Keynote 1: Welcome and Opening Remarks
Jim Zemlin, Executive Director of The Linux Foundation welcomed everyone to the conference and gave a short overview of the current state of OSS.
He stressed that “companies who don’t use OSS won’t be able to sustain in the long run.” GitHub hosts an astounding 38 Million repositories, the OpenStack cloud platform project is propelled forward by 49 engineering teams. The OSS cycle usually starts with a hobbyist project, which is built into a product by a company which then proceeds to make profit and launches new projects. This cycle needs to be accelerated and supported. The ecosystem around projects is important, think about governance, developer community, technical infrastructure, intellectual property.
Keynote 2: Incremental Revolution – What Docker Learned from the Open-Source Fire Hose
Solomon Hykes, Founder, CTO and Chief Product Officer of Docker talked about how contrary to common perception, Docker was 9 years in the making before being released in 2013. After a brief overview of Docker, he discussed what the Docker team has learned from OSS and gave a technology outlook at the end.
Docker seems themselves as providing tools for “mass innovation” and to provide the stack to “program the internet”, a sort of “digital supply chain” for hobbyists and infrastructure professionals alike.
The numbers of Docker pulls are remarkable, 1 Million in 2014, 1 Billion in 2015 and as of October 2016 we reached 6 Billion. Docker has more than 50 projects and more than 1200 patches per month. Suddenly forced to look for help to cope with this amount of demand and activity they had to brainstorm around things like BDFL (Benevolent Dictator For Life), DCO (Developer Certificate of Origin, like a Contributor License Agreement), project ownership and received much needed assistance from the Linux Foundation.
He stressed the importance of separating the church (OSS project) and the state (employer). Also, they focus on interface stability and obsess over user and products, spinning out OSS components along the way; “users first, plumbing later”.
He mentioned the added OCI support in Docker 1.11 and integrating the SwarmKit orchestration layer into the engine itself with 1.12.
Previously, developing software using Docker on Mac was too cumbersome and deploying engines on AWS too complex. Therefore they built Docker for Mac which is already released and Docker for AWS which is in private beta at the moment. It is integrated into Docker Cloud and will provide AWS CloudFormation templates and access to the Docker Swarm in the system tray menu no the developer’s laptop. The team at Conductant (who built Aurora and was acquired by Docker) will make sure the system will be able to update the infrastructure in production by describing changes to the cluster declaratively.
Accompanied by cheers he then proceeded to open-source InfraKit live on stage.
Keynote 3: OpenSDS – An Industry-Wide Collaboration for SDS Management
Steven Tan, Chief Architect and Cameron Bahar, SVP and Global CTO of Huawei talked about their new OpenSDS initiative that tackles the ongoing issue of how to best build distributed storage in the container world. Bringing companies like Intel, Dell and hopefully others on board, they would like to provide a common abstraction for software defined storage. Huawei employs contributors to Hadoop, OpenStack and the Linux Kernel.
Currently, there are many options available: Flocker, Ceph, libstorage, Cinder, Manila, Swift, Huawei OceanStor and others. They all need to be integrated into service discovery, orchestration layer and scheduler. Some of them are well integrated into larger systems like Kubernetes, Docker, Mesos or OpenStack.
Focusing on Kubernetes, the common options are GlusterFS, Ceph or Flocker. They differ in terms of project phase, maturity and problem domain. Either they solve a problem too broad or too narrow. This is where OpenSDS comes into play. An OpenSDS Agent functions as an adapter for SDS, NAS, SAN, Cloud and block storage options and abstracts them all into a single storage pool.
Intel works on OpenSDS since three years, the announcement was just done, so there will hopefully be more to come. SDS is the biggest unsolved problem in the container world at the moment.
Keynote 4: Collaboration Beyond Code
Jilayne Lovejoy, Principal Open Source Counsel at ARM talked about risks of OSS usage in companies and announced the new OpenChain Project that can help them be compliant.
Many companies already went through the pains of figuring out how to deal with OSS. It makes sense to provide common guidelines to use as a starting ground. OpenChain is about protecting IP, license management, license compliance, confidentiality, copyrights and ownership. She works for the SPDX project at the Linux Foundation which introduced an automated way of adding license meta-data to source files to simplify compliance management. Prominent companies are members of the OpenChain initiative: Adobe, HP, Cisco, Siemens, Qualcomm, Arm.
On a side note, she is a trained lawyer and inserted a couple of entertaining jokes about it in her talk. I can’t help but wonder how exhausting it must be for lawyers to always have to be sarcastic and self-loathing when talking to non-lawyers, so sad.
OpenChain centers around these three points:
- Specification: How OSS oversight can and should be organized in a company
- Curriculum: How engineers and managers can be educated on the issues.
- Conformance: Certify companies for OSS compliance. First only with a self-conformance on the website (available soon?), later more formal
She summarized the main action points for becoming conformant:
- Employ a FOSS Policy and offer training
- Define a person responsible for FOSS and advertise them internally and externally as the FOSS contact
- Review & Approve FOSS content produced by employees
- Deliver FOSS documentation and artefacts (see SPDX)
- Understand FOSS community engagement (just dumping the code might not be enough)
- Certify adherence using the self-conformance quiz only
My take: The OpenChain project looks like a good starting point for OSS compliance for small and medium sized companies that do not have the resources to hire a bunch of copyright lawyers. They can benefit from more experienced entities that have been doing this for years.
Keynote 5: Blurring the Lines: The Continuum Between Containers and VMs
Graham Whaley, Senior Software Engineer at the Intel Open Technology Center talked about the binary choice between a container and a virtual machine. He would rather prefeer a continuum.
End-users like application developers might not usually care much about the details, it just needs to work. Containers are clearly more friendly than VMs, just because of ease of setup and startup time. But VMs do not have to be big. Projects like rkt can also start VMs using KVM with very little overhead and insanely fast.
Intel is working on a technology called Clear Containers. The mission is to make VM behave like containers, to make them “thin enough”. They currently achieve below 50 ms boot time. There is still around 50 Mb overhead per instance because of the kernel and user space files, but in a lot of cases this is good enough. As they progress, this footprint will likely decrease. They are now looking into the next 10fold improvement.
My take: Alongside other projects from the community like rkt, this is an exciting project coming from the big league. Definitely put this on the watch list.