Memory gc and scheduler behavior

Memory GC and Scheduler Behavior Deep Dive

Overview

User-perceived performance in Android is strongly affected by heap behavior, GC events, scheduler decisions, and context-switch overhead.

ART GC behavior

GC can impact frame timing through:

• pause windows
• concurrent mark/scan CPU usage
• allocation churn from UI and data pipelines

Reduce allocation pressure in hot paths before trying low-level tuning.

Scheduler impact

Linux CFS plus cgroup and priority controls define CPU distribution. Misuse of priorities can improve one path while starving others.

Thread priority anti-patterns

• elevating background work close to UI priority
• long blocking calls on priority-sensitive threads
• lock inversion between mixed-priority workers

Context-switch overhead

In IPC-heavy systems, excess process hops and synchronous chains increase context-switch and wakeup overhead.

Mitigation:

• reduce unnecessary synchronous boundaries
• batch operations
• minimize binder nesting in critical paths

Measurement strategy

Correlate:

• frame metrics
• GC event timings
• scheduler run queue pressure
• context-switch counts and tail latency

Interview guidance

Strong answers connect allocator and scheduler behavior to concrete user-level SLOs, not only microbenchmarks.