The Problem: Computational Overhead in Traditional Web Servers

┌───────────────────────────────────────────────────┐
│                 TRADITIONAL WEB SERVER                 │
├───────────────────────────────────────────────────┤
│                                                        │
│  Request 1 Arrives                                     │
│     ↓                                                  │
│  Load Heavy Libraries (Unicode tables, NLP models...)  │ ← EXPENSIVE!
│     ↓                                                  │
│  Initialize Data Structures                            │ ← EXPENSIVE!
│     ↓                                                  │
│  Process Request (Finally!)                            │ ← TINY PART
│     ↓                                                  │
│  Generate Response                                     │
│     ↓                                                  │
│  Cleanup & Unload Everything                           │ ← WASTEFUL!
│                                                        │
│  Request 2 Arrives... REPEAT THE WHOLE CYCLE!          │
│                                                        │
└───────────────────────────────────────────────────┘

The Core Problem: You spend 80% of your time loading libraries and only 20% doing actual computation. For a text analysis API using powerful libraries like Softanza, this means:

• Loading Unicode normalization tables on every request
• Reinitializing pattern matching engines
• Reconstructing language models
• Rebuilding object hierarchies

The stzAppServer Solution: Persistent Computational Power

stzAppServer flips this paradigm completely. Instead of treating computational libraries as disposable per-request resources, it treats them as persistent computational engines:

+--------------------------------------------------------+
│                    STZAPPSERVER                        │
+--------------------------------------------------------+
│                                                        │
│   +------------------------------------------------+   │
│   │         PERSISTENT ENGINE (Always Ready)       │   │
│   │  • Unicode Tables Loaded                       │   │
│   │  • Softanza Object System Active               │   │
│   │  • Language Models Initialized                 │   │
│   │  • Pattern Engines Ready                       │   │
│   +------------------------------------------------+   │
│                          ↑                             │
│                          │                             │
│  Request 1 -> Parse -> Route -> Execute -> Response    │ ← FAST!
│  Request 2 -> Parse -> Route -> Execute → Response     │ ← FAST!
│  Request 3 -> Parse -> Route -> Execute → Response     │ ← FAST!
│                                                        │
└───────────────────────────────────────────────────┘

Architecture Overview

stzAppServer consists of four core components working together:

    ┌-----------------------------------┐
    │            stzAppServer           │
    │         (Main Orchestrator)       │
    └-----------------┬-----------------┘
                      │
    ┌-----------------┼-----------------┐
    │                 │                 │
    │   ┌-------------v-------------┐   │
    │   │      stzComputeEngine     │   │
    │   │   (Persistent Softanza)   │   │
    │   └-------------┬-------------┘   │
    │                 │                 │
    │   ┌-------------v-------------┐   │
    │   │      stzContextPool       │   │
    │   │   (Memory Management)     │   │
    │   └-------------┬-------------┘   │
    │                 │                 │
    │   ┌-------------v-------------┐   │
    │   │       stzAppRouter        │   │
    │   │    (Request Routing)      │   │
    │   └-------------┬-------------┘   │
    │                 │                 │
    │   ┌-------------v-------------┐   │
    │   │     stzReactiveSystem     │   │
    │   │   (Non-blocking I/O)      │   │
    │   └---------------------------┘   │
    │                                   │
    └-----------------------------------┘

Understanding Key Concepts

What is a "Context"?

A context is an isolated execution environment where a request gets processed. Think of it as a dedicated workspace that contains:

• Memory space for variables and objects
• Access to the pre-loaded Softanza engine
• Isolated state that doesn't interfere with other requests
• Resource limits (memory, execution time)

haro

# Each context works like this:
oContext = new stzProcessingContext()
oContext {
    # Has access to pre-loaded Softanza
    oStr = new stzString("Hello World")  # Instant access!
    
