2023.06.26 WEB+DB PRESS

Magazine WEB+DB PRESS Vol.135: Interview “How a Large On-Premise Environment was Migrated to the Cloud”, Seongwon Youn and Daisuke Nakano [ part 1 ]

※This article is machine translated.

It first appeared in a Japanese magazine, and after Cocone received permission, it was translated into English and published.
first appearance: WEB+DB PRESS Vol.135 ( Gijutsu-Hyoron Co., Ltd. )

Explore Cocone’s engineering capabilities to open up a new era of Web3

Top-level performance in the world’s favorite digital world

Part 3: How a Large On-Premise Environment was Migrated to the Cloud

In this issue, we take a closer look at Cocone’s infrastructure. Cocone has a mix of traditional on-premise and cloud environments. We asked Mr. Seongwon Youn and Mr. Daisuke Nakano of the Infrastructure Office about how they use the two different environments and how they have lifted and shifted to the cloud.

Magazine WEB+DB PRESS Vol.135: Interview "How a Large On-Premise Environment was Migrated to the Cloud", Seongwon Youn and Daisuke Nakano [ part 1 ]

– Please introduce yourself.

Seongwon Youn (Youn): My name is Seongwon Youn, General Manager of Cocone Development Division. I am also the head of the Infrastructure Office. I have a long experience in development and DBA (Database Administrator).

Daisuke Nakano (Nakano): My name is Daisuke Nakano, Deputy General Manager of the Infrastructure Office. I have experience in building and maintaining web-based infrastructure from my previous jobs, such as server racking and middleware tuning.

– What is the evolution of Cocone’s infrastructure environment?

Youn: In the past, Cocone mainly used an on-premise environment. This was because for large scale and stable services, the on-premise environment was less expensive than cloud computing of the same scale, and it was fine to keep the on-premise environment. However, the on-premise environment is expensive, especially in terms of initial costs and operational man-hours. Another drawback is the lack of flexibility in adding servers and changing specifications.

Nakano: Around 2017, we started using cloud rather than on-premise environment for newly released services. While we often chose AWS for services for Japan, Google Cloud had the advantage in verification for global deployment, so we tended to choose Google Cloud for services with a global perspective. Later, AWS added Global Accelerator and other improvements were made to make any cloud work globally.

Youn: Using both AWS and Google Cloud was also meant to avoid relying solely on either cloud vendor.

Lift large on-premises environments to the cloud

– Have you migrated any systems from on-premise to the cloud?

Youn: Although this is a group company and not Cocone itself, there is a case in which we migrated the operating environment for a new service to the cloud over the course of a year and a half. The main purpose was to solve the issues of operation system and cost. There were several hundred servers, and the scale of the project was quite large, so we proceeded in two stages. I remember making many trips to the site.

Nakano: In the first stage, we moved the web server and application server to the cloud, and left the database server on-premise (while moving the data center). Strictly speaking, MySQL was moved to the cloud (Amazon EC2), while the main Oracle Database remained on-premise. The reason is that we had packed a lot of things into a high-specification server, so we could not migrate immediately. The cloud and on-premise were connected by a dedicated line (AWS Direct Connect).

Youn: The second step was to migrate the remaining on-premise Oracle Database to Amazon RDS for Oracle. We took the time to analyze the contents of the database and carefully migrated it. Since the service had been running for a long time, a complete change would require man-hours and time on the development side. It was a choice and concentration. As a result, the database was reorganized and overall costs were cut in half. We also formed an infrastructure organization for the group companies.

– What were some of the challenges?

Nakano: In the hybrid state during the first stage of migration, even though the system was connected via leased lines, there were more delays than in the on-premise system. This was beyond the system’s tolerance and caused timeouts and other problems.

Youn: As we proceeded with the analysis, we found some loose ends in the logic for handling the database. On-premise, with high specs, a little waste was not a problem. We were then able to solve the problem by reducing the number of calls from the database and extending the time until timeout so as to reduce the delay.

Nakano: Compared to on-premise systems, which operate within the same facility, the cloud system has a distance even within the same availability zone, so there may be some latency.

Youn: The member who handled this matter received an award from Cocone’s in-house award system. Infrastructure is generally regarded as ” as a matter of course, ” but I am glad to see that infrastructure activities are also properly evaluated at Cocone.

Blurring the boundaries with the development team

– What is the current structure of the Infrastructure Office?

Nakano: When I joined the company, there were three of us, but now there are eight. In the past, it also served as an internal help desk, but that has been transferred to the General Affairs Department. On the other hand, we formed a new security team, and I am also a member of the security team.

Youn: In the past, the development side would make a request to the infrastructure with ” specifications and configurations ” and then work on it, but from now on, I think it would be good if the boundaries between the organizations are blurred. In fact, now that the number of infrastructure members has increased, the infrastructure team is joining the development team from the early stages of development and designing together. However, the number of infrastructure members is still very small, so they are still working concurrently.

Nakano: It would be good if the development team can understand the server configuration by having both parties work together on the design, and conversely, it is important for the infrastructure team to have some understanding of how the application works and how it operates in order to perform performance tuning.

– What does the security team do?

Youn: In the past, we have taken individual security measures, but now that Cocone’s organization has grown and we are also dealing with Web3 virtual currency and blockchain, we have decided to establish a new security organization.

Nakano: We are working to bring the company closer to DevSecOps, starting with strengthening internal security awareness, and then strengthening the vulnerability assessment system.

The next article will delve deeper into the features of Cocone’s infrastructure.