SK Planet Develops Cloud-Based Search System

SK Planet developed its own search system utilizing Elasticsearch and Kubernetes in an AWS environment to quickly provide search functionality for its services. This system operates as a SaaS and has been applied to major services such as OK Cashbag, Syrup, and V Coloring.

Architecture and Key Features

The search system is divided into AWS Service and Application layers. All applications run on EKS (Elastic Kubernetes Service), store data in S3, and process indexing and queries in the ES Cluster. A pipeline that delivers real-time search data via Kafka has also been built.

Key features include a unified search API, an autocomplete API, and a popular search terms API. In particular, through Dictionary management functionality for each service, operators can directly configure user dictionaries, synonym dictionaries, and correction dictionaries, which are periodically reflected in Elasticsearch. Data transmission supports two methods: batch (S3 upload) and real-time (Kafka).

Outlook for Integration with Generative AI

SK Planet analyzed that the search field has been significantly impacted since the release of ChatGPT. Noting the trend of major companies such as Google, Microsoft, Naver, and Kakao incorporating generative AI into their search services, SK Planet is currently examining the possibility of expanding its current keyword-search-centered system into conversational search and other forms. Going forward, the company plans to continue research aimed at providing new search experiences through such integration with AI technology.