"Commencis Cloud solutions helped us to build robust, highly-available and cost-optimized hybrid big data platform on AWS with minimum effort. Thanks to this project, our Kubernetes workload performs and scales smoothly. Also, IT efforts have been dramatically reduced with the help of infrastructure-as-code approach." Hakan Türköner - Cloud And Infrastructure Manager
About Ay Marka Retailing
Ay Marka Retailing, is one of Turkey’s leading textile retail companies with its famous and exclusive retail brands. They are a pioneer in the Turkish retail industry, paving the way for the latest fashion trends from around the world.
Through their brands, which meet the style expectations of a variety of consumers, they continue to target consumers’ needs and wants at every stage from the range of products to the after-sales services at both physical and online stores.
Ay Marka Retailing gives direction to Turkey’s brand apparel and accessories industry through brands such as Network, Que, Divarese and George Hogg.
The Challenge
Ay Marka’s DataCollection project is a hybrid streaming data architecture that collects events from several channels to a central location and those events on several targets.
The aim was to understand their customer behaviors, to make them meaningful with the help of analytical reports and to provide Ay Marka’s customers a better overall shopping experience.
Output
Ay Marka Project is based on microservices that runs on Kubernetes cluster, the popular open-source container management system. From the project perspective, agility and flexibility were top priorities. With the help of EKS & EC2 services, the team adopted these key priorities successfully to the project. The usage of EKS API made it easier to manage the containers with common kubernetes commands. At the same time, Elastic Load Balancing and Auto Scaling helped the application scale up/down when it was needed. As one of the most high-traffic e-commerce websites in Turkey, it is very important to have this ability during special days or campaign periods. Tens of millions customer events are expected to be processed and stored in a single day, so the system needs to be ready for such high loads at any time.
Also, Amazon Elastic Container Registry (Amazon ECR) is used to store and manage the project’s container images.
As mentioned before, DataCollection project persists the data on several targets and one of them is Simple Storage Service (S3). All the event data is replicated on S3 buckets while persisting on PostgreSQL at the same time. After a predefined threshold, all the data in S3 is moved to AWS’ archival service, Glacier, to minimize the storage cost.
For relational database, Ay Marka uses a managed PostgreSQL RDS instance. All the customer events are stored in this database server and with the help of RDS, data is kept highly available and secure in the production environment. It’s crucial for the project to keep these customer events available at all times. The project also uses Amazon Simple Queue Service to create communication between distributed microservices. It is used to queue the events which will later be processed by the consumer services. This helps to throttle sudden peaks in the application. In terms of hybrid architecture, Ay Marka connects its AWS VPC to its on-premise data center via AWS site-to-site VPN connection. After establishing this VPN connection, the application moves the transformed customer events to an on-premise data warehouse server. After that, BI users run various reports to make sense of this collected customer data which is generated from several channels.
The Benefits
With the help of managed AWS services, Ay Marka created a robust hybrid-cloud architecture which is highly available, scalable, fault-tolerant and secure. Duration of new deployments and the cost of the compute resources have been significantly reduced thanks to this project. Ay Marka will continue to improve their customers’ user experience and reduce the costs and efforts of its IT resources.