A Short History of the Future - Part IIPosted on Thursday, July 30, 2015 in Blog By David Rusling
In the last episode, I talked about how Linaro got started and how the mobile market matured. In this blog I write about how Linaro evolved.
Start Ups are Fun
Having worked for ARM as it grew from startup to established business, I wanted to be part of a startup and, boy, has it been a wild ride. Linaro started 5 years ago with just 6 members and an initial staff of around 50 to more than 30 companies and 200 staff. We have had the usual challenges along the way, but we have stayed true to our founding principles. First, the ARM partnership needs a place to collaborate, a place that is outside the firewalls of Linaro member companies. That collaboration is not just talking about things (especially not talking about things in secret), it is about engineering. Secondly, that open source and open standards are the key to everyone’s business moving forward.
Along the way, I have learned that simplest is always the best. Asking “How can we do this in a simpler way?” is a very good habit. Like all companies, we have learnt from our mistakes. We have a unique problem in that we are running open source development sponsored by our members. We want our members to have visibility into what we are doing, the ability to dig into details and, as a result of this, our engineering teams have grown. At any one time we have over 100 epics (big things) that we’re working on collectively. This is really hard to visualise let alone manage. Realising this, we are moving to a more mission focused “what we are trying to achieve” approach. We call this ‘Lead Projects’ but my nickname is ‘Back to the Future’ as it is really about member and community involvement around key areas that we want to focus on. One example is open source security.
Early in 2012 Linaro was approached by a couple of companies, one an equipment manufacturer and the other a distribution provider. Having seen what Linaro was doing elsewhere, they asked if Linaro could be used as a way of working together on the software needed in the enterprise market. This led to the formation of the Linaro Enterprise Group (LEG) in late 2012. We used the same model of a technical steering committee directly driving the work that was needed. Whilst the Linaro TSC ‘owns’ the central engineering teams and supports segment groups, each segment group runs their own affairs. This ‘franchising’ of Linaro has helped us scale (and focus) our efforts and we now have four segment groups, Enterprise, Home, Networking and Mobile and we are thinking about launching a fifth group around embedded / IoT.
The Linaro Enterprise Group, led by Andrea Gallo has set a high standard for Linaro groups. It has worked steadily on foundation technology, including OpenJDK and UEFI / ACPI. The performance and stability of OpenJDK has improved dramatically and their UEFI work has helped establish ARM as a plausible player in data centers. LEG has also followed standards as they started to emerge - one of which is OpenStack, which fits with its focus on enabling the ARM architecture in big data. Along the way, LEG has done much to enable ARMv8 64 bit computing.
The Network Group, LNG, led by Raj Murali took a different approach. One of the values of ARM in this space is to create SoCs that integrate network acceleration technology. This means that ARM based networking equipment can be very different, which presents a programming challenge. Open Data Plane (ODP) was the solution. When we started Linaro, I was very wary of Linaro being a standards organisation but I am happy to say that this has been highly successful with ODP being widely adopted. The next phase is to ensure that adoption grows and is recognised by the various networking standards bodies.
In a similar way to the enterprise group, Linaro’s home group (LHG), led by Mark Gregotski started as a response to Comcast’s RDK, Comcast’s effort to standardise set top box architecture. Comcast realised Linaro’s potential for collaboration amongst ARM based set top box suppliers and requested that Linaro did some enablement work. Android TV interest is growing too and Linaro is looking at work here.
The last (at least for the moment) Linaro group has been the Linaro mobile group. Linaro when it was formed had a very mobile focus with most of the core work actually being for mobile. As groups have been created, the nature of that core engineering has changed and rather than lose focus on mobile we created a separate mobile group. We also wanted to create the dynamics that groups, with their mixture of OEMs and SoC providers. This gives us very informed discussions that quickly decide what areas are most important to work on together. LMG, run by Tom Gall, has become the largest group in Linaro and is blazing the trail for Lead Projects mostly staffed by extra engineers volunteered by members. These volunteer engineers are in addition to their assigned engineers. This group is actively working on the next versions of Android for mobile and on Project ARA.
Talking and Meeting
Linaro continues its tradition of 6 monthly engineering cycles punctuated by a face to face conference we call ‘Connect’. It is a good name, it is the place to connect with other members, other engineers and companies. The first ‘Connect’ that Linaro was at was an Ubuntu Developer Summit (UDS) in Belgium in May 2010. Actually, Linaro was not there - it was still in the process of forming. We liked the engineer driven discussions and so adopted the format and evolved it for our own use. The first real Connect (in Cambridge, England naturally) was around 120 people (including Russell King) and the next one (in San Francisco in September) is likely to be around 500 people. We are not quite at the dry ice and rock music stage, but these are serious events. I believe that companies do not ‘get’ Linaro until they attend a Connect. I am not entirely sure this is true, but I don’t believe that any company has joined Linaro without first going to a Connect. Despite the size of the event, the mornings are engineering driven, technical discussions and the afternoons are hacking sessions (also known as ‘making things work’) and committee meetings. Sadly, I’m in the committee meetings.
We also have regular teleconferences. Remember telephones with wires? We started that way but we now use Google Hangouts and BlueJeans as our main conferencing tools. Here I would really like to thank all those committee members who join from around the world, often at crazy times due to the timezones.
F.A.B Brains, F.A.B. The next episode looks at how embedded is evolving into IoT and what we should do about it.