Skip to content

Latest commit

 

History

History
861 lines (671 loc) · 28 KB

performance.md

File metadata and controls

861 lines (671 loc) · 28 KB

Akavache Performance Guide

This comprehensive guide covers performance analysis, benchmarks, best practices, and optimization techniques for Akavache V11 vs V10.

Akavache V10 vs V11 Performance Summary

Quick Performance Comparison

V11 Wins

  • Bulk Operations: 10x+ faster than individual operations
  • GetOrFetch Pattern: Sub-linear scaling (1.5ms to 45ms for 100x data)
  • Memory Consistency: More predictable allocation patterns
  • In-Memory Performance: 122ms for 1000 complex operations
  • Architecture: Modern builder pattern with better error handling

Equivalent Performance

  • Cache Type Operations: All persistent caches within 2% of each other
  • Read Operations: Generally comparable with V10
  • Object Serialization: SystemTextJson matches or exceeds V10 performance
  • Memory Usage: Similar allocation patterns across versions

V11 Trade-offs

  • Large Sequential Reads: Up to 8.6% slower in some cases
  • Initialization Overhead: Builder pattern adds slight complexity
  • Package Dependencies: More granular package structure

Key Numbers

Operation Small (10) Medium (100) Large (1000)
GetOrFetch 1.5ms 15ms 45ms
Bulk Operations 3.3ms 4.5ms 18ms
In-Memory 2.4ms 19ms 123ms
Cache Types ~27ms ~255ms ~2,600ms

Migration Decision Matrix

Factor V10 to V11 Migration Recommended
New Projects Always
Performance Critical Yes (with SystemTextJson)
Legacy Data Compatibility Yes (with Newtonsoft BSON)
Large Sequential Reads Evaluate (8.6% slower)
Developer Experience Highly Recommended

Bottom Line

Akavache V11 delivers architectural improvements with comparable performance. The new features (multiple serializers, cross-compatibility, modern patterns) provide significant value with minimal performance impact.

Recommendation: Upgrade to V11 for all new projects and consider migration for existing projects that would benefit from the architectural improvements.

Akavache Performance Benchmark Report - V10 vs V11 Comprehensive Comparison

This report provides a detailed performance comparison between Akavache V10.2.41 and Akavache V11.0 (current development version), covering all major functionality areas and use cases.

Executive Summary

Akavache V11.0 represents a significant architectural advancement over V10.2.41, introducing:

  • New builder pattern for initialization
  • Multiple serializer support (System.Text.Json, Newtonsoft.Json, BSON variants)
  • Cross-serializer compatibility
  • Enhanced modularity with separate packages
  • Improved settings management

The performance analysis shows comparable or improved performance across most operations, with V11's architectural benefits providing significant developer experience improvements.

Test Environment

  • Framework: .NET 9.0
  • OS: Windows 11 (10.0.26100.4770/24H2/2024Update/HudsonValley)
  • Benchmarking Tool: BenchmarkDotNet v0.15.2
  • Runtime: .NET 9.0.8 (9.0.825.36511), X64 RyuJIT AVX2
  • Test Data: Complex objects with Guid, string, int, and DateTimeOffset properties

Architectural Differences

Akavache V10.2.41

// Legacy initialization
BlobCache.ApplicationName = "MyApp";
// Ready to use: BlobCache.UserAccount, BlobCache.LocalMachine, etc.
  • Single serialization approach
  • Traditional singleton pattern
  • Monolithic package structure

Akavache V11.0

// Modern builder pattern
CacheDatabase.Serializer = new SystemJsonSerializer();
CacheDatabase.Initialize(builder =>
    builder.WithApplicationName("MyApp")
           .WithSqliteDefaults());
// Ready to use: CacheDatabase.UserAccount, CacheDatabase.LocalMachine, etc.
  • Multiple serializer options
  • Explicit configuration
  • Modular package design
  • Enhanced type safety

Performance Results Summary

V11 Performance Characteristics (Latest Benchmark Results)

Based on comprehensive benchmarks across different operation types and data sizes:

Small Dataset Performance (10 items)

Operation Mean Time Memory Allocation Notes
GetOrFetchObject 1.476 ms 167.47 KB Fast object retrieval with fallback
InMemoryOperations 2.402 ms 227.2 KB Excellent in-memory performance
BulkOperations 3.261 ms 66.08 KB Efficient bulk processing
UserAccount/LocalMachine/Secure ~27 ms ~227 KB Consistent across cache types

Medium Dataset Performance (100 items)

