Module tiering 

Hot/Warm/Cold Tiering System for Vector Indices

This module provides a production-grade tiering system that automatically manages vector indices across different storage tiers based on access patterns, size, and performance requirements.

Tiering Strategy

• Hot Tier: In-memory, frequently accessed indices (sub-millisecond latency)
• Warm Tier: Memory-mapped or SSD-backed indices (millisecond latency)
• Cold Tier: Compressed, disk-based indices (second latency, loaded on demand)

Features

• Automatic tier promotion/demotion based on access patterns
• Configurable policies (LRU, LFU, cost-based, adaptive)
• Integration with monitoring and analytics
• Support for gradual tier transitions
• Predictive tier management using ML
• Multi-tenant resource allocation

Example

use oxirs_vec::tiering::{TieringManager, TieringConfig, TieringPolicy};

let config = TieringConfig {
    hot_tier_capacity_gb: 16.0,
    warm_tier_capacity_gb: 128.0,
    cold_tier_capacity_gb: 1024.0,
    policy: TieringPolicy::Adaptive,
    ..Default::default()
};

let mut manager = TieringManager::new(config)?;

// Add an index
manager.register_index("embeddings_v1", index)?;

// Access triggers automatic tier management
let results = manager.query_index("embeddings_v1", query_vector, k)?;

// Check tier statistics
let stats = manager.get_tier_statistics();
println!("Hot tier: {:.2}% utilized", stats.hot_tier_utilization * 100.0);

Re-exports

pub use config::TieringConfig;

pub use manager::TieringManager;

pub use metrics::TierMetrics;

pub use policies::TierTransitionReason;

pub use policies::TieringPolicy;

pub use types::IndexMetadata;

pub use types::StorageTier;

pub use types::TierStatistics;

Modules

access_tracker

Access pattern tracking for tiering decisions

config

Configuration for hot/warm/cold tiering system

manager

Main tiering manager that coordinates tier operations

metrics

Metrics collection for tiering system

policies

Tiering policies for index placement and transition decisions

storage_backends

Storage backends for different tiers

tier_optimizer

Tier optimization using ML and statistical analysis

types

Core types for hot/warm/cold tiering system