Redis Caching Strategy - AIUsagePlatform API Portal

⚠️ Current State

The platform currently has no distributed caching layer. All dashboard and analytics queries hit the database directly, causing performance degradation at scale.

❌ No Cache
Every request queries database

⚠️ N+1 Queries
Aggregation endpoints trigger multiple queries

⚠️ Cold Start
Dashboard slow on first load

🎯 Target
70% cache hit rate, 50ms response

🎯 Proposed Architecture

Redis Caching Architecture ═══════════════════════════════════════════════════════════════════════════════ ┌─────────────────────────────────────────────────────────────────────────────┐ │ CURRENT (No Cache) │ ├─────────────────────────────────────────────────────────────────────────────┤ │ │ │ Client → API Controller → Service → Repository → SQL Server │ │ │ │ │ │ │ 100-500ms query latency │ │ │ ▼ │ │ │ (cold data on every request) │ │ │ │ └─────────────────────────────────────────────────────────────────────────────┘ ↓ ┌─────────────────────────────────────────────────────────────────────────────┐ │ PROPOSED (With Redis Cache) │ ├─────────────────────────────────────────────────────────────────────────────┤ │ │ │ Client → API Controller → CacheInterceptor → Service → Repository → DB │ │ │ │ │ │ │ Cache Hit? │ Cache Miss │ │ │ ──────────────────────────────│─────────────│ │ │ Yes → Return cached │ No → Query │ │ │ (5-15ms) │ DB │ │ │ │ │ │ │ │ │ ▼ │ │ │ │ Store in │ │ │ │ Redis │ │ ▼ │ (TTL) │ │ ┌──────────────┐ │ │ │ │ Redis │◄─────────────────────┘ │ │ │ Cluster │ │ │ │ │ │ │ │ • Dashboard │ │ │ │ KPIs (5m) │ │ │ │ • Analytics │ │ │ │ (15m) │ │ │ │ • Sessions │ │ │ │ (1h) │ │ │ │ • User │ │ │ │ (1h) │ │ │ └──────────────┘ │ │ │ └─────────────────────────────────────────────────────────────────────────────┘

📦 Cache Key Strategy

Data Type	Cache Key Pattern	TTL	Invalidation
Dashboard KPIs	`dashboard:{orgId}:kpi`	5 min	On session create
Top Organizations	`dashboard:top-orgs:{date}`	15 min	Daily refresh
User Analytics	`analytics:user:{userId}:{date}`	1 hour	On session end
Organization Analytics	`analytics:org:{orgId}:{date}`	15 min	Hourly refresh
Session List	`sessions:{orgId}:{page}:{size}`	5 min	On new session
Export Status	`export:{jobId}:status`	30 min	Job completion
User Profile	`user:{userId}:profile`	1 hour	Profile update
Plan Features	`plan:{planId}:features`	24 hours	Plan change

🔧 Implementation Plan

Phase 1: Infrastructure

□ Provision Redis cluster (Azure Cache for Redis)
□ Add StackExchange.Redis package
□ Configure connection strings
□ Health checks for Redis

Priority: High Not Started

Phase 2: Cache Service

□ ICacheService interface
□ RedisCacheService implementation
□ Serialization (JSON/MessagePack)
□ Compression for large objects

Priority: High Not Started

Phase 3: Integration

□ Cache-aside in DashboardController
□ Cache-aside in AnalyticsController
□ Cache invalidation on writes
□ Cache warming background job

Priority: Medium Not Started

Phase 4: Advanced

□ Response caching middleware
□ Cache tags for bulk invalidation
□ Distributed locks for race conditions
□ Circuit breaker for Redis failures

Priority: Low Future

💻 Code Example

// Proposed ICacheService interface
public interface ICacheService
{
    Task<T> GetOrCreateAsync<T>(string key, 
        Func<Task<T>> factory, 
        TimeSpan expiration);
    
    Task RemoveAsync(string key);
    Task RemoveByPatternAsync(string pattern);
}

// Usage in DashboardController
public async Task<IActionResult> GetDashboard()
{
    var orgId = GetCurrentOrganizationId();
    var cacheKey = $"dashboard:{orgId}:kpi";
    
    var data = await _cacheService.GetOrCreateAsync(cacheKey,
        () => _dashboardService.GetDashboardAsync(orgId),
        TimeSpan.FromMinutes(5));
    
    return Ok(data);
}
        

📊 Expected Performance Improvements

Endpoint	Current (No Cache)	With Cache (Hit)	Improvement
/api/admindashboard	150-400ms	10-25ms	~90% faster
/api/analytics	200-800ms	15-30ms	~92% faster
/api/analytics/trends	300-1000ms	20-40ms	~95% faster
/api/usersmanagement	100-300ms	10-20ms	~88% faster

⚠️ Risks & Mitigations

⚠

Cache Stampede

Multiple requests cache miss simultaneously. Mitigation: Use async locking or cache warming.

⚠

Stale Data

Users may see outdated data. Mitigation: Proper TTL selection and explicit invalidation.

⚠

Redis Failure

Cache becomes unavailable. Mitigation: Circuit breaker, fallback to DB, degrade gracefully.

⚠

Memory Pressure

Redis eviction under load. Mitigation: LRU policy, memory limits, monitoring alerts.