В MWS Cloud запустили платформу хранения больших данных для обучения ИИ

The platform allows you to store and process data such as sales and order information, demand for goods and logistics, information about payments and services provided. Tools for training and inference of ML and large language models can be launched on top of the platform.

MTS Web Services emphasized:

The platform allows for 40% more efficient use of storage by eliminating data duplication between different systems, increases the efficiency of maintenance personnel by 2.5 times due to centralization and automation of administration functions, and reduces the calculation time of analytical data marts by 2 times.

Various solutions can be used as infrastructure for deploying the platform, including Kubernetes, object S3 compatible storage. The platform also includes tools for centralizing access control, auditing and encryption, and dynamic masking of sensitive data.

The platform is compatible with Greenplum and Postgres, and supports a wide range of data, including open formats Apache Parquet and Iceberg.

Service administration is carried out through a single interface, allowing you to centrally manage users, clusters and scaling. Dynamic change of computing cluster resources is supported, ensuring continuity of operation and no downtime during scaling.