Content System — infrastructure used to prepare billions of banners for selection in the ad engine and subsequent display. Statistics — these are services designed to collect and aggregate statistics on the interaction of millions of users with billions of banners. Both systems work with huge data streams on thousands of servers, using a BigRT core with exactly once guarantees. Statistics (Rtstat, RTABT) — a data showcase for advertisers and partners, an infrastructure for rapid experiments and analysts. Content System (Caesar, Plutonium) — we are building the infrastructure of the future for getting banners into selection within 30 minutes of their creation. Both projects are characterized by high loads: we process streams measurable in gigabytes per second and messages in millions per second.

Our stack:

• Main code in C++ * Python for infrastructure and tools * YT for storage and MapReduce processing * BigRT for data stream processing

What tasks await you

Development and support of the data stream processing platform Both in Statistics and the Content System, BigRT and YT are used for stream processing of events and for storing the results. The systems process gigabytes of data and millions of messages every second. That's quite a lot — accordingly, we have tasks for optimizing the hardware resources consumed by the systems. We also have many tasks for migrating old processes to the new architecture: almost everything created on MapReduce we are moving to BigRT.

Testing, acceptance, and operation of our systems We are infrastructure teams, and we test and release all changes ourselves. This motivates us to improve our release and acceptance processes. And we also constantly invest time in increasing the stability of our systems.

We expect you to

• Have a good knowledge of C++ and STL
• Know classical algorithms and data structures
• Have experience in industrial development
• Understand the key aspects of building distributed systems
• Be familiar with Linux