Improved resource-related cost efficiency by about 50% through effective utilization of resources
With the deployment of the container platform running Red Hat OpenShift Service on AWS on Amazon EC2, resource efficiency has improved significantly.
“By using the Kubernetes API Horizontal Pod Autoscaler, we introduced a mechanism where new pods are automatically launched when resource usage reaches a certain threshold, and conversely, pods are removed when no longer needed,” said Kato. “This simultaneously resolved issues with scaling out resources and being unable to scale in, enabling us to respond more quickly and efficiently.”
Previously, it was necessary to always secure ample resources to handle spikes in demand, leading to significant infrastructure costs. However, with OpenShift effectively utilizing surplus resources, cost efficiency has improved by approximately 50%.
“For instance, in the past if there was ever a sudden need to secure resources for a campaign, it was impossible to respond unless resources were maximized in advance due to the time-consuming scaling,” said Watanabe. “But now, with the greatly improved scalability, we can respond quickly to requests like increasing resources for even just one day for a large-scale distribution.”
Significantly reduced operational burdens and eliminated time spent on recovery
Another major advantage of switching to Red Hat OpenShift Service on AWS was the improved operational ease it brought in various situations.
“We were setting up the integration between Amazon Cognito, which is the authentication infrastructure, and Red Hat OpenShift Service on AWS, and since we were not yet familiar with container platforms, there was some trial and error. However, thanks to the GUI in OpenShift, that trial and error never became a serious issue. By simply changing the values in the GUI, resources were automatically updated, allowing us to move forward without stress,” said Fuuki Yoshizawa, Core Staff of the Communication Platform Department, commenting about the ease of development.
Kato has also found recovery after operations to be easier, not just during the development stage.
“With OpenShift’s automatic container recovery, failures are far less likely to occur in the first place, and the system’s fault tolerance has further improved, which has greatly reduced the frequency of daily operational tasks,” said Kato. “Previously, when a failure caused the system to stop, we had to manually extract processing information from the logs and handle recovery. Now, however, the system is designed to automatically recover business processes in tandem with OpenShift’s automated recovery, so time-consuming manual recovery is no longer necessary.”
Deployment time reduced by 98.89%, development period shortened by about 20%
To accelerate deployment tasks, KDDI implemented Argo CD, a tool that supports the OpenShift GitOps Operator, and automated key tasks. As a result, deployment can now be completed with the push of a button in about two minutes.
“Until recently, deployment used to take around three hours, so this 98.89% reduction is dramatic,” said Kato. “The shorter deployment time not only improves the customer experience but also helps prevent human error and potential failures.”
The reduction in deployment time has also influenced releases. Watanabe notes that it has “changed our motivation”. “Previously, when commercial operations took hours, we were reluctant to release frequently, since it felt inefficient. Now that the work can be completed in just a few minutes, the mindset has shifted toward carrying out releases proactively. This has accelerated the feedback loop, speeding up the cycle of improvements and fixes, and ultimately improving quality,” said Watanabe.
Currently, verification releases are carried out about once a week, while commercial releases have increased from once every three months to about once a month.
In addition, with the adoption of Red Hat OpenShift Service on AWS, KDDI shifted from deploying Java EE–based applications on JBoss EAP to using Quarkus, a lightweight Java framework. The Quarkus libraries have also made it easier to introduce Apache Kafka to promote loose coupling through asynchronous processing.
“With the adoption of Quarkus, the introduction of Apache Kafka feels like it has become smoother. By leveraging Quarkus libraries, we were able to improve development efficiency and shorten development periods by around two months, which is about 20% less compared to before,” said Kato.
Improved system visibility for developers with a new development framework
KDDI also undertook a complete overhaul of its monitoring and observability framework, moving forward with planned development that included metrics and log design.
“With Red Hat’s proposals and support, we built a monitoring platform on OpenShift by integrating Prometheus and Grafana. Previously, our focus was mainly on tracking overall data throughput, but with this new platform we can now visualize the information needed for operations in greater detail and with greater efficiency, such as usage conditions and performance indicators for each system and service. By leveraging this, we believe we can optimize operations, enhance service quality, and support future business growth,” said Watanabe.
In this project, Red Hat provided comprehensive proposals and support not only for Red Hat OpenShift Service on AWS, but also for middleware and the peripheral tools.
“It was reassuring to have Red Hat’s experts supporting us with their extensive container experience. Red Hat started by asking, ‘What kind of future do you envision?’ They were there with us from the environment design phase, even before basic development, sharing our ideals and goals,” said Watanabe.
“We have been able to build a healthy development framework,” said Kato. “Red Hat does not insist solely on its own products but constantly proposes whatever might work to achieve our ideal system. That type of consistency is pretty rare.”