    # Isolated variables - won't affect other contexts
    cResult = oStr.Reversed()
    
    # Clean up happens automatically
}

Why contexts matter:

• Isolation: Each request runs independently
• Performance: No library loading overhead
• Concurrency: Multiple requests processed simultaneously
• Resource Control: Memory and CPU limits per request

What is a "Pool"?

A context pool is a collection of pre-created contexts that are reused across requests. Instead of creating/destroying contexts for each request, the pool manages a fixed number of reusable contexts.

Context Pool Management:
┌-------------------------------------┐
│  Available Contexts   │  Active     │
│  +-----+  +-----+     │  +-----+    │
│  | Ctx |  | Ctx |     │  | Ctx |    │
│  |  3  |  |  4  |     │  |  1  |    │
│  +-----+  +-----+     │  +-----+    │
│                       │             │
│                       │  +-----+    │
│                       │  | Ctx |    │
│                       │  |  2  |    │
│                       │  +-----+    │
└-------------------------------------┘

Flow:
1. Request arrives → Borrow available context
2. Process request in borrowed context  
3. Return context to available pool

Pool benefits:

• Performance: No context creation/destruction overhead
• Memory efficiency: Fixed memory footprint
• Resource control: Limit concurrent processing
• Scalability: Handle thousands of requests with limited resources

Component Breakdown

1. stzAppServer _ the Main Orchestrator

TODO

2. stzComputeEngine - The Persistent Brain This component loads Softanza once at startup and keeps it ready:

oComputeEngine = new stzComputeEngine()
oComputeEngine {
    PreloadStringEngine()    # Unicode tables, patterns ready
    PreloadObjectSystem()    # Class definitions cached
    PreloadCollections()     # Data structures initialized
    PreloadNLP()             # Language models loaded
}

3. stzContextPool - Smart Memory Management Manages reusable execution contexts:

oPool = new stzContextPool()
oPool {
    CreateContexts(20)           	 # Pre-create 20 contexts
    SetContextMemoryLimit("100MB")   # Limit per context
    SetContextTimeout(30)            # Max 30 seconds per request
}

4. stzAppRouter - …

TODO

5. stzReactiveSystem - Non-blocking Processing Built on event-driven architecture for handling thousands of concurrent requests without blocking:

oReactive = new stzReactiveSystem()
oReactive {
    SetMaxConcurrentRequests(5000)
    SetIOThreads(4)
    EnableKeepAlive()
}

Real-World Usage Examples

Basic Text Processing API

# Create the server
app = new stzAppServer()

# Simple route with immediate Softanza access
app.Get("/analyze", func oRequest, oResponse {
    cText = oRequest.Query("text")
    
    # Full Softanza power instantly available!
    oStr = new stzString(cText)
    
    aResult = [
        :length = oStr.NumberOfChars(),
        :words = oStr.Words(),
        :reversed = oStr.Reversed(),
        :languages = oStr.DetectedLanguages(),
        :sentiment = oStr.SentimentAnalysis()
    ]
    
    oResponse.Json(aResult)
})

app.Start(3000)  # Server ready with Softanza loaded!

Advanced Linguistic Processing

app = new stzAppServer()

app.Post("/linguistic-analysis", func oRequest, oResponse {
    cDocument = oRequest.Body()
    
    # Complex analysis using persistent Softanza engine
    oText = new stzText(cDocument)
    
    aAnalysis = [
        :statistics = [
            :characters = oText.NumberOfChars(),
            :words = oText.NumberOfWords(),
            :sentences = oText.NumberOfSentences(),
            :readability = oText.ReadabilityScore()
        ],
        :structure = [
            :headings = oText.Headings(),
            :quotes = oText.QuotedSubStrings(),
            :urls = oText.URLs(),
            :emails = oText.EmailAddresses()
        ],
        :linguistic = [
            :language = oText.DetectedLanguage(),
            :keywords = oText.ExtractedKeywords(),
            :entities = oText.NamedEntities(),
            :topics = oText.TopicAnalysis()
        ]
    ]
    
    oResponse.Json(aAnalysis)
})

app.Start(8080)

