Skip to content

Day 25 - Hybrid Robustness & Update Coalescing

Overview

This chapter enhances the hybrid (IRQ + polling) model by introducing:

  • Update coalescing
  • Global cooldown policy
  • Detailed runtime observability

The goal is to prevent excessive updates while maintaining data freshness.

Key Concepts

1. Event vs Update vs Skip

Each trigger (IRQ / polling / on-demand) is categorized into:

  • Event: trigger occurs
  • Update: cache is refreshed
  • Skip: update is suppressed by policy
event_count >= update_count
skip_count = event_count - update_count

2. Global Cooldown

A new parameter:

min_update_interval_ms

Defines the minimum time between two successful updates.

Behavior:

  • If last update is too recent → skip
  • Applies to ALL triggers (IRQ / poll / on-demand)

3. Update Policy

Final decision logic:

should_update =
    trigger_allowed_by_mode
    AND trigger_specific_interval
    AND not_within_cooldown

4. Trigger-specific Rules

IRQ

  • Allowed when irq_enable == true
  • No intrinsic interval limit
  • Controlled mainly by cooldown

Polling

  • Allowed when polling_enable == true
  • Uses polling_interval_ms
  • In hybrid mode:
  • Acts as fallback when data is stale

On-demand

  • Active only when IRQ and polling are disabled
  • Uses cache_interval_ms

5. Cooldown Helper

myi2c_sensor_is_within_update_cooldown_locked()

Responsibilities:

  • Check elapsed time since last update
  • Compare with min_update_interval_ms
  • Return whether update should be suppressed

Observability Design

New counters:

IRQ

  • irq_event_count
  • irq_update_count
  • irq_skip_count

Polling

  • poll_event_count
  • poll_update_count
  • poll_skip_count

On-demand

  • ondemand_update_count
  • ondemand_skip_count

Debug Interface

Example:

mode=hybrid
irq_event_count=924
irq_update_count=82
irq_skip_count=842
poll_event_count=356
poll_update_count=239
poll_skip_count=116

This allows clear analysis of:

  • Trigger frequency
  • Update efficiency
  • Coalescing effectiveness

Design Insights

  • IRQ may generate frequent events but be heavily rate-limited
  • Polling provides stability under high IRQ load
  • Cooldown acts as a global rate limiter
  • System favors "eventual consistency" over strict immediacy

Key Takeaways

  • Separate event, update, and skip metrics
  • Use cooldown to prevent redundant work
  • Hybrid systems require careful policy layering
  • Observability is essential for debugging behavior