 Hi, today we're going to talk about accelerating hybrid cloud and multi-cloud deployments with PipeCD, a unified CD solution. I'm Kenta. I work for cyberagent.inc. I'm a maintainer of PipeCD. I love the option developer productivity. You can find me on GitHub or social media. The deploying applications to hybrid and multi-clouds has some challenges. Do we need different tools to deploy different kinds of applications? Do we need to learn the new tools? Do we how to keep consistency in delivery process in on-prem or public clouds? Ensuring security and observability across these varied platform adds another layer of complexity. Common questions we face include how to effectively store credentials and manage access control to various delivery tools alongside the strategic deployment in multiple clouds and environments. The third challenge is that how do we strategically deploy applications in multiple environments and multiple clouds? Here, PipeCD comes in. The vision of PipeCD is the one city for all. It is designed to overcome these challenges according to our experiences to deliver various kinds of applications to different clouds with many teams. PipeCD is a GitHub-style continuous delivery tool, which is particularly designed with a focus on handling multiple applications and providers. PipeCD scales in large organizations. PipeCD is an OSS project and joint CNSF sandbox in 2022. PipeCD supports a range of applications. Currently, we support Kubernetes, AWS ACS, Lambda, GCP Cloud Run, and Terraform. It also integrates with Kubernetes existence. If you use Customize, Helm, or Istio, you can migrate PipeCD without using any CRDs. PipeCD offers a unified and simple user interface across various applications. PipeCD has a unique architecture, which gives us a superpower to deploy a variety of teams and environments. There are two components, Control Plane and PipeD agent. Control Plane is a stateful component which store data in database and file store, while PipeD agent is a stateless. It's a single binary so that you can install easily to anywhere. Based on a company's experience, PipeCD is designed keeping platform and product teams in mind that ensure scalability and security. Platform team manages Control Plane, product team install a PipeD agent to their own clusters. Of course, there's no problem for one team to operate two components. Let's talk about how PipeCD manages credentials, deploying application in hybrid and mold clouds. To deploy applications to the cluster, you need to give credentials to a CD tool. With PipeCD, you give credentials to PipeD agent. PipeD agent use credentials to access Git and cluster API to deploy applications. The Control Plane can never see the credentials. It just provides console and scheduling each deployment. Only PipeD agent send a request to the Control Plane, so you don't have to set up firewall of application cluster for PipeD agent to accept inbound requests. Also, PipeCD supports multiple security features, such as building secret management, role-based access control, and a single sign-on. So let's move on to how PipeCD achieved progressive delivery. Here we say progressive delivery is the next step of continuous delivery. The automation to deploy new versions to a subset of user and they are evaluated in terms of correctness and performance before rolling them out to all of the users and rollback if not matching some key metrics. In a deployment, if nothing bad happens, it rolls out the new version to all cluster. If something unexpected happens, it rollbacks to the previous version immediately. Not only Kubernetes applications, you can do the same with other kinds of applications, such as Cloud Run on this slide. In analysis stage, PipeCD collects metrics from what we call analysis provider. Various providers, including Prometheus and Datalog, are supported. PipeCD can execute deployments to over applications one by one, ensuring each deployment finishes successfully. This is a deployment chain feature. What does this feature give benefits to us? For example, we can automate deployments from Dev to Prod. We can deploy applications to Dev at a first, then automatically deploy it to Prod. Another example is that we can deploy the new version in a certain area, and then if everything goes well, roll out to different regions. This always minimizes the impact of end users. Additionally, PipeCD provides software delivery metrics. Currently, we have deployment frequency and challenge failure rate, so you can get our insight of developer pro activities in one place. Talking about feature directions, looking forward, we plan to support more application types as plugins so that community can build their own provider to satisfy their needs. We also aim to enable custom scripts in deployments, which enhances flexibility. Further integration with various open source software ecosystem is also on our roadmap. So if you are interested in PipeCD, go to pipec.dev and try it. Thank you for your listening. Bye-bye.