Analytics Dashboard
🧠 Philosophy
Understanding how your deployed workflows perform at runtime is essential for optimizing compute usage, debugging bottlenecks, and improving reliability.
That’s why the INTELLITHING Analytics Dashboard provides live and historical system metrics — powered by Kubernetes — directly tied to your deployed workflows. No need to integrate third-party observability tools or navigate opaque infrastructure layers.
Whether you're investigating a slowdown, verifying resource limits, or just ensuring stability, the dashboard gives you real-time visibility into container behavior.
🔑 Key Concepts
| Concept | Description |
|---|---|
| Pod Metrics | Resource-level statistics from the running container |
| Date Range Filter | Choose a custom time window for data exploration |
| Live Graphs | View real-time usage patterns as your app is running |
| Resource Categories | Metrics are grouped by CPU, memory, GPU, and container-level info |
| Alignments & Aggregation | All metrics are normalized to 1-minute resolution for clarity |
📘 Key Definitions
| Metric | Meaning |
|---|---|
| CPU Usage Time | Actual CPU time consumed by the container (in cores) |
| CPU Request Utilization | Percent of allocated CPU request being used |
| CPU Limit Utilization | Percent of CPU limit being used |
| Memory Used | Memory consumption in GB |
| Memory Limit Utilization | Percent of memory limit currently used |
| Container Restart Count | Number of times the container restarted during the time window |
| GPU Duty Cycle | Percent of time the GPU was actively in use (if GPU enabled) |
| GPU Memory Used | Amount of GPU memory consumed |
| Total GPU Memory | Allocated GPU memory (useful for capacity planning) |
🧩 Dashboard at a Glance
One-line Summary
Example Usage
- You deploy a workflow to production.
- Head to the Analytics Dashboard.
- Choose a date/time range.
- Click "Refresh Data".
- Review graphs for CPU, memory, and restarts.
- Spot performance trends or stability issues quickly.
⚙️ How Analytics Fits into INTELLITHING
The Analytics tab provides Kubernetes-native observability, integrated directly into your project UI.
- Every deployment is backed by containerized infrastructure.
- These containers emit real-time metrics — CPU usage, memory load, restarts, and GPU activity (if used).
- INTELLITHING collects and displays these metrics using a clean, user-friendly interface.
- You get instant feedback on how well your runtime is performing — without touching kubectl or Prometheus.
🧭 Using the Analytics View
📅 Date Filter
Choose the time window you want to inspect.
- From: Start date/time
- To: End date/time
- Then click "Refresh Data" to reload graphs.
This allows you to explore past performance for debugging or trend analysis.
📈 Interpreting Graphs
Each graph is labeled and auto-updated. Some helpful tips:
| Graph | What to Look For |
|---|---|
| CPU Usage Time | Spikes or flatlines — may indicate heavy load or idling |
| CPU Request/Limit Utilization | Percent-based comparison vs. assigned resources |
| Memory Used | Track memory growth or memory leaks |
| Memory Limit Utilization | Ensure you’re not near the max (100%) |
| Container Restart Count | Anything > 0 suggests instability or failure |
| GPU Metrics | Only visible if GPU is provisioned |
Values like "0.78" on memory used = 0.78 GB used at that time point.
🚨 Common Scenarios
| Situation | What to Check |
|---|---|
| App is slow | Look at CPU Usage Time, Memory Used |
| App crashes or restarts | Check Container Restart Count, Memory Limit Utilization |
| Model seems underutilized | Review GPU Duty Cycle or CPU Request Utilization |
| Memory leak suspected | Look for steady upward trend in Memory Used |
| Too much allocated but unused | Low CPU/Memory utilization — consider scaling down |
🔐 Permissions
| Feature | Required Role |
|---|---|
| Access to Analytics | All users with access to the project |
| Build/Deploy to see metrics | Requires actual deployment to exist |
No metrics will be shown unless a valid deployment exists for the selected date range.
💡 Best Practices
| Tip | Why It Helps |
|---|---|
| Always check after deploy | Verify that resource limits and usage align |
| Investigate restarts immediately | Often caused by crashes, memory limits, or bad configs |
| Monitor during load tests | Helps determine real-world CPU/memory usage |
| Use date range to compare trends | Spot regressions across deployments |
| Keep limits generous but safe | Avoid crashing under load or wasting idle capacity |
💬 Summary
- The Analytics Dashboard gives you deep insight into real-time container performance.
- Metrics are broken down by CPU, Memory, GPU, and Container Health.
- You can choose custom date ranges and explore visual trends interactively.
- This view helps debug, optimize, and confidently run production-grade workflows — all from inside INTELLITHING.