of reference customers from both and fed it to research database Quid to analyze which companies (with reported funding events) are using these cloud providers. This excludes the large enterprise and government customers that AWS has, and excludes the smallest companies that have never taken outside funding. The result for AWS was 148 companies that have taken a total of $8B of funding. According to Quid, just 24 GCE-powered companies had received a total of $760M of funding. This is a crude comparison but it shows that AWS is currently far bigger than GCE in the startup space, and the startups are clustered in different markets. We can revisit this measure next year to see how it changes.
The rest of the IaaS market is in turmoil. Most vendors don’t appear to have traction, are for sale, or are going through technology transitions. One exception to watch over the next year is Digital Ocean, which is growing rapidly by pursuing a disruptive strategy of appealing directly to developers. One year ago, 10,000 websites were running on Digital Ocean’s cloud; today it’s 200,000, a 20X increase. The company is taking a big chunk of the low-end, Web hosting market from IBM Softlayer and Rackspace. But Digital Ocean is itself a startup. It’s going to have to raise a lot of money to support future growth, or be acquired by someone with deep pockets and an interest in developer driven IT.
Enterprises have two paths to private cloud and are taking both of them in parallel. One is to take their existing investment in virtualization with VMware or Microsoft and follow the hybrid-cloud technology route into vCHS or Azure. This gives customers a single-vendor, hybrid-cloud solution with relatively mature technology and strong support for Windows apps, but with vendor lock-in at a premium price.
The other path is based on open-source technologies and AWS. Initial enthusiasm for OpenStack hit the reality of significant support costs to make early versions work, but as it matures, OpenStack is doing an excellent job getting every vendor to cooperate in a single data-center automation standard, the “south bound” APIs that control and automate hardware. OpenStack is still weak on the “north bound” APIs that provide functionality for developers, and isn’t moving fast enough to catch up with AWS, Azure or GCE. My prediction: In the future, everything you run in the data center will be branded or include technology from OpenStack, but despite the best efforts of the OpenStack consortium the standard will fragment into similar but incompatible stacks
In the Platform as a Service market (PaaS) Cloud Foundry has a dominant position. It’s been endorsed by most of the large enterprise vendors, and is being used or evaluated by most of the enterprise customers I speak to. CIOs see PaaS as a way to provide developers with an agile environment that they have some ability to control and impose policy on. The disruption in PaaS is coming from Docker. Everyone has adopted the Docker container, and Cloud Foundry is going to have to do the same.
Think of Docker’s technology as a disaggregated PaaS, where each component is useful and non-threatening to other vendors, but its strategy is to use developer enthusiasm as a way to do an end-run around the rest of the PaaS vendors, and create a way to make applications portable. Without Docker, an application would be portable from one Cloud Foundry installation to another. But with Docker, the application is no longer tied to a specific PaaS.
Finally, on the SaaS front, I wanted to examine which enterprise oriented SaaS sectors were receiving the most funding from investors today. To me this is important because for end- users of cloud, perhaps the most important questions today are around usage. Is a particular cloud technology just getting developer adoption at startups, or has it matured to become enterprise ready?
A Battery analysis using data from Quid shows that, very generally, investment in SaaS companies has been growing over the last two years. More recently, investors seem very focused on areas like business analytics (big data), talent management and file sharing. Often the big numbers in these sectors are being driven by mega- investments in a handful of companies. Still, it will be fascinating to re-examine this next year to see how investment patterns have changed.
I’ve had a lot of interesting conversations in my first six months with Battery Ventures, and this is an exciting time, as cloud really takes off in the enterprise. I’m looking forward to the next year, and I’m paying particular attention to the rise of Go as the primary language for building infrastructure components; the move from monolithic applications to micro-services; and the adoption of DevOps and continuous delivery by “Lean Enterprises”.
See you next year.
Image © 2014 by Jakub Mosur, used with permission.