Talks

A fast (typically under a microsecond) and exact (Java Memory Model) distributed key/value store.

Azul made a really strong Hardware Transactional Memory, and yet in practice it was not good enough to run Java synchronized blocks concurrently — and Cliff will tell you why it didn't work and why HTM/STM in general will not allow concurrent execution in practice.

This talk's main purpose is to introduce protocols of distributed transactions invented in the last few years, which can be implemented on client side on any cloud storage supporting compare-and-set.

Alexey will tell about problems of adapting classic machine learning algorithms for distributed execution from his experience of working with Apache Spark ML, Apache Mahout, Apache Flink ML and creating Apache Ignite ML.

We revisit the underlying theory behind Paxos, weakening its original requirements and generalising the algorithm.

Nikita presents Lin-Check, a tool for testing concurrent data structures in JVM-based languages.

This talk surveys the theory and practice of blockchain-based distributed systems from the point of view of classical distributed computing, and how treating blockchains as a distributed computing problem simplifies everything.

How java.util.concurrent.SynchronousQueue and java.util.concurrent.Exchanger work, and why you should care if you write multithreaded code.

Attendees will be expected to watch videos of Leslie's talks "Programming Should Be More Than Coding" and "If You're Not Writing a Program, Don't Use a Programming Language", and Leslie will answer questions related to the topics of these talks. Each attendee should prepare at least one question.

We'll discuss several tricks we use at OK for working with big graphs.

We discuss Witness Replicas, a replication scheme used in Spanner and Megastore, and Apache Cassandra implementation of this concept, called Transient Replication and Cheap Quorums.

We'll explore transaction planning algorithm which Yandex Database (YDB) transactional system is based on. We'll see which entities participate in transactions, who sets global order for the transactions, and how transaction atomicity, reliability and high-level isolation are achieved.

For providing fault tolerance of Yandex Database, we use our own algorithm of achieving distributed consensus as well as several technical solutions explored in this talk.

Learn how to make a wait-free data structure.

An update on the latest research on Conflict-free Replicated Data Types (CRDTs).

Ori will introduce the formal underpinning of the C/C++ concurrency model from 2011 and the key ideas behind it. He will discuss some of the flaws of the model, ways of correcting it, and some remaining open problems.

Dmitry will look inside of the Go scheduler and expose how magic of goroutines is implemented.

We'll tell the story of the path that we went through when designing concurrency libraries for Kotlin programming language.