Information systems architectures: models and real-world cases
Pierre Henriet — Architectural guide analyzing monoliths, SOA, microservices, event-driven systems (EDA), and composable models through verified 2023-2025 European cases (Adyen, Michelin, SNCF Connect, France Travail, NHS, Spotify, Zara).
In enterprise engineering, software architecture and information systems (IS) architecture are often conflated. Software architecture organizes an individual application: its internal code structure, design patterns, and execution speed. In contrast, information systems architecture orchestrates an enterprise's broader socio-technical fabric: business capabilities, data governance, multi-vendor systems (ERP, CRM, SCM), legacy mainframes, integration backbones, and public interoperability frameworks.
Selecting the wrong IS architecture creates lasting technical debt, operational friction, and business misalignment. From high-throughput modular monoliths to event-driven meshes and composable packaged capabilities, each paradigm embodies distinct engineering trade-offs. This guide examines the primary information systems architectures and reviews their execution across European enterprises and public institutions (including Adyen, Michelin, SNCF Connect, France Travail, NHS England, Spotify, and Zara) through recent 2023 to 2025 implementations.
1. Theoretical foundations: what defines an information systems architecture?
An information system is a socio-technical structure designed to collect, process, store, and distribute information for enterprise decision-making and operational control. Three foundational frameworks govern modern IS architecture:
- The TOGAF standard (10th edition) BDAT model: The Open Group Architecture Framework organizes enterprise architecture into four connected layers:
• Business architecture: Strategy, governance, organizational structure, and core operational processes.
• Data architecture: Physical and logical data assets, master data management (MDM), and data movement.
• Application architecture: Software portfolio blueprint, interface contracts, and relationships to business processes.
• Technology architecture: Multi-cloud infrastructure, networking, middleware, and cybersecurity topology. - Conway's law and the inverse Conway maneuver: Formulated by Melvin Conway in 1968: "Organizations which design systems are constrained to produce designs which are copies of the communication structures of these organizations." Modern engineering applies the inverse Conway maneuver: organizing small, autonomous cross-functional teams to naturally induce decoupled technical architectures.
- The Zachman framework: A two-dimensional ontology mapping six stakeholder perspectives (planner, owner, designer, builder, subcontractor, functioning enterprise) across six core interrogatives (what, how, where, who, when, why).
2. Comparative paradigm matrix: core IS architectures
The table below summarizes the technical, operational, and organizational characteristics of the five primary architectural models:
| Architecture paradigm | Coupling and communication | Data management model | Deployment and scalability | Primary operational trade-off |
|---|---|---|---|---|
| 1. Monolithic (classic and modular) | Tightly coupled; in-process memory calls; direct function execution | Single unified RDBMS; ACID transactions; shared schema | Single build artifact; vertical scaling; entire app deployed together | Rapid initial velocity vs. blast radius and organizational bottlenecks at scale |
| 2. Service-oriented architecture (SOA) | Loosely coupled via enterprise service bus (ESB); SOAP/WSDL; canonical XML schemas | Siloed legacy data stores; centralized data transformation in the bus | Distributed coarse services; coordinated enterprise release cycles | System interoperability vs. ESB becoming a complex, brittle monolithic bottleneck |
| 3. Microservices architecture | Decoupled bounded contexts; lightweight APIs (REST / gRPC); API gateways | Database-per-service; polyglot persistence (SQL, NoSQL, cache); eventual consistency | Independent CI/CD pipelines; horizontal auto-scaling via containers (Kubernetes) | High team autonomy vs. distributed operational overhead (tracing, latency, partial failures) |
| 4. Event-driven architecture (EDA) | Temporally and spatially decoupled; publish/subscribe; distributed event logs | Immutable event streams; CQRS; materialized read projections and state stores | Independent consumer scaling; asynchronous non-blocking real-time processing | Sub-second responsiveness vs. eventual consistency and complex replay semantics |
| 5. Composable enterprise and MACH | API-first; headless frontends; packaged business capabilities (PBCs); cloud-native SaaS | Domain data products; federated data mesh; specialized SaaS persistence | Modular plug-and-play components; continuous independent SaaS release cycles | Rapid business adaptability vs. multi-vendor governance and contract fragmentation |
3. Deep dive into architectural models and verified industry case studies
A. The monolith: why high-performing European leaders still rely on it
While microservices dominate tech discussions, monoliths remain efficient and cost-effective when built with clear boundaries. Monolithic systems run in a single process, sharing memory and a database.
Enterprises increasingly favor the modular monolith: an approach enforcing strict boundaries and domain isolation at compile time within one codebase, avoiding distributed network latency.