Operation Mean Time Memory Allocation Performance Notes
GetOrFetchObject 15.179 ms 1,427.63 KB Linear scaling
InMemoryOperations 18.796 ms 2,025.13 KB Good scalability
BulkOperations 4.504 ms 243.73 KB Bulk advantage clear
InsertWithExpiration 245.723 ms 671.27 KB Expiration overhead minimal
Cache Type Operations ~255 ms ~2,000 KB Consistent performance

Large Dataset Performance (1000 items)

Operation Mean Time Memory Allocation Scalability
GetOrFetchObject 45.034 ms 14,053.05 KB Sub-linear scaling
InMemoryOperations 122.826 ms 19,740.02 KB Excellent for large datasets
BulkOperations 17.992 ms 2,015.54 KB Best scaling performance
Cache Types ~2,600 ms ~19,800 KB Predictable large-scale performance
InvalidateObjects 7,141.040 ms 17,071.66 KB Expected invalidation cost

Key Performance Insights

Performance Strengths of V11

  1. Bulk Operations Excel:

    • 1000 items processed in just 17.992 ms
    • Minimal memory overhead (2,015.54 KB)
    • 10x+ faster than individual operations

  2. GetOrFetch Pattern Optimization:

    • Highly optimized for cache-miss scenarios
    • Scales sub-linearly: 10 items (1.5ms) → 1000 items (45ms)
    • Excellent for real-world usage patterns

  3. Memory Efficiency:

    • Consistent memory usage across cache types
    • Predictable allocation patterns
    • No memory leaks or excessive overhead

  4. In-Memory Performance:

    • 122ms for 1000 complex object operations
    • Ideal for session data and temporary caching

V11 vs V10 Comparison Highlights

Read Performance

Based on earlier comparative runs:

  • V11 shows 1.8-3.4% faster read performance for smaller datasets
  • More consistent performance with lower standard deviations
  • Bulk reads perform equivalently between versions

Write Performance

  • Sequential writes: Comparable performance
  • Bulk writes: V11 shows significant advantages with new bulk API
  • Object serialization: SystemTextJson in V11 outperforms legacy serialization

Memory Usage

  • Generally equivalent or better memory efficiency in V11
  • More predictable allocation patterns
  • Garbage collection performance improved

Serializer Performance Comparison

System.Text.Json (V11 Default)

  • ~260ms for 100 complex objects (serialization + storage + retrieval)
  • 2,029.27 KB memory allocation
  • Best choice for new applications

Cross-Serializer Compatibility Benefits

V11 can read data written by different serializers:

// Write with one serializer
CacheDatabase.Serializer = new NewtonsoftBsonSerializer();
await cache.InsertObject("key", data);

// Read with another serializer
CacheDatabase.Serializer = new SystemJsonSerializer();
var retrieved = await cache.GetObject<MyData>("key"); // Still works!

Cache Type Performance Analysis

All cache types show remarkably consistent performance in V11:

Cache Type 100 Items 1000 Items Use Case
UserAccount 253.847 ms 2,603.138 ms User settings, preferences
LocalMachine 257.036 ms 2,642.941 ms Cached API data, temporary files
Secure 255.278 ms 2,620.867 ms Credentials, sensitive data
InMemory 18.796 ms 122.826 ms Session data, frequently accessed

Key Insight: Persistent cache types perform within 2% of each other, showing excellent architecture consistency.

Advanced Operation Performance

GetOrFetch Pattern (Cache-Miss Scenario)

  • 45.034 ms for 1000 operations
  • Ideal for API data caching with fallback
  • Efficient handling of cache misses

Mixed Operations (Real-World Simulation)

  • 3,552.625 ms for 1000 complex operations
  • Includes: Insert → Read → Update → Verify cycle
  • Demonstrates excellent real-world performance

Expiration Handling

  • 2,604.595 ms for 1000 items with expiration
  • Minimal overhead for time-based invalidation
  • Essential for API caching scenarios

Migration Performance Impact

From V10 to V11

  • < 5% performance difference for most operations
  • Significant improvements in consistency
  • Better error handling and recovery

Breaking Changes

  1. Initialization pattern requires code changes
  2. Must explicitly choose serializer
  3. Package structure requires dependency updates

Migration Benefits

  1. Future-proof architecture
  2. Better performance monitoring
  3. Enhanced debugging capabilities
  4. Serializer flexibility

Best Practices for Optimal Performance

1. Choose the Right Serializer

// For new applications (fastest)
CacheDatabase.Serializer = new SystemJsonSerializer();

// For V10 compatibility (most compatible)
CacheDatabase.Serializer = new NewtonsoftBsonSerializer();

2. Use Bulk Operations When Possible

// Instead of multiple individual inserts
await cache.InsertObjects(largeDataSet); // Much faster!

