Where in the spring to talk about Apache Ignite, distributed databases and distributed machine learning

At the mitap, dedicated to the birthday of Open Data Science, Yuri Babak, Head of the development team at GridGain, will tell about distributed TensorFlow with Apache Ignite:

- Apache Ignite as a distributed data source for TensorFlow;
- Distributed training of the model on the TensorFlow cluster over Apache Ignite;
- Inferens TensorFlow models on Apache Ignite cluster
')
Program and Registration

Cluster replication allows you not to worry about the performance of distributed systems. But what to do if the light in the data center is turned off, all the clusters will immediately start to fail (and each in its own way), and then someone will make a very serious mistake?

Ivan Rakov, Senior Software Engineer in GridGain, will tell you what replication options are possible, how transactional integrity is implemented in GridGain, and how it protects the cluster from disasters.

Program and Registration

Dmitry Govorukhin, Senior Software Engineer at GridGain, will tell you about personal experience in developing disk intensive applications and the rakes collected during the development of Apache Ignite Native Store.

Program Committee Commentary:

Reports on efficient file handling in Java are quite rare, which is surprising for such basic functionality. This report fills this gap, immediately concentrating on the practical details of achieving the desired performance.

Read more about the report.

Ivan Rakov - Disasters are no longer terrible: how we did asynchronous transactional replication in GridGain

When your business depends on the life of the cluster, it is good to have a spare just in case - besides the option of switching in case of an accident, you can read metrics and analytics on it. In the case when your application code uses transactions to maintain the necessary invariants, it will be even better if the replica cluster maintains transactional integrity. Read more

Stanislav Lukyanov - We support the development of several versions of the product in Git

We studied and tried different approaches, understood what works for us and what does not. As a result, we have formed our own Git process, which allows us to support dozens of previous versions, avoid regressions and compatibility issues, and remain flexible in the release of maintenance- and hot fix-releases. Read more