Auchan started the platform for work with big data in a cloud of Mail.ru Cloud Solutions

On November 17, 2020, it became known that ASHAN deployed an analytical platform for Big Data in the public облакеMail.ru Cloud Solutions. The solution will be used both for training artificial intelligence and ML models, and for ad-hoc analytics for product and functional teams.

Based on the Big Data cloud platform, product AI solutions will be developed and industrialized, including services for supply chain management, pricing and loyalty programs. Data laboratories will also be deployed in the Big Data Platform ecosystem, in which product and functional teams will conduct research and ad-hoc analytics.

Previously, the retailer used the same database for analytics, training ML models and supporting business processes. The high system load that occurs when working with analytical projects has negatively affected the performance of other services. To launch a full-fledged data platform and handle more complex tasks from the Data Science area, the company was looking for a partner who offers the right components as a service in the cloud. In order to quickly scale and consolidate all analytical areas on a single platform, ASHAN decided to launch Big Data Platform in the cloud of Mail.ru Cloud Solutions.

When choosing a supplier, ASHAN was guided by two important criteria: the availability of all the necessary platform components and the optimal value for money. In addition, it was necessary to ensure compliance with 152-FZ on personal data.

ASHAN launched a pilot project in which the Mail.ru Cloud Solutions platform demonstrated compliance with company requirements, performance expectations and deployment speed. The Big Data Platform was based on the Cloud Big Data service with Apache Hadoop, which transferred 25 terabytes of raw data needed to build AI data models. Building data storefronts and ETLs between different Big Data Platform components was implemented using the Spark component. The managed database is ClickHouse used to store hot data, ad-hoc analytics, and data storefronts for AI solutions. As a massive parallel system, the company connected Arenata DB based on the Greenplum DBMS. Its use will facilitate research for logistics, commercial management and marketing of ASHAN.

The company plans to train tens of thousands of ML models, so it turned to the Kubernetes technology from MCS. The automatic cluster scaling feature allows you to use hundreds of learning machines and automatically reduce power during interruptions in use.

On the MCS platform, the company built a big data pipeline, assembled for the task from the finished components of PaaS (Platform as a Service) as an analytical designer. The launch of the Big Data platform in the cloud of Mail.ru Cloud Solutions allowed ASHAN to scale quickly and easily to the needs of the business through the use of cluster technologies, parallel computing and Kubernetes auto-scaling.

We see that only the transition to the cloud can meet the demand of a fast-growing business to scale its IT infrastructure. Customers understand the value of a cloud platform where they get advanced technology by click. The business can focus on solving its problems, and not invest in the development of tools from scratch, - commented Ilya Letunov, head of the platform Mail.ru Cloud Solutions and Tarantool

Mail.ru Cloud Solutions met our internal requirements for technology stack, reliability, flexibility, and information protection. The demand forecast is our first solution that we migrated and deployed on our cloud Big Data Platform. We expect that the industrial use of this solution in the cloud will allow us to reduce inventory in stores by at least 5% and increase sales by 2% or more. In addition, using the Kubernetes-based micro-service architecture will allow us to effectively manage our resources in the cloud and optimize the costs of developing, implementing and supporting our AI solutions. - says Alexander Dorofeev, head of Big Data ASHAN Retail Russia