3. Leverage Cache Type Appropriately

// Fast temporary data
await CacheDatabase.InMemory.InsertObject("session", data);

// Persistent user data
await CacheDatabase.UserAccount.InsertObject("settings", userSettings);

4. Optimize with GetOrFetch Pattern

var data = await cache.GetOrFetchObject("api_data", 
    () => httpClient.GetFromJsonAsync<ApiData>(url));

Recommendations

Choose V11 for New Projects

  • Similar or better performance
  • Modern, maintainable architecture
  • Future-proof design
  • Enhanced developer experience

Migrate from V10 When

  • You need multiple serializer support
  • You want better error handling
  • You require cross-serializer compatibility
  • Performance is not the primary concern (impact minimal)

Consider Carefully for

  • Large sequential read workloads (8.6% slower in some cases)
  • Legacy data requiring specific serialization
  • Applications with extremely tight performance budgets

Performance-Critical Applications

  • Use System.Text.Json serializer
  • Leverage bulk operations wherever possible
  • Consider InMemory cache for frequently accessed data
  • Implement proper expiration strategies

Conclusion

Akavache V11.0 successfully delivers on its architectural improvements while maintaining excellent performance characteristics. The comprehensive benchmarks show:

  • Performance : Generally equal or better than V10
  • Consistency : Much more predictable performance patterns
  • Flexibility : Multiple serializer options for optimization
  • Architecture : Modern, maintainable, and extensible design

Bottom Line: V11 is a compelling upgrade that provides architectural benefits without significant performance regression. The slight trade-offs in specific scenarios are more than offset by the improved developer experience, better error handling, and future-proof design.

Performance data collected using BenchmarkDotNet v0.15.2 on .NET 9.0. Results represent typical performance and may vary based on hardware configuration, data characteristics, and usage patterns.

Performance Optimization Tips

1. Choose the Right Serializer

// ✅ ALWAYS use System.Text.Json for optimal V11 performance
// This is faster than V10 across all scenarios and significantly faster than V11 Newtonsoft
AppBuilder.CreateSplatBuilder()
    .WithAkavacheCacheDatabase<SystemJsonSerializer>(builder =>
        builder.WithApplicationName("MyApp")
               .WithSqliteProvider()
               .WithSqliteDefaults());

// ⚠️ For V10 compatibility with large datasets, consider Newtonsoft BSON
// (Only if you need V10 format compatibility - otherwise use System.Text.Json)
AppBuilder.CreateSplatBuilder()
    .WithAkavacheCacheDatabase<NewtonsoftBsonSerializer>(builder =>
        builder.WithApplicationName("MyApp")
               .WithSqliteProvider()
               .WithSqliteDefaults());

2. Use Bulk Operations When Possible

// ✅ Use batch operations for multiple items
await CacheDatabase.UserAccount.InsertObjects(manyItems);

// ✅ Bulk retrieval
var keys = new[] { "key1", "key2", "key3" };
var results = await CacheDatabase.UserAccount.GetObjects<MyData>(keys).ToList();

// ❌ Avoid individual operations in loops
foreach (var item in items)
{
    await CacheDatabase.UserAccount.InsertObject($"key_{item.Id}", item); // Slow
}

3. Set Appropriate Expiration Times

// ✅ Set reasonable expiration for cached data
await CacheDatabase.LocalMachine.InsertObject("temp_key", data, 30.Minutes().FromNow());

// ✅ Use different expiration policies by data type
await CacheDatabase.LocalMachine.InsertObject("api_cache", apiData, 1.Hours().FromNow());
await CacheDatabase.LocalMachine.InsertObject("image_cache", imageBytes, 1.Days().FromNow());

// ✅ Don't expire user settings (unless necessary)
await CacheDatabase.UserAccount.InsertObject("user_preferences", prefs); // No expiration

4. Use Appropriate Cache Types

// ✅ Use InMemory cache for frequently accessed data
await CacheDatabase.InMemory.InsertObject("hot_data", frequentData);

// ✅ Use LocalMachine for cacheable data
await CacheDatabase.LocalMachine.InsertObject("api_response", apiData);

// ✅ Use UserAccount for persistent user data
await CacheDatabase.UserAccount.InsertObject("user_settings", settings);

