There were a few really interesting happenings today, running the gamut from container orchestration to agricultural robots. Because there’s no real unifying theme (well, except for maybe the catch-all of “software is eating the world”) I’ll just dive right into each.
I didn’t see this one coming, although I guess it makes some sense given the public-market situations with competitors Cloudera and Hortonworks. Both of those companies have been on a roller-coaster ride in terms of share prices (Cloudera in just 4 months, Hortonworks over nearly 3 years) but are trending up at the moment. If you can afford to wait and see what happens for a while longers, it’s probably a smart decision.
Perhaps MapR is waiting to see what happens with MongoDB’s impending IPO
, too. A third open source, enterprise software company borne of the NoSQL/big data era going public to lackluster results could scare off investors.
I have no idea what MapR’s financial situation looks like, so a generic list of possibilities might look like this:
- MapR holds off on an IPO and puts itself in a position to look like a smart, and safe, bet by comparison.
- MapR holds off on an IPO and gets its ship in order so it doesn’t look like a complete mess.
- MapR gets acquired. Honestly, I would have predicted more M&A in this space over the past few years, so maybe we’re due. I could be a smart target for any number of companies that need a better data story as IoT, AI and other data-centric use cases pick up.
Financial speculation aside, though, MapR has been doing a lot of work on the product side to differentiate itself. This includes announcing products for everything from hybrid cloud storage to containers, and from edge computing to deep learning. Solid adoption of those technologies could do a lot to help MapR future-proof its business story and, like both Cloudera and Hortonworks are trying to do, position itself as much more than just a Hadoop company.
I have a lot of thoughts about this because I used to work at Mesosphere (I left nearly 9 months ago and do not own shares) and now spend a lot time thinking and writing about Kubernetes. In general, I think this is very smart thing to do.
As to the actual news …
The very abbreviated lay of the land is that Mesosphere sells a product (there’s also an open source version) called DC/OS that’s built on a foundation of Apache Mesos and another open source project called Marathon. DC/OS can launch and manage Docker containers and non-containerized applications, and can run entire distributed systems on top of it (data systems like Kafka, Spark and Cassandra are very popular). Technically, DC/OS could run Kubernetes clusters as well and, since Google open sourced Kubernetes in 2014
, Mesosphere has been back and forth on how much effort it should put into making that happen.
There’s certainly a competitive angle to this, but it’s also just a matter of resources. If you’re Mesosphere in 2015-2016 and trying to sell enterprise software, you can either have your engineers work on improving your core technology, or you can have them work on a technologically tricky integration with Kubernetes. If you have good technology internally, this is an easy choice.
Fast-forward to 2017, and Kubernetes is mature and in high demand. So Mesosphere works with Google and builds a Kubernetes package to run on DC/OS. Given the relatively young container-orchestration market
and the customer base Mesosphere had amassed without Kubernetes support, the timing shouldn’t be much of an issue.
Getting back to that misleading headline … I believe that when you’re a software vendor, there’s really no "bowing to” or competing against foundation-managed open source projects like Kubernetes. The Apache Software Foundation and the Linux Foundation are not charging for software licenses.
You only compete against other software vendors. And once a project reaches critical mass, money will start flowing into the vendors that support it. You can either get your piece of that pie, or you can leave it for somebody else.
Or why every Hadoop vendor embraced Apache Spark. Or why every cloud provider offers OSS technologies as managed services in addition to their competing proprietary services.
Because they want to make money. If you can add a feature or integration that customers really want without sacrificing the integrity of your core technology, you’d almost be crazy not to do it.
If you’re unfamiliar with Blue River
, it makes machines (I guess you could call them robots) that identifies, and kills, weeds among rows of crops. John Deere, of course, is a huge company in the business of selling farm equipment.
Blue River is a prime example of why so many people are high on machine learning and AI for specific applications
, and are skeptical of companies promising general-purpose AI platforms. Blue River’s computer vision models are focused on one thing—identifying weeds—and they do it very well. While attempts to add data science or machine learning to massive agricultural datasets seem to have fallen flat
, Blue River’s machines are reportedly at work in 15 percent of the lettuce crop in the United States.