Thank you for contacting us. We will get back to you shortly.
October 06, 2015 - by Casey Bisson
The first ever MesosCon Europe hackathon is right around the corner. Teams and ideas are already forming for development at the hackathon, so it's high time we shared some more details about the event.
Hopefully we all get a chance to build something silly, or fantastic, or just work through a really frustrating bug, but this is a hackathon, so the goal is to meet people, have fun, and learn from each other. That said...
Triton is the open source, container-native data center automation solution. Joyent uses Triton (formerly SmartDataCenter) to power our public cloud, and enterprises of all sizes (including enterprises of one) use it to manage private clouds. Because Triton offers real, multi-tenant bare metal container security, it eliminates the need for VMs that complicate traditional container deployments and solves the risk that private clouds suffer without security boundaries between tasks.
Triton's SmartOS container hypervisor offers trusted security by design and has been tested by nearly a decade of use powering Joyent's public cloud. In early 2015 Joyent introduced native, bare metal Docker support that treats an entire data center as a single, elastic Docker host. Together with best of breed storage and network virtualization, including VXLAN user defined networking, Triton makes Docker uncomplicated and scalable. Try deploying a whole database cluster for yourself, you'll have it running in minutes.
But Triton is narrowly focused on providing the best infrastructure for modern applications, and we believe application scheduling is a layer above the infrastructure and firmly a part of the application itself. The large number of Mesos frameworks to serve various needs is a testament to that: one size definitely does not fit all. We've been working with Mesos, however, and we're enjoying the combination of Mesos' outstanding scheduling features with Triton's leading data center automation.
Our work with Mesos so far changes the relationship between Mesos and the infrastructure. Rather than running the Mesos Slave on each compute node, something that would be awkward and insecure in a multi-tenant cloud, the slave interfaces with the provisioning tools for a whole data center. This requires relatively few changes in Mesos, and no changes in the frameworks we've tested on top of Mesos. The big win here is that this means you only pay for the tasks you schedule, and they run on bare metal where they belong. You don't have to provision and pay for VMs (or bare metal) before scheduling anything, and that means you don't pay for unused or inefficiently used VMs or bare metal that you'd have to pay for in a typical Mesos deployment.
The implementation today only supports Docker containers, and we've discovered a few places where Mesos has some baked-in assumptions about running on a single-tenant host, but these projects are all open source and compatibly licensed, so we believe these are solvable and there's lots of room to move forward. In short, we're excited to join the Mesos community and eager to work together to improve Mesos, an already great application scheduler, for public and private clouds.
We believe that Triton is ideal for a wide variety of applications and workloads, but what distinguishes it is that it's truly container-native. In addition to the community need and people’s choice award categories, the Joyent team will be looking for projects that demonstrate container-native advantages and may offer bonus awards for them. Some ideas about what we have in mind include:
We can't wait to see your hacks, and will be toasting the winners at an After-Hack party Friday evening at the Habourmaster Pub. See you in Dublin!
Credits may only be used for Triton services in the Joyent public cloud. Credits are non-transferable and expire October 31, 2016. Maximum monthly usage limited to $1,000. Recipient must accept Joyent's standard terms of service and register their user information with Joyent staff during the event. ↩ ↩