Karolis Scerbiakas

Understanding Service Level Indicators (SLIs).

Karolis Ščerbiakas -

Introduction:

In the world of SaaS, ensuring the reliability and performance of applications is paramount. Service Level Indicators (SLIs) are a crucial component of this process, providing a quantifiable measure of the quality of service being provided. For developers and operations teams understanding and implementing effective SLI metrics is essential for maintaining a competitive edge and delivering a best user experience.

SLIs in Detail:

Service Level Indicators (SLIs) are specific, measurable metrics used to gauge the performance and reliability of a service. They form the foundation of Service Level Objectives (SLOs) and Service Level Agreements (SLAs), which are basically a formal promise made to customers.
SLIs typically focus on key aspects of service performance such as availability, latency, throughput, error rates, and system saturation.

Key Metric Types.

  1. Availability: Measures the uptime of your service and its ability to serve requests. An SLI for availability often track the percentage of time the service is accessible to users within a given period. This can be something simple as hitting endpoint and getting 2xx response back or it can consist of multiple checks throughout the services.
  2. Latency: Assesses the time taken to respond to requests. Latency SLIs are crucial for user experience, as unjustified delays lead to dissatisfaction and churn.
  3. Throughput: Evaluates the number of requests your service can handle within a specific timeframe. This SLI helps to figure out if the service(s) can scale fast enough and manage varying loads.
  4. Error Rate: Quantifies the rate of failed requests. An effective SLI for error rates helps in identifying and mitigating issues that can affect service reliability.
  5. Saturation: Measures the utilization of resources. For example in Kubernetes deployments, this could involve tracking pod memory or CPU usage to prevent overloading and ensure smooth scaling.

SLIs in your Application.

Implementing SLIs requires observability, monitoring and custom metrics tooling. Here's how to approach it:

  1. Define Your SLIs: Begin by identifying what is critical for your users and what constitutes a successful interaction with your service. This will guide you in selecting the appropriate SLIs.
  2. Use Observability Tools: Tools like Prometheus, Grafana Mimir, DataDog and NewRelic are most popular for collecting and visualizing metrics.
  3. Create Custom Metrics: Use these to capture SLIs that are specific to your business logic or application architecture. For example create mail_send_rate to know percentage of successfully sent emails.
  4. Set Up Alerts: Configure alerting based on your SLIs to notify you when performance deviates from expected thresholds. Kubernetes' built-in Horizontal Pod Autoscaler (HPA) can also use custom metrics to automatically scale your application in response to demand. Consider using keda for exactly that.
  5. Rinse and Repeat: Regularly review your SLIs and adjust them as your service evolves. The dynamic nature of SaaS products means that SLIs should be periodically reassessed to ensure they remain relevant and actionable.

How to choose what metrics to use when setting SLI.

Simplicity: Keep your SLIs simple and focused.

Complex metrics can be difficult to interpret and act upon. For example derivative metric such as page views per session. User might not find a product right away and browse a platform more than usual, it won't hurt the service.

Actionability: Ensure that your SLIs lead to actionable insights.

If an SLI doesn't inform a decision or trigger a response, it may not be valuable.

Alignment: Align SLIs with business objectives.

The performance metrics should reflect the goals and priorities of the organization.

Automation: Automate the collection and analysis of SLI data as much as possible.

This reduces the risk of human error and frees up resources for other tasks.

Conclusion:

Effective SLI metrics are a cornerstone of high-performing SaaS products. SLIs provide a framework for monitoring, scaling, and improving applications. Remember to keep your SLIs simple, actionable, transparent, aligned with business goals, and as automated as possible to maintain a robust and reliable service offering.