What to Expect
Learn more about what happens after you integrate your cloud & monitoring data.
Last updated
Learn more about what happens after you integrate your cloud & monitoring data.
Last updated
Congratulations on starting your autonomous journey! It's important to remember that autonomous is the end goal, not the starting point. We recommend a Crawl-Walk-Run approach to allow your team time to get familiar with and trust Sedai. The system supports this approach through its mode settings. When you integrate any cloud account or Kubernetes cluster, by default Sedai starts in the "Crawl" phase and runs in Datapilot mode.
By default, the system will not make any changes to your resources until you modify the mode from the Settings > Resources page.
The following diagram explains Sedai's different modes and how they correlate to each phase of Crawl-Walk-Run:
So what does Crawl-Walk-Run actually look like? Below is an outline of what you can expect during the first month and setup phase:
Integrate cloud/monitoring data. Sedai automatically imports supported cloud resources and prioritizes relevant metrics.
π Duration: A few hours (depending on the size of your accounts/clusters)
Preliminary analysis. Sedai will build models that capture the behavior of each resource, and use that information to predict optimization opportunities for cost savings. During this time, Sedai will also start to flag potential availability issues.
π Duration: If backfilling metric data, a few hours; otherwise, a few days (the UI will start to populate early on, but we recommend waiting to review for stronger insights)
Review Datapilot opportunities. After the system has been running for a few days, you can explore predicted cost/performance gains from resource configuration changes. Opportunities represent optimal end-state configurations that maximize potential savings. They are intended to guide your team's understanding of how Sedai works and inform decision making when starting to optimize resources.
π Duration: ~1 week
Review Copilot results. Once Sedai completes an optimization, the system will analyze resource behavior in various traffic conditions to ensure the new configuration does not negatively impact performance. The system leverages reinforcement learning with its analysis, and proposes additional optimization phases.
π Duration: ~1 week
Execute recommendations with Copilot. Sedai will ask for permission to continue optimizing resources with incremental changes, and evaluate the results at each step. You can closely review its suggestions and results. π Duration: Up to you! You're in control β this time is all about getting familiar with Sedai and comfortable with how it works
Enable Autopilot. Once you're comfortable with how Sedai operates, we recommend choosing resources that you're ready to let Sedai autonomously manage. Once you update resource settings, Sedai will continue to assess configurations and make incremental changes on its own. Keep in mind that the system will only act if it passes strict safety checks.
It's up to your team on how fast (or slow) you want to progress through each phase. We recommend starting off with pre-production resources to get comfortable, but to truly understand Sedai's full capabilities, it's best to explore production resources. Go at your own pace β our team is here to help.
It takes time for Sedai to understand your data. As the system learns, you can continue to explore the platform and customize it to work best for your team. View the onboarding checklist for a list of tasks you can do during the first few weeks to help you get familiar with Sedai:
Try Copilot. Select opportunities for resources your team is comfortable allowing Sedai to modify (such as from a dev account). For these resources, you will update the optimization setting to run in Copilot mode (learn more), and click to execute each opportunity. If the operation passes Sedai's rigorous safety checks, the system will apply changes. It's important to note that when Sedai first optimizes a resource, it only makes a small change to start. The system evalutes behavior after a new configuration is applied and uses reinforcement learning to identify subsequent optimization phases. Through each phase, Sedai reevaluates the optimal end state and adapts as needed based on how a resource responds to change. π Duration: ~1-2 weeks Pro tip: Create a group of resources to try out Copilot on.