Talks

Google Street View began with a single video camera hand held by Larry Page. It launched in 2007 and covered 5 cities. Since then it has grown to span 7 continents and over 50 countries with 5 million unique miles driven and 20 petabytes of imagery captured. Collections now include indoor, underwater, snowmobile, backpack and bicycle. The scaling up has been a logistics and a software challenge and we will cover some of the highlights, low lights, software architectures and lessons learned in Street View and Google at large.

Craig is Senior Software engineer on the Street View team at Google. After finishing a PhD in control theory (with a fair dose of CS) he spent some time developing wireless vehicle to vehicle communication applications for Mercedes-Benz Research and Development. In 2009 he joined Street View and started work on estimating panorama location and orientation using onboard sensors (GPS, IMU, laser, wheels), computer vision (feature matches, loop closing), map matching and other hackery. He enjoys the fact that it's a multi-dimensional project that is both enabled and challenged by its scale. When he is not enjoying it, he goes running.

The video for this talk is available on YouTube

Simon and Ben will discuss their experience building Vumi, an open source mobile messaging platform that's built for scale. The talk will explore how to create an environment where team members can effectively build and manage a complex software product, while existing in a challenging client-driven services organization.

Simon enjoys exploring details and finding ways to make the world a better place. From a childhood filled with weekends on the rocky beaches of South Africa’s west coast and fortuitous early access to the GW-BASIC reference manual, he made his way through high school and into academia, where he spent a lot of time learning details like quantum mechanics and general relativity. Somehow he completed an MSc in Applied Mathematics, despite discovering the wonders of Linux and open-source software in the middle. Since leaving academia, Simon has written code for a variety of commercial and scientific projects ranging from online newspapers to petrochemical companies and from bioinformatics to radio astronomy. In his spare time Simon contributes to and maintains a number of open source software projects, writes the odd computer game or two, engages in old-fashioned pen-and-paper gaming and tries to get out to the swimming pool or squash court every now and then. Simon is currently the lead developer of Vumi, a scalable mobile messaging platform developed by the Praekelt Foundation and recently organized the first two PyConZA conferences.

Simon and Ben will discuss their experience building Vumi, an open source mobile messaging platform that's built for scale. The talk will explore how to create an environment where team members can effectively build and manage a complex software product, while existing in a challenging client-driven services organization.

Ben creates environments that enable talented people to build cool stuff. Recently back in South Africa after a 14-year stint in the States, Ben now runs product for Vumi, an open source mobile messaging platform, owned by the Praekelt Foundation. When not on IRC, Ben's typically trying to keep up with his two (soon to be three!) kids.

Caching has been a 'hot' topic for a few years. But caching takes more than merely taking data and putting it in a cache : the right caching techniques can improve performance and reduce load significantly. But we'll also look at some major pitfalls, showing that caching the wrong way can bring down your site. If you're looking for a clear explanation about various caching techniques and tools like Memcached, Nginx and Varnish, as well as ways to deploy them in an efficient way, this talk is for you.

Wim has been working with open source technologies since 1997 and has been involved in open source projects such as OpenX and PHPCompatibility. Next to web development, he's worked with a wide range of technologies (from database clusters to Internet backbone design) and focuses a lot of his time on high-scalability projects and on coaching/training PHP engineers.

The video for this talk is available on YouTube

Performance improvements in modern browsers coupled with the acceptance of javascript as a first class language gives us a compelling platform to move a bulk of what used to be done earlier on the server side to the client. This allows the distribution of processing power and removing some obvious bottlenecks on the server side. This talk aims to share our experience with the advantages of using this approach along with its limitations. We talk about single page web applications, client side consumption of RESTful services, performance gotchas, single code base for designers and developers and its effect on building a portable client and scalable server code.

Zabil works as an Agile consultant at ThoughtWorks with a focus on using the right set of technologies and tools as the enabler for successful delivery of large-scale projects. He was also a part of Thoughtworks Social Impact programme where he contributed to many open source projects like RapidFTR and OpenMRS. His career as a consultant spans 15 years on various technologies and processes-related web development, Enterprise systems and Agile practices. He is currently based in the ThoughtWorks Johannesburg office. When he is not a self-confessed lazy programmer, he is an avid runner and climber.

Performance improvements in modern browsers coupled with the acceptance of javascript as a first class language gives us a compelling platform to move a bulk of what used to be done earlier on the server side to the client. This allows the distribution of processing power and removing some obvious bottlenecks on the server side. This talk aims to share our experience with the advantages of using this approach along with its limitations. We talk about single page web applications, client side consumption of RESTful services, performance gotchas, single code base for designers and developers and its effect on building a portable client and scalable server code.

