Struct WorkloadPredictor 

pub struct WorkloadPredictor { /* private fields */ }

Workload predictor using simple ML models

Implementations

impl WorkloadPredictor

pub const MIN_SAMPLES: usize = 10

Samples required before a prediction is meaningful.

pub fn new() -> Self

pub fn update_metrics(&mut self, metrics: &PerformanceMetrics)

pub fn sample_count(&self) -> usize

Number of utilization samples observed so far.

pub fn can_predict(&self) -> bool

Whether enough history has accumulated for WorkloadPredictor::predict_workload to answer.

pub fn predict_workload(&self, horizon: Duration) -> Result<f32>

Predict mean GPU utilization horizon from now.

Errors

When fewer than WorkloadPredictor::MIN_SAMPLES samples have been recorded. Earlier revisions returned a hard-coded 0.75 here, which the auto-scaler then acted on as though it were a measurement.

Trait Implementations

impl Default for WorkloadPredictor

fn default() -> Self

Returns the “default value” for a type.