// ✅ Use Secure for sensitive data
await CacheDatabase.Secure.InsertObject("api_key", apiKey);

5. Optimize Object Storage

// ✅ Use specific types instead of object when possible
await CacheDatabase.UserAccount.GetObject<SpecificType>("key"); // Good

// ❌ Avoid generic object retrieval when type is known
await CacheDatabase.UserAccount.Get("key", typeof(SpecificType)); // Slower

// ✅ Avoid storing very large objects
// Instead, break them into smaller chunks or use compression

// ✅ Use compression for large data sets
var compressedData = await CompressAsync(largeData);
await CacheDatabase.LocalMachine.Insert("compressed_key", compressedData);

6. GetOrFetch Pattern Optimization

// ✅ Use GetOrFetch for cache-miss scenarios
var data = await CacheDatabase.LocalMachine.GetOrFetchObject("api_data", 
    () => httpClient.GetFromJsonAsync<ApiData>(url));

// ✅ Set reasonable expiration for fetched data
var weatherData = await CacheDatabase.LocalMachine.GetOrFetchObject("weather",
    () => weatherApi.GetCurrentWeather(),
    DateTimeOffset.Now.AddMinutes(30));

7. Proper Initialization

// ✅ Initialize once at app startup
public class App
{
    static App()
    {
        AppBuilder.CreateSplatBuilder()
            .WithAkavacheCacheDatabase<SystemJsonSerializer>(builder =>
                builder.WithApplicationName("MyApp")
                       .WithSqliteProvider()
                       .WithSqliteDefaults());
    }
}

// ❌ Don't initialize multiple times

8. Proper Shutdown

// ✅ Always shutdown Akavache properly
public override void OnExit(ExitEventArgs e)
{
    CacheDatabase.Shutdown().Wait();
    base.OnExit(e);
}

9. Testing Performance

// ✅ Use in-memory cache for unit tests
[SetUp]
public void Setup()
{
    AppBuilder.CreateSplatBuilder()
        .WithAkavacheCacheDatabase<SystemJsonSerializer>(builder =>
            builder.WithApplicationName("TestApp")
                   .WithInMemoryDefaults());
}

[TearDown]
public void TearDown()
{
    CacheDatabase.Shutdown().Wait();
}

Reproducing Performance Benchmarks

Platform Requirements

Benchmark reproduction requires Windows hosts. Linux/macOS are not supported due to Windows-specific projects and dependencies used in the benchmark harnesses.

Prerequisites

  • .NET 9.0 SDK
  • Windows operating system
  • PowerShell 5.0+ (for automation script)

Running Compatibility Tests

# From the solution root directory
.\src\RunCompatTest.ps1

Running Performance Benchmarks

# V11 benchmarks (current)
cd src
dotnet run -c Release -p Akavache.Benchmarks/Akavache.Benchmarks.csproj

# V10 comparison benchmarks  
dotnet run -c Release -p Akavache.Benchmarks.V10/Akavache.Benchmarks.V10.csproj

Important Notes:

  • Results vary by hardware configuration and system load
  • Benchmarks are indicative, not absolute measurements
  • Large benchmark runs can take 10-30 minutes to complete
  • Some benchmark projects use BenchmarkDotNet which requires Windows-specific optimizations

Known Limitations

V11 Performance Considerations

  • Large Databases with Newtonsoft.Json: V10 outperforms V11 when using legacy Newtonsoft serialization with very large datasets
  • Sequential Read Performance: Up to 8.6% slower than V10 specifically when using the legacy Newtonsoft.Json serializer (System.Text.Json does not have this limitation and performs better than V10)
  • Package Dependencies: More granular package structure may require careful workload management

Platform Limitations

  • Linux/macOS Build: Benchmark projects and compatibility tests require Windows due to platform-specific dependencies

Summary

For optimal V11 performance:

  1. Always use System.Text.Json serializer - faster than V10 across all scenarios
  2. Use bulk operations for multiple items - 10x+ performance improvement
  3. Choose appropriate cache types - InMemory for hot data, LocalMachine for API caches
  4. Set reasonable expiration times - balance between freshness and performance
  5. Initialize once at startup - avoid repeated initialization overhead
  6. Proper shutdown - ensure data consistency and resource cleanup

V11 delivers architectural improvements with excellent performance when following these best practices. // This is faster than V10 across all scenarios .WithAkavacheCacheDatabase()

// ⚠️ Only use Newtonsoft.Json for V10 compatibility .WithAkavacheCacheDatabase()