Simba is a software developer at ThoughtWorks, based in the Johannesburg office. He has contributed to educational software in South Africa that visualises Matric results across the country. With 7 months experience in software development, Simba is keen to learn and enjoys working with technology that makes a difference in society. He is also an avid gamer with a thirst for general knowledge.

We can make servers look alike with Puppet, Chef, Ansible and most probably other machinery that we are going to invent. And it makes us happy, because we like conformity and consistency. Life is easier like that, we believe or hope. Yet, at ScaleConf 2013, I noticed that every company that told their story increases people proportional to increases in servers. How did that work out? Did we make all people the same - like our servers? Do we want that in the first instance? And if we don't want clones, what do we want? How do we scale a culture if value diversity? Which emerges first - consistency in values, or consistency in technology? What is the tail that is wagging a body? I will explore these and other challenges - the really tough stuff that continuous delivery pipelines don't solve. I hope you will help me answer those questions? I believe that when we understand this, we can achieve scale and balance. That's what I'm after - equilibrium at scale.

Aslam is an African software developer, by birth, by choice and for life. He believes that all of software is just design and values code as an expression of design. The same goes for architecture too - it''s all design. The one thing he has learnt, though, is that design is hard, very, very hard, because achieving a harmonious balance is difficult. The same with scalability - it throws us into extreme states, way off balance. That is the challenge - finding a new equilibrium. You can read his blog at http://f3yourmind.net

The video for this talk is available on YouTube

Sending notifications to millions of devices is quite challenging. Using native push notification services comes with the burdens of different APIs per platform, the need to manage the individual device registrations, the complexity of message personalization, and the need to massively scale when sending millions of notifications simultaneously. Notification hubs take a different approach: they're based on pub/sub and designed to elastically scale. They provide message templates for personalization and a single API across iOS, Android and Windows.

Beat is part of Microsoft's technical evangelism and development team (ted), where he architects, designs, codes and delivers leading edge apps and services to showcase the Microsoft platform, with a special focus on cloud. He started to focus on cloud computing in 2008 and just loves the opportunities it provides to businesses of all sizes. Before joining Microsoft, he was an independent architect and developer and was involved in a wide variety of projects, ranging from real-time building control systems, best-selling shrink-wrapped products to large scale CRM and ERP systems.

The video for this talk is available on YouTube

For Facebook’s 10 year anniversary, we produced over 700 million personalized “look back” videos. Without any marketing or promotions within Facebook, these videos were shared by hundreds of millions of people in the first two days. To make this happen, we scanned through billions of years of timelines and pre-rendered petabytes of videos. I’ll talk about how the videos were rendered and the systems used to scale it, how we managed to do this in just 3 weeks, including a few days for the infrastructure just to compute and store the output, and how Facebook's company culture enabled a group of relative strangers to band together and iterate quickly on solutions for this frankly herculean project.

Chris is a Production Engineer at Facebook. He works with large scale software deployments running on, amongst other things, CentOS, Chef, Python, designer drip coffee, Ruby, RedBull, vi, bash, cable ties and duct tape, using a large infrastructure of Facebook's OpenCompute hardware.

A look at the elements required by Elasticsearch to turn a simple inverted index into an auto-clustering, horizontally scalable real time search and analytics engine. The talk will start from first principles, explaining how an inverted index works, how to make an inverted index suitable for real time search, how to scale that out, how to add reliability and failover to the cluster, and design patterns to use the cluster most efficiently

Clinton was the first user of Elasticsearch and wrote the Perl API back in 2010. When Elasticsearch formed a company in 2012, he joined as a consultant and the maintainer of the Perl modules. Now Clinton spends a lot of his time designing the user interfaces and writing about Elasticsearch, and is currently working on a book about it with O'Reilly Publishers. He studied medicine at UCT in Cape Town and now lives in Barcelona.

The video for this talk is available on YouTube

Yuppiechef delights people daily by delivering stunning products, handwritten cards and quality service. This is an architectural look at where Yuppiechef has come from and how we're dealing with rapid growth while still keeping our top-notch service levels. We'll explore some of our experiences and highlight a couple of key pieces that are helping us to scale with very limited resources.

Deon is a senior software developer at Yuppiechef.co.za, the leading e-commerce store in South Africa. He is a technology enthusiast, loves programming concepts and languages and has spent 14 years in development on websites, mobile applications, payment switches, warehouse systems, desktop software and development tooling. Occasionally spotted with an Arduino in the one hand and soldering iron in the other.

The video for this talk is available on YouTube