CrystalClear Time Series Analysis Engine —
Correlation-Based Forecasting for AIOps Resource Management

Predicting resource usage for every microservice in an application — without manually tuning a model for each one.

Applied the Multi-Layer Correlation patent to its algorithms, CrystalClear Time Series Analysis Engine is the forecasting core of Federator.ai. It uses cross-correlation between an application’s primary workload and its microservices to generate fast, accurate resource predictions — even across hundreds of metrics at once.

Why Traditional Forecasting Falls Short

Most time series tools were built to model one metric at a time. Microservice environments don’t work that way.

One Model Per Metric

Traditional forecasting tools analyze each microservice metric in isolation — identifying, estimating, and validating a separate model every time. With hundreds of microservices, this doesn’t scale.

Blind to Dependencies

Without seeing how microservices relate to the application’s overall workload, forecasts miss the bigger picture. A spike in one service often signals what’s coming for others — but isolated models can’t catch that.

Too Slow for Real-Time Use

Manual model tuning and validation take seconds per metric. When predictions need to drive live autoscaling decisions, that delay is the difference between proactive and reactive.

What It Is

CrystalClear Time Series Analysis Engine is ProphetStor’s correlation-based forecasting approach. Instead of modeling every metric from scratch, it uses the relationship between an application’s primary workload and its microservices to generate predictions directly.

High Correlation — Fast Path

When a microservice’s resource usage closely tracks the application’s primary workload, CrystalClear builds a prediction model directly from that relationship — skipping the slow model-identification process entirely.

Low Correlation — Feature-Based Path

When the relationship is weaker, CrystalClear falls back to analyzing the metric’s own trend, seasonality, and change points to build a tailored forecast.

How CrystalClear Time Series Analysis Engine builds predictions: correlation analysis routes each microservice to either primary-feature or app-feature modeling

Why It Matters

Scales to Hundreds of Microservices

Predictions are generated automatically across an entire application, without per-metric configuration.

Faster Than
Traditional Tools

By reusing correlation structure instead of rebuilding models from scratch, CrystalClear produces forecasts in a fraction of the time.

Built for Real-World Kubernetes

Restarts, scaling events, and irregular workloads are common in Kubernetes. CrystalClear’s feature-based fallback keeps forecasts usable even when patterns get messy.

Frequently Asked Questions

It forecasts resource usage for every microservice in an application by analyzing how closely each one correlates with the application’s primary workload.
No. It automatically chooses between a correlation-based model or a feature-based model depending on each metric’s relationship to the primary workload.
Microservice-based applications running in Kubernetes or other cloud-native environments, as part of Federator.ai’s broader resource optimization and autoscaling capabilities.

Please select the software you would like a demo of:

Federator.ai GPU Booster ®

Maximizing GPU utilization for AI workloads and doubling your server’s training capacity

Federator.ai ®

Simplifying complexity and continuously optimizing cloud costs and performance