### 2. Use Batch Operations

```csharp
// ❌ Slow - Individual operations
foreach (var item in items)
{
    await cache.InsertObject($"key_{item.Id}", item);
}

// ✅ Fast - Batch operation (10x+ faster)
var keyValuePairs = items.ToDictionary(item => $"key_{item.Id}", item => item);
await cache.InsertObjects(keyValuePairs);

3. Choose Appropriate Cache Types

// ✅ InMemory for frequently accessed data
await CacheDatabase.InMemory.InsertObject("hot_data", frequentData);

// ✅ UserAccount for user-specific persistent data
await CacheDatabase.UserAccount.InsertObject("user_prefs", userPreferences);

// ✅ LocalMachine for shared application data
await CacheDatabase.LocalMachine.InsertObject("app_config", configuration);

4. Set Appropriate Expiration Times

// ✅ Short expiration for frequently changing data
await cache.InsertObject("stock_price", price, TimeSpan.FromMinutes(1));

// ✅ Long expiration for stable reference data
await cache.InsertObject("country_list", countries, TimeSpan.FromDays(30));

// ✅ No expiration for settings and configuration
await cache.InsertObject("user_settings", settings);

5. Use Specific Types Instead of Object

// ✅ Fast - Specific type
await cache.GetObject<SpecificType>("key");

// ❌ Slower - Generic object
await cache.Get("key", typeof(SpecificType));

6. Optimize Object Size

// ❌ Large objects slow down serialization
public class HeavyObject
{
    public byte[] LargeByteArray { get; set; } // Multiple MB
    public List<ComplexNestedObject> Items { get; set; } // Thousands of items
}

// ✅ Break large objects into smaller chunks
public class OptimizedObject
{
    public string Id { get; set; }
    public string DataReference { get; set; } // Reference to chunked data
}

// Store large data separately in chunks
await cache.InsertObject("data_chunk_1", chunk1);
await cache.InsertObject("data_chunk_2", chunk2);

7. Use UpdateExpiration for Lifetime Extension

// ❌ Expensive - Re-serializes entire object
var data = await cache.GetObject<LargeObject>("key");
await cache.InsertObject("key", data, newExpiration);

// ✅ Efficient - Only updates metadata (up to 250x faster)
await cache.UpdateExpiration("key", newExpiration);

BSON for Performance

When to Use BSON

  • Large objects - BSON is more efficient for complex data structures
  • Binary data - Better handling of byte arrays and binary content
  • Performance critical - BSON serialization is typically 20-40% faster for large objects
// System.Text.Json BSON (recommended)
AppBuilder.CreateSplatBuilder()
    .WithAkavacheCacheDatabase<SystemJsonBsonSerializer>(builder => /* ... */);

// Newtonsoft.Json BSON (for V10 compatibility)
AppBuilder.CreateSplatBuilder()
    .WithAkavacheCacheDatabase<NewtonsoftBsonSerializer>(builder => /* ... */);

Memory Optimization

1. Monitor Memory Usage

public class CacheMemoryMonitor
{
    public static void LogMemoryUsage(string operation)
    {
        var beforeGC = GC.GetTotalMemory(false);
        GC.Collect();
        var afterGC = GC.GetTotalMemory(true);
        
        Console.WriteLine($"{operation}: {beforeGC:N0} -> {afterGC:N0} bytes");
    }
}

2. Clear Unnecessary Data

// Periodically clean up expired entries
await cache.Vacuum();

// Clear specific data types when no longer needed
await cache.InvalidateAllObjects<TemporaryData>();

// Clear InMemory cache during memory pressure
if (MemoryPressure.IsHigh())
{
    await CacheDatabase.InMemory.InvalidateAll();
}

3. Use Weak References for Large Objects

public class SmartCacheWrapper<T> where T : class
{
    private readonly WeakReference<T> _weakRef = new(null);
    private readonly string _cacheKey;
    private readonly IBlobCache _cache;
    
    public async Task<T> GetObject()
    {
        if (_weakRef.TryGetTarget(out var cached))
            return cached;
            
        var data = await _cache.GetObject<T>(_cacheKey);
        _weakRef.SetTarget(data);
        return data;
    }
}

Performance Monitoring

1. Measure Cache Hit Rates

public class CacheMetrics
{
    private long _hits = 0;
    private long _misses = 0;
    
    public double HitRate => _hits / (double)(_hits + _misses);
    
    public async Task<T> GetWithMetrics<T>(string key)
    {
        try
        {
            var result = await cache.GetObject<T>(key);
            Interlocked.Increment(ref _hits);
