Case Study: Migrating Data Centers

Gulliver is a leading Japanese used car distribution business. The Company sells used cars at showrooms & travel assessment centers while cars that are not sold after 7 to 14 days are distributed through auctions. The fast growing business owned 420 showrooms worldwide in 2013 and is expected to increase this number to 1600 showrooms by 2018 (800 in Japan and 800 in overseas). Gulliver was founded in 1994 and operates both in Japan and USA.

As a fast growing business, Gulliver had been held down by the existing inefficient on-premises infrastructure. The system suffered from low performance and high costs, some key servers even reached max capacity and broke down.

Gulliver wanted to migrate both its internal and external systems to the cloud for better performance, high availability, and accessibility while saving maintenance and operational costs. The migration had to be fast, reliable (with zero downtime), and within budget. The high number of applications which were formerly ran on 22 sub-systems with over 110 servers required a repeatable assessment & migration process to a cloud environment capable of storing massive data sizes.

With the help of FPT Software, an Advanced Consulting Partner with the AWS Partner Network (APN), Gulliver considered the advantages of AWS and chose AWS as the most suitable infrastructure solution. The excellent AWS network infrastructure allowed Gulliver to connect its customer datacenter to the AWS datacenter with AWS Direct Connect. The openness and flexibility of the Amazon Cloud allowed the company to find a tailored solution to its complex IT challenges. Gulliver already operated in the AWS environment before and was happy to enhance its AWS integration.

FPT Software, which was responsible for environment assessment, technical design, application and data migration, cloud deployment and maintenance, approached the solution in 4 steps:

In Step1 the current system that was hosted at the Gulliver’s data center was investigated, finding an application framework using various technologies like .NET, Java and PHP that ran on multiple Microsoft, SunOS and Linux operating systems. The complex database and data system included 22 sub-systems with over 110 servers on premises.

Continuing the process, in the 2nd Step, Gulliver’s various applications were analyzed and divided into 3 categories:

• ‘Easy to Move’ – applications, like internal only apps or legacy applications that had little integration and could be moved quickly to virtual machines with minor reconfiguration were re-hosted in the AWS environment.

• ‘Web Applications’ that could be reconfigured to leverage Auto Scaling, Amazon Simple Storage Service (S3) or Amazon Simple Queue Service (SQS) with little effort were able to be re-factored by restructuring existing computer codes.

• ‘Optimize for Cloud’ – applications were complex apps that needed to be updated & standardized before they were migrated on the AWS platform & Amazon Relational Database Service (RDS) for better performance and cost-efficiency.

During step 3 Gulliver & FPT Software designed a comprehensive application and an intelligent data migration process based on its rich experience. In this process all stages from infrastructure provisioning to optimization were covered using best practices of cloud platforms. Various AWS storage types including Amazon RDS and Amazon Elastic Block Store (EBS) were used to assure flexibility and reliability in the new cloud environment. Gulliver uses Amazon Elastic Load Balancing (ELB) for better performance and higher fault tolerance in its applications.

In step 4 an own failover clustering mechanism was designed allowing the system to operate on a back-up server in case of failure or downtime. FPT developed a data migration script and manually re-configured all subsystems for streamlined integration without re-factoring the application code.

Although the project is still ongoing, Gulliver has started to see some initial success. With the AWS’ infrastructure, the new system runs well with no downtime and its performance has been increased significantly. The problem of critical on-premises overcapacity has been solved while maintenance and infrastructure costs have been considerably reduced. The new cloud-based infrastructure is now able to grow with the Gulliver enterprise with remarkable scalability.