Cloud database business Mesosphere has released version 1.11 of its 'Datacenter Operating System' (DC/OS) , aimed at slashing the cost of running multiple clouds, preventing vendor lock-in, simplifying management for platforms like Cassandra and Kafka and delivering Kubernetes 'as a service'.
"The vision myself and my cofounders always had was essentially building a cloud in a box, a cloud you can deploy anywhere," cofounder and CTO Tobias Knaup tells Computerworld UK.
Mesopshere was founded in 2013 by former lead engineer at Airbnb Tobias Knaup; the co-creator of open source distributed system management project Apache Mesos Benjamin Hindman who brought the platform to Twitter and is chief architect at Mesosphere; and Florian Leibert as CEO.
DC/OS was originally conceived to do what it says on the tin – to make the management of data centres, cloud, containers and distributed systems all simple to manage under one umbrella package.
Knaup says upgrades have been led by customer demand, and new to DC/OS 1.1 are improved security features plus a realignment of the product towards simplifying edge and multi-cloud operations, including tools for cloud bursting, region management and remote office infrastructures.
Mesosphere has also automated 21 steps for Kubernetes deployment, from provisioning compute resources through to generating config and keys, provisioning network routes and deploying addons – which the company describes as 'Kubernetes-as-a-service'.
Kubernetes was added to DC/OS in September 2017, and this latest release aims to make deploying it freer from the perceived complexities it carries with it.
"The idea is to make Kubernetes as easy to run as possible," Knaup explains. "We deploy pure-play Kubernetes and we don't hide it behind any APIs or remove things, our goal is to deploy it in a secure, production-grade way and integrate it with the other workloads on DC/OS."
This might look like running containers on Kubernetes and then connecting them to a Cassandra database that sits on DC/OS too. The Kubernetes stack can be installed through the DC/OS command line tools or a GUI, and the overhaul includes non-disruptive upgrades plus transport layer security and self-healing disaster recovery.
"Non-disruptive upgrades are one of the biggest pain points for customers – how you upgrade a distributed system without taking down all your containers," Knaup says.
"So we've automated that... Distributed systems are fairly complex and Kubernetes is no exception, so to fix things when it's broken usually requires manual steps and deep knowledge of the technology, but we've automated a lot of that."
The new release now includes edge and multi-cloud federation so customers can, for example, run a large set of compute at a core location and a smaller set of compute at an edge location, such as a branch office. "For some of our customers this is a hospital in a remote location, for another customer this is a cruise ship out on the ocean," Knaup says.
It also allows customers managing multiple cloud providers to configure their applications without making any underlying infrastructure changes, rather than having to take into account the different APIs for different clouds or switching tools. The idea is to manage it all from DC/OS.
The product itself takes the form of a simple control panel, and features tabs for a resource allocation overview dashboard, services, jobs, and catalogue for platform services like Tensorflow, Cassandra or Apache HDFS. Another tab allows for selecting regions.
"If you think about it the cloud provider in this case almost becomes a setting, you don't have to take on a multi-month project to use a new cloud provider, you simply sign up, configure that region in DC/OS, and your users are ready to deploy applications there," Knaup says.
There are also added capabilities in data security to protect data in transit and to authenticate users.
"Security has always been a priority for us – we have had financial services customers since the very beginning, so this is top of mind for them," Knaup says. "We've added capabilities for data security, so a lot of our customers are running big workloads.. databases, message queues, and so they really need to protect that data in transit and to authenticate users, and also other services that want to connect to those data services.
"We have made it essentially a checkbox to enable security for a lot of these workloads like Kafka, like HDFS, so security is really easy to set up for customers."
Ultimately, what does the CTO think are going to be the eye-catching details for IT buyers, CIOs and decision-makers?
"Cloud platform services aren't cheap," Knaup asserts. "A price that CIOs pay is that they give up flexibility, because those data services have proprietary APIs – and a message queue on AWS has a different API than a message queue on Azure and on Google Cloud.
"The other alternative is doing it all themselves – hiring a Kafka expert and a Cassandra expert, which isn't a feasible route for most companies. So DC/OS gives them the best of both worlds, they get a cloud-like experience where things deploy very easily and they're highly automated, but they're in control, they can choose on which infrastructure it runs, and they don't lock themselves into a cloud, because it is standard Kafka and standard Kubernetes."
Mesosphere has attracted attention from top-tier enterprise businesses like Microsoft and HPE, both of which have plugged significant funding into the business.
Now it counts "over 125 paying enterprise customers," Knaup says, plus "30 percent of the Fortune 50 companies", including in finance, technology and telecommunications.
When prompted that the company must surely be on the edge of an IPO, Knaup, perhaps predictably, returns a "no comment" and laughs.
But he does say that the business seems to have found a "sweet spot of applications" where "three things come together".
The first of these is delivering microservices and developer agility; the second, managing real-time data services and distributed systems; and delivering hybrid cloud the third.
"DC/OS is uniquely positioned to enable all these three things," he says. "I don't know another platform that can do these things well in a production-proven way."