Performance Comparison

Traditional Approach (Per-Request Loading)

Request Timeline:
┌------------------------------------------------------------┐
│ Load Libs | Init Data | Process | Cleanup |  Response      │
│   800ms   |   200ms   |  50ms   |  100ms  |    5ms         │
│           |           |         |         |                │
│ <--------------------- 1,155ms total --------------------> │
│                                           │                │
│ Only 50ms of actual computation!          │                │
└------------------------------------------------------------┘

stzAppServer Approach (Persistent Engine)

Request Timeline:
┌-----------------------------------------------------------┐
│ Parse | Route | Execute |  Response                       │
│  2ms  |  1ms  |  50ms   |    5ms                          │
│       |       |         |                                 │
│ <------------ 58ms total ---------->                      │
│                         │                                 │
│ 20x faster response!    │                                 │
└-----------------------------------------------------------┘

What Problems Does stzAppServer Solve?

1. Computational API Performance

Traditional web servers are optimized for simple CRUD operations. When you need rich text processing, complex algorithms, or linguistic analysis, the overhead becomes crushing.

Before stzAppServer: - Text analysis API: 1.2 seconds per request - 90% overhead, 10% actual work - Poor scalability under load

With stzAppServer: - Same analysis: 60ms per request - 20x performance improvement - Scales to thousands of concurrent requests

2. Resource Waste

Loading the same heavy libraries thousands of times per day wastes: - CPU cycles on initialization - Memory allocation/deallocation churn - Disk I/O for library loading - Network latency under load

3. Developer Frustration

Powerful libraries become "too expensive" to use in web APIs, forcing developers to: - Choose weaker but faster alternatives - Implement caching layers - Pre-process data offline - Compromise on functionality

How stzAppServer Differs from Alternatives

vs. Traditional Web Servers (Apache, Nginx + PHP/Python/Ruby)

Traditional Server:
┌-----------------------------------------┐
│  Request > Load > Process > Unload      │
│             ^                 ^         │
│             |                 |         │
│         EXPENSIVE         WASTEFUL      │
└-----------------------------------------┘

stzAppServer:
┌-----------------------------------------┐
│     Request > Process (Engine Ready)    │
│                 ^                       │
│                 |                       │
│             EFFICIENT                   │
└-----------------------------------------┘

vs. Application Servers (Tomcat, .NET, etc.)

While application servers do keep some components loaded, they're not optimized for computational workloads:

Traditional Application Servers: - Focus on business logic and database connectivity - Computational libraries still loaded per application, not per server - Not designed for rich text/linguistic processing

stzAppServer: - Purpose-built for computational workloads - Softanza engine is the first-class citizen - Every request has immediate access to full computational power

vs. Specialized Solutions (Wolfram, Jupyter, etc.)

┌------------------┬------------------┬------------------┐
│      Wolfram     │     Jupyter      │   stzAppServer   │
├------------------┼------------------┼------------------┤
│ + Computational  │ + Interactive    │ + Computational  │
│ x Proprietary    │ x Not for APIs   │ + Open Source    │
│ x Expensive      │ x Single User    │ + Free           │
│ x Math-focused   │ x Notebook UI    │ + General Web    │
│ x Complex        │ x No Scaling     │ + Simple & Fast  │
└------------------┴------------------┴------------------┘

Use Cases Where stzAppServer Excels

1. Natural Language Processing APIs

# Instant access to powerful NLP without loading overhead
app.Post("/nlp", func oRequest, oResponse {
    oText = new stzText(oRequest.Body())
    oResponse.Json([
        :entities = oText.NamedEntities(),
        :sentiment = oText.SentimentAnalysis(),
        :summary = oText.GenerateSummary(),
        :translation = oText.TranslateTo("french")
    ])
})

