AWS Case Study: Interpark

Interpark is Korea’s first online shopping mall service opened in 1996. Based on the years of experience, technical expertise and strong brand power, the company became one of the  e-commerce leaders in the country. Interpark provides a one-stop online shopping experience that offers a variety of goods and services from books to cultural event tickets to tour products to serve the changing life styles of customers. Its mission is to create a next-generation e-commerce environment that integrates IT technologies like AI and big data analytics with smart shopping experience.

Interpark wanted to manage book, shopping, ticket, and tour product data in one place to collect and analyze big data and provide  services and benefits personalized to each customer. It used to have an open-source commercial Hadoop platform on-premises for big data analytics. However, it was looking for other solutions because of the high license costs and operational issues. The high costs of licenses were preventing the company from using all of the servers in the IDC, making it almost impossible to separate development, staging and production environments. This was the reason why they were suffering development and operational issues. In addition, Interpark did not have engineers who could build and run a big data platform thereby had to seek for tech support outside whenever they had an issue, so it took a long time to fix the issues. Sungyoon Lee, head of the data platform team at Interpark, says, “We did not have any in-house big data platform experts, so we had to find support outside whenever our platform failed, and it took long time to fix the failures. One time, we even had to shut down our development and production environments because of severe sever failures. This was a big problem because we are an e-commerce business that operates 24/7.”

Interpark compared the costs and performance of on-premises commercial Hadoop and Amazon EMR and decided to move to AWS that had years of cloud computing experience and a strong ecosystem of APN partners. 

Interpark began to migrate to Amazon EMR while still operating some services on on-premises Hadoop. This actually allowed the company to compare the performance differences between the existing platform and Amazon EMR, run an integration test of Amazon EMR with the production environment, and verify potential performance and security improvements. AWS and GS Neotek supported Interpark throughout the migration process from architecture design to operation of the service. The company completed the migration of all company data from the IDC servers to Amazon EMR in December, 2020. Now all of its data is processed by Amazon EMR. Before, the resources and storage were interlocked on the old on-premises Hadoop, making recoveries taking longer and leaving the resources and storage unfunctional until the recoveries are  completed. Also the high costs of Hadoop was another problem because the company had to identify the maximum resources available for a scheduled job prior to creating any cluster, not to mention that it had to rely on a vendor to build a big data platform. On the other hand, Amazon EMR separates resources from storage, enabling faster recoveries. In addition, its scale in/out function provides steady performance while reducing costs at the same time.

“We separated development, staging, and production environments completely and by projects with Amazon EMR and built an environment that enabled higher operational reliability because it allowed us to avoid impacts on other services,” says Lee. “We also reduced and optimized costs with Auto Scaling and Amazon EC2 Spot Instances.” Interpark uses Apache Spark to aggregate big data, Amazon Simple Storage Service (Amazon S3) for data storage, and Amazon EMR as a data processing platform.     

Since the migration to Amazon EMR in December, 2020, Interpark started processing all of its data on Amazon EMR and saw zero server failures ever since. It also received support from solutions architects from AWS and GS Neotek on the use of AWS technologies. It now understands the technologies better and uses them with confidence. Furthermore, it successfully completed the migration in the scheduled timeframe by modernizing its architecture with the help of AWS consulting services.

Amazon EMR allowed the company to separate workloads at a project level, avoiding impacts on other services and enabling stable operation. Its flexibility also brought in 50% cost savings. The service separates compute from storage, scales them separately, and provides a single view of resources, allowing Interpark developers to focus on development and increase productivity. Lee says that because his team has built up the experience on Amazon EMR now, they can now address issues themselves. According to him, work productivity has increased dramatically because they could get support quickly from AWS and GS Neotek in case of a server failure.

Lee explains, “On our old platform, we cloud not separate our development and production environments. Moreover, we had to find someone who could help us outside of organization if there was any server problem, and all business operations stopped until the servers were recovered. There was also a growing need for big data analytics. All things combined, we needed a solution that could address this issues for us. Amazon EMR provided all of the things we needed, so there was no reason not to use Amazon EMR.”

With the successful migration to Amazon EMR, Interpark now plans to further optimize costs and resources. It also plans to build data lakes on AWS to analyze data more effectively and create business insights.