Mesosphere raises $73.5M with Microsoft participating | Hacker News Reader

Mesosphere raises $73.5M with Microsoft participating

venturebeat.com

136 pointsmwilcox13 comments

"Mesosphere started in 2013 and is based in San Francisco. The startup has more than three dozen customers, a spokesperson told VentureBeat."

Hope those "three dozen" customers are paying a lot. Its quite hard to make the numbers add up on an 'open core' business model.

Considering Apple is one of their customers- I'd say they're doing alright.

And Verizon

http://www.prnewswire.com/news-releases/verizon-selects-meso...

We are using Mesosphere Enterprise DCOS at Valassis Digital, they are a great group of people to work with.

We believe this type of infrastructure to be the future and are already reaping benefits after just having launched it into our production infrastructure a few months ago.

Could we have launched it without their help and packaging, probably, but there are simply not enough experienced individuals out there to ramp up quickly. I see this analogous to running Hadoop in its early days and having Cloudera support as a safety net.

I am sure closed source schedulers and other for pay tools to round out the ecosystem will provide a sufficient business model for them to be successful.

Shameless plug: If you are interested in working in this type of environment, ping me (email in profile). Looking for Devops, Node and frontend (react, etc.) engineers. Seattle, WA; Livonia, MI; San Francisco, CA

> These new tools could help Mesosphere further distinguish itself from other companies — namely Docker and Mesosphere — that deploy applications in containers, which are lightweight alternatives to more traditional virtual machines.

Mesosphere needs to distinguish itself from Mesophere? :-)

That's the latest strategy. Compete with yourself. Win-win.

Is the Pardot signup for Beta broken? I've filled in info, but it keeps telling me it'll redirect me to put in more info. I've not received a verification or anything.

How this happened: - Mesos was largely the brainchild of Andy Konwinski (also co founder of Databricks - both Spark and Mesos were born in the AmpLab at Berkeley). Ben Hindman then interned/worked at Twitter and they needed cluster management. Ben used Andy's stuff (but contributed a lot to it). The mesosphere co-founders also used mesosphere after their experience at Twitter. They figured that everyone else would need it too.

- The key to this is that Andy interned at Google the year before, and worked on 'borg' and Omega. These are schedulers used at Google. Andy then took ideas from Borg and, to some degree, Omega, and just made it outside of Google.

- So this is basically technology that is 1-2 generations older than what Google has. I dont know what they have, but while Andy et al. were building Mesos, Google et al. weren't just sitting there, warming there asses.

We know what Google was doing, or at least doing and willing to make public: http://kubernetes.io/

I thought kubernetes was created by the Google cloud team and that a combined Borg and Omega is what's used internally?

Right. Google still uses Borg extensively in production. I am not sure if Omega ever saw the light of day to be used in production.

Kubernetes is mainly used as part of GKE but is not used internally at _all_.

Also worth reading: http://www.nextplatform.com/2016/03/22/decade-container-cont...

Great article, thanks!

Do you know if the Dataflow API is actually used internally as well? That's another thing that's a little confusing.

Almost nothing that you say here is true. I should know. I am good friends with all three of the original authors of Mesos: Andy, Ben, and Matei together. We sat next to each other for years while working on our Ph.Ds at Berkeley -- first in the ParLab/RADLab, and then in the AMPLab.

As a quick disclaimer I should mention I am now an employee of Mesosphere. I joined Mesosphere because I believe in the technology they are building, and it is certainly not 1-2 years behind everyone else. In fact, I left my job at Google to come here.

To touch on the exact points that you make. Andy would agree that Mesos was not purely his own "brainchild". Ben, Matei and Andy all worked on it together as a class project in the Spring of 2009 (I was also in this class, so was able to witness it all first hand).

Moreover, many of the ideas in Mesos were largely an extension of some of Ben's previous work on Lithe: a multi-level scheduling framework for managing multiple parallel libraries in a single application. I even took some of Ben's Lithe work and extended it for part of my Ph.D work. If anything, Mesos was the brainchild of Ben more than Andy in this respect.

As time progressed, Andy and Matei decided to focus more on Spark than on Mesos itself. This has lead to the (very successful) Databricks startup they created. Ben, on the other hand, decided to focus more on Mesos itself, bringing it to Twitter and now to Mesosphere in order to commercialize it.

Andys internship at Google has almost nothing to do with how Mesos has progressed over time. The internship itself was significantly after the initial Mesos paper was published, and focused mostly on evaluating the performance of Omega over Borg. I know this because I saw him give a talk on his work on Omega shortly after completing the internship. To claim that Andy took the ideas from Borg/Omega and applied them directly to Mesos is simply false.

To claim that Mesos is somehow stuck in the past is also false. It's true that Google and the AMPLab collaborated closely in the early days with people like John Wilkes and others giving lots of feedback, but Mesos has grown quite a bit since then. They haven't been simply sitting around warming their asses either.

Reply on news.ycombinator.com