Architectures You've Always Wondered about

Have you ever browsed to a site like Facebook or Amazon and wondered what sort of software architecture they may have used, or what insights their teams must have after solving such complex and large-scale problems? This track will give you the opportunity to learn directly from some of the most well-known and high-volume web applications in the world. In previous QCon conferences, this track has featured presentations by Twitter, eBay, Amazon, Facebook, LinkedIn, and Netflix.

Day of week:

Wednesday

Location:

Grand Ballroom - Salon D

Host:

Peter Bell

Presentations

Decomposing Twitter: Adventures in Service-Oriented Architecture

5:10pm - 6:00pm

By: Jeremy Cloud

Hundreds of millions of users world-wide, billions of writes and trillions of reads per day -- Twitter operates at a scale that very few software systems ever achieve, and it must do so within the context of 'now'. Near real-time delivery of the relentless stream of tweets is one of Twitters highest architectural priorities, with tweets from Presidents, Popes, celebrities, scientists, and maybe even your mom, all being delivered to millions of followers within seconds. This talk will delve into the service-oriented-architecture we have developed at Twitter, some approaches we take for maintaining high levels of concurrency, and briefly touch on some functional design patterns we employee to manage code complexity in a rapidly evolving system.
Dropbox Photos - Building abstractions on top of the file system

1:40pm - 2:30pm

By: Ramesh Balakrishnan

"Dropbox is moving away from a file-folder centric view of user data to a content-centric view. This talk will focus on a couple of the abstractions we've built on top of the file system - metadata indexing and virtual collections - that enable rich interactions with the content type in question. Photos is the first content vertical to take this approach - abstracting away the file-folder view from the user but, underneath the covers, staying constantly in sync with it. We'll go into some of the technical details of the general-purpose infrastructure we've built in this regard and the product experiences they enable."
Facebook Messages: Backup & Replication Systems on HBase

4:00am - 4:50am

By: Nicolas Spiegelberg

Facebook Messages is built on top of HBase, an open source distributed database system under active development.Facebook focuses on moving fast and being bold. To empower this philosophy, Facebook infrastructure needs robust solutions so developer mistakes, power failure, or hardware upgrades minimally affect user experience. We have built robust backup & replication systems for HBase that allow us to transition access to petabytes of user data in minutes. This presentation will talk about the various systems that have been designed for this goal, how they have iteratively evolved as the product matured, and algorithmic scaling challenges to handle the 500TB of data added toFacebook Messagesper month.
Lessons from Building and Scaling LinkedIn

10:20pm - 11:10pm

By: Jay Kreps

As LinkedIn has grown from a small application to a network of 200 million professionals the underlying technology has grown too. This talk will cover the evolution of LinkedIn's architecture and a few of the key lessons we have learned. This includes how we grew our technology stack from a monolithic application to a distributed set of services, how we moved from a single centralized database to distributed data stores, and how we scaled our development practices to support hundreds of engineers.
Making the internet a better place: Scaling AppNexus

2:50pm - 3:40pm

By: Mike Nolet

Our mission is to make the internet a better place. We help content producers like news, social, and gaming sites monetize better, and we help users discover new products and services. In order to achieve this goal, our entire business depends on data, reliability, and speed. The past three years weve experienced 'hold on for dear life!' growth, and this presentation will give an uncensored dive into our architecture and lessons weve learned along the way. In particular, well dig into the buy vs build connundrum, with a focus on: Our data pipeline and our transition from out of the box to the horizontally scalable Hadoop; our load balancer-free DNS architecture built in GSLB and Keepalived; our real-time data streaming infrastructure, built primarily in C and processing 30 terabytes of data every day.
Architectures Unconference

11:30am - 12:40pm

By: Floyd Marinescu

Meeting each other in the context of conversations we care about is why at QCon we are giving prime time to the unconference format; it creates a space for peer-to-peer learning, collaboration and creativity.

First we’ll father in a circle to be guided through creating an agenda using open space technology. Then there will be three three 20 minute time slots, each with 5 concurrent break out discussions that were designed by you.

Attendees often know more than the speakers! Bring your questions, ideas - be prepared to be surprised.