2. Real-time Text Analysis

Perfect for applications requiring immediate text processing: - Content moderation systems - Live translation services
- Search query analysis - Social media monitoring

3. Educational Platforms

Rich linguistic analysis for: - Grammar checking - Writing assistance - Language learning - Literature analysis

4. Business Intelligence

Text mining and analysis for: - Document classification - Customer feedback analysis - Market research - Compliance monitoring

Configuration and Optimization

Pool Configuration

app = new stzAppServer()
app.ConfigureContextPool([
    :size = 50,                     	# Number of contexts in pool
    :memoryLimitPerContext = "200MB",  	# Memory limit per context
    :timeoutSeconds = 60,              	# Max execution time
    :preloadEngines = true             	# Pre-initialize Softanza in each context
])

Performance Tuning

app.ConfigurePerformance([
    :maxConcurrentRequests = 10000,
    :ioThreads = 8,
    :keepAliveTimeout = 30,
    :compressionEnabled = true,
    :cachingEnabled = true
])

Monitoring and Health Checks

app.Get("/health", func oRequest, oResponse {
    oResponse.Json([
        :status = "healthy",
        :uptime = app.Uptime(),
        :requests_served = app.RequestCount(),
        :active_connections = app.ActiveConnections(),
        :context_pool = [
            :total = app.ContextPoolQ().TotalCount(),
            :available = app.ContextPoolQ().AvailableCount(),
            :active = app.ContextPoolQ().ActiveCount()
        ],
        :memory_usage = app.MemoryUsage(),
        :cpu_usage = app.CPUUsage()
    ])
})

Getting Started

Basic Setup

# Load the framework
load "stzAppServer.ring"

# Create your server
app = new stzAppServer()

# Define routes with full Softanza power
app.Get("/", func oRequest, oResponse {
    oResponse.Json([
        :message = "Welcome to stzAppServer!",
        :engine = "Softanza Persistent Computational Engine",
        :status = "Ready for complex computations"
    ])
})

# Start the server
app.Start(3000)
? "Server running at http://localhost:3000"

Advanced Configuration

app = new stzAppServer()

# Configure the server for production
app {
    SetPort(8080)
    SetMaxConnections(5000)
    EnableSSL("cert.pem", "key.pem")
    SetLogLevel("production")
    
    # Configure context pool
    ConfigureContextPool([
        :size = 100,
        :memoryLimit = "500MB",
        :timeout = 30
    ])
    
    # Add middleware
    UseMiddleware("cors")
    UseMiddleware("compression")
    UseMiddleware("logging")
}

The Future of Computational Web Applications

stzAppServer represents a paradigm shift from request-response web servers to persistent computational engines. This opens possibilities for:

• AI-powered APIs that don't sacrifice performance for intelligence
• Real-time linguistic processing at web scale
• Rich educational platforms with instant feedback
• Advanced text analysis accessible to any application
• Reliable and higly reatcive entreprise platforms ready for critical business scenarios

By treating computational power as a persistent asset rather than a per-request burden, stzAppServer enables a new class of web applications that were previously impractical due to performance constraints.

The age of choosing between "fast" and "smart" in web APIs is over. With stzAppServer, you can have both.

stzAppServer: Where computational richness meets web-scale performance.

Citer

Mansour Ayouni. « stzAppServer: The Persistent Computational Web Engine Revolution ». Softanza narrations, 2025. https://mayouni.github.io/stzsite/fr/narrations/stzappserver-article.html

Non exécuté : son code ne se charge pas tel qu'il est écrit. L'article est publié tel qu'il est écrit ; ses blocs sont des illustrations. Le code des articles est du code Haro. Là où un article a été écrit avant que la langue prenne son nom actuel, le nom a été changé et rien d'autre. Genre, domaine et série lus dans le nom du fichier, année tirée du premier commit : dérivés, jusqu'à ce que l'article porte son propre en-tête. Le fichier sur GitHub(s'ouvre dans un nouvel onglet)