Struct ConcurrencyMetrics 

pub struct ConcurrencyMetrics { /* private fields */ }

Concurrency metrics for resource manager monitoring

Educational: Observability-Driven Tuning

These metrics answer key questions:

• “Are we CPU-saturated?” → cpu_saturation_percent()
• “Are we I/O-saturated?” → io_saturation_percent()
• “What’s causing latency?” → wait time histograms
• “How much memory are we using?” → memory_used_bytes

Use these to guide tuning decisions!

Implementations

impl ConcurrencyMetrics

pub fn new(cpu_tokens: usize, io_tokens: usize, memory_capacity: usize) -> Self

pub fn update_cpu_tokens_available(&self, available: usize)

Update CPU tokens available (from ResourceManager)

pub fn cpu_tokens_available(&self) -> usize

Get CPU tokens available

pub fn cpu_saturation_percent(&self) -> f64

Get CPU saturation percentage

Educational: What is saturation?

Saturation = (tokens_in_use / total_tokens) × 100

• 0%: Idle, not utilizing resources
• 50%: Good utilization
• 80-90%: High utilization, approaching saturation
• 100%: Fully saturated, tasks waiting

pub fn record_cpu_wait(&self, duration: Duration)

Record CPU token wait time

pub fn cpu_wait_p50(&self) -> u64

Get CPU wait time percentile

pub fn cpu_wait_p95(&self) -> u64

pub fn cpu_wait_p99(&self) -> u64

pub fn update_io_tokens_available(&self, available: usize)

Update I/O tokens available (from ResourceManager)

pub fn io_tokens_available(&self) -> usize

pub fn io_saturation_percent(&self) -> f64

pub fn record_io_wait(&self, duration: Duration)

Record I/O token wait time

pub fn io_wait_p50(&self) -> u64

pub fn io_wait_p95(&self) -> u64

pub fn io_wait_p99(&self) -> u64

pub fn update_memory_used(&self, bytes: usize)

pub fn memory_used_bytes(&self) -> usize

pub fn memory_used_mb(&self) -> f64

pub fn memory_capacity_bytes(&self) -> usize

pub fn memory_utilization_percent(&self) -> f64

pub fn worker_started(&self)

pub fn worker_completed(&self)

pub fn active_workers(&self) -> usize

pub fn tasks_spawned(&self) -> u64

pub fn tasks_completed(&self) -> u64

pub fn update_cpu_queue_depth(&self, depth: usize)

Update CPU queue depth

Educational: Observing Backpressure

Queue depth reveals whether workers can keep up with the reader:

• Depth near 0: Workers are faster than reader (good!)
• Depth near capacity: Workers are bottleneck (increase workers or optimize stages)
• Depth at capacity: Reader is blocked (severe backpressure)

pub fn cpu_queue_depth(&self) -> usize

Get current CPU queue depth

pub fn cpu_queue_depth_max(&self) -> usize

Get maximum CPU queue depth observed

pub fn record_cpu_queue_wait(&self, duration: Duration)

Record time a chunk waited in CPU queue

pub fn cpu_queue_wait_p50(&self) -> u64

Get P50 (median) CPU queue wait time in milliseconds

pub fn cpu_queue_wait_p95(&self) -> u64

Get P95 CPU queue wait time in milliseconds

pub fn cpu_queue_wait_p99(&self) -> u64

Get P99 CPU queue wait time in milliseconds

pub fn reset(&self)

Reset all metrics (for testing/benchmarking)

Trait Implementations

impl Debug for ConcurrencyMetrics

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.