            return result;
        }
        catch (KeyNotFoundException)
        {
            Interlocked.Increment(ref _misses);
            throw;
        }
    }
}

2. Track Operation Performance

public class PerformanceTracker
{
    public static async Task<T> TimeOperation<T>(string operationName, Func<Task<T>> operation)
    {
        var stopwatch = Stopwatch.StartNew();
        try
        {
            return await operation();
        }
        finally
        {
            stopwatch.Stop();
            Console.WriteLine($"{operationName}: {stopwatch.ElapsedMilliseconds}ms");
        }
    }
}

// Usage
var data = await PerformanceTracker.TimeOperation("Cache Get", 
    () => cache.GetObject<MyData>("key"));

3. Monitor SQLite Database Size

public class CacheSizeMonitor
{
    public static long GetDatabaseSize(string dbPath)
    {
        if (File.Exists(dbPath))
            return new FileInfo(dbPath).Length;
        return 0;
    }
    
    public static void LogCacheSizes()
    {
        var basePath = CacheDatabase.GetCacheDirectory();
        
        Console.WriteLine($"UserAccount: {GetDatabaseSize(Path.Combine(basePath, "userAccount.db")):N0} bytes");
        Console.WriteLine($"LocalMachine: {GetDatabaseSize(Path.Combine(basePath, "localMachine.db")):N0} bytes");
        Console.WriteLine($"Secure: {GetDatabaseSize(Path.Combine(basePath, "secret.db")):N0} bytes");
    }
}

Benchmarking Your Application

1. Benchmark Framework

[MemoryDiagnoser]
[SimpleJob(RuntimeMoniker.Net90)]
public class CacheBenchmarks
{
    private IBlobCache _cache;
    private TestData[] _testData;
    
    [GlobalSetup]
    public void Setup()
    {
        _cache = new InMemoryBlobCache(new SystemJsonSerializer());
        _testData = GenerateTestData(1000);
    }
    
    [Benchmark]
    public async Task InsertObjects()
    {
        var dict = _testData.ToDictionary(d => d.Id.ToString(), d => d);
        await _cache.InsertObjects(dict);
    }
    
    [Benchmark]
    public async Task GetObjects()
    {
        var keys = _testData.Select(d => d.Id.ToString()).ToArray();
        var results = await _cache.GetObjects<TestData>(keys);
    }
}

2. Run Benchmarks

# Install BenchmarkDotNet
dotnet add package BenchmarkDotNet

# Run benchmarks
dotnet run -c Release --project YourBenchmarkProject

Known Performance Limitations

1. Large Databases with Newtonsoft.Json

  • Issue: V10 outperforms V11 when using legacy Newtonsoft serialization with very large datasets
  • Solution: Use System.Text.Json for better performance

2. Sequential Read Performance

  • Issue: Up to 8.6% slower than V10 specifically when using Newtonsoft.Json serializer
  • Solution: System.Text.Json does not have this limitation and performs better than V10

3. Cross-Platform Differences

  • Issue: Performance varies between platforms due to different SQLite implementations
  • Solution: Test on target platforms and optimize accordingly

Performance Reports

For comprehensive performance analysis:

Reproducing Benchmarks

Platform Requirements

Benchmark reproduction requires Windows hosts. Linux/macOS are not supported due to Windows-specific dependencies.

Prerequisites

  • .NET 9.0 SDK
  • Windows operating system
  • PowerShell 5.0+ (for automation script)

Running Performance Benchmarks

# V11 benchmarks (current)
cd src
dotnet run -c Release -p Akavache.Benchmarks/Akavache.Benchmarks.csproj

# V10 comparison benchmarks  
dotnet run -c Release -p Akavache.Benchmarks.V10/Akavache.Benchmarks.V10.csproj

Running Compatibility Tests

# From the solution root directory
.\src\RunCompatTest.ps1

Performance Best Practices Summary

  1. Use System.Text.Json for optimal performance in V11
  2. Batch operations when working with multiple items
  3. Choose appropriate cache types based on access patterns
  4. Set reasonable expiration times to balance freshness and performance
  5. Monitor cache hit rates and adjust strategies accordingly
  6. Use UpdateExpiration instead of re-inserting data when extending lifetime
  7. Break large objects into smaller, manageable chunks
  8. Clear unnecessary data regularly to maintain performance
  9. Monitor memory usage and implement pressure relief mechanisms
  10. Test on target platforms as performance characteristics vary

Next Steps