- European case study — Adyen (Netherlands, 2023–2024): Amsterdam-based fintech leader Adyen processed over €970 billion in 2023 and surpassed €1.2 trillion in 2024. Adyen maintains a single optimized monolithic Java core on PostgreSQL. By keeping financial transactions in-process, Adyen secures instant ACID consistency, sub-millisecond execution, and eliminates distributed network failure modes across global card networks.
- Public sector case study — UK Government Digital Service (GOV.UK, 2023–2024): Serving over 500 million visits annually, the UK central portal relies on monolithic publishing engines. While integrating the 2024 One Login digital identity service, GOV.UK preserved its monolithic core for official content to ensure atomic publishing, strict accessibility (WCAG 2.1 AA), and 99.99% availability during national crises.
B. Service-oriented architecture (SOA): enterprise and public sector integration
SOA solved enterprise fragmentation in the 2000s by linking disparate ERPs, mainframes, and CRM systems. It introduced coarse-grained business services exposed over standardized protocols (SOAP, WSDL) and coordinated through an enterprise service bus (ESB).
- European case study — SNCF Connect and Tech (France, 2023–2024): Managing French passenger rail requires connecting legacy reservation systems (Resarail / Amadeus) with real-time seat availability and mobile ticketing. SNCF operates a modernized SOA and API mediation platform handling over 1.3 billion visits and 170 million tickets annually, coordinating transactional inventory between high-speed trains and European partner platforms.
- Public sector case study — France Travail (France, 2024): Established in January 2024 to succeed Pôle Emploi, France Travail deployed its national employment platform ("Réseau pour l'Emploi"). Built on an API platform and SOA mediation backbone, the agency links legacy relational databases with social security services (URSSAF) and sovereign authentication (FranceConnect) to calculate allowances and coordinate job matching in real time.
- European public sector case study — European Digital Identity Architecture (2024–2025): Under the eIDAS 2.0 regulation adopted in 2024, the European Commission developed a federated interoperability architecture. It connects national identity registries, banks, and public administrations across 27 EU member states using open mediation buses and standardized verifiable credentials.
- The structural flaw of classic SOA: Teams often embedded complex business logic into the bus, creating "smart pipes and dumb endpoints." The ESB turned into an unmaintainable bottleneck managed by an overburdened integration team.
C. Microservices architecture: decentralized domain autonomy
Microservices emerged to resolve monolithic deployment bottlenecks and centralized ESB delays. Systems decompose into small, autonomous services structured around bounded contexts from Eric Evans' domain-driven design (DDD). Each service owns its database and communicates via lightweight protocols (REST, gRPC) or asynchronous events.
- European case study — Spotify (Sweden, 2023–2024): An early adopter of autonomous engineering squads and microservices, Spotify advanced toward platform engineering to tame microservice sprawl. To govern thousands of internal services across Stockholm, London, and New York, Spotify developed and open-sourced Backstage (now a graduated CNCF project). Backstage provides a centralized software catalog and developer portal that restores architectural visibility and API ownership without restricting team autonomy.
- Public media case study — BBC iPlayer (UK, 2023–2024): To absorb live-streaming peaks — including the 2023 Eurovision Song Contest, the King Charles III coronation, and 2024 election coverage —, the BBC decoupled its broadcasting core into serverless cloud microservices on AWS. Ingestion, recommendations, and video delivery scale independently, serving over 7.5 billion streaming requests across 2023 and 2024 without service degradation.
D. Event-driven architecture (EDA): asynchronous real-time streaming
Synchronous request-response chains (REST/HTTP) create tight temporal coupling: if one service slows down, the entire sequence stalls. Event-driven architecture decouples systems in time and space by publishing immutable events to an append-only distributed log (such as Apache Kafka, Apache Pulsar, or AWS Kinesis).
- European case study — Michelin (France, 2023–2025): Michelin transformed manufacturing and logistics into an event-driven system on Apache Kafka. Sensor telemetry from European tire plants, automated warehouse robots, and supply chain hubs stream continuous events. Subscribed microservices process inventory movements, predictive machinery maintenance, and dispatch schedules in real time, minimizing production line stoppages.
- European financial case study — European Central Bank (ECB) TARGET Instant Payment Settlement (TIPS, 2023–2024): Settling real-time payments across the eurozone requires instant, non-blocking execution. In preparation for the 2024-2025 EU instant payment mandate, TIPS operates 24/7/365, settling transactions in central bank funds in under 0.2 seconds using event-driven streaming and ISO 20022 messaging across eurozone central and commercial banks.
E. The composable enterprise and MACH architecture
Defined by Gartner, the composable enterprise delivers agility by assembling modular packaged business capabilities (PBCs) instead of customizing monolithic ERP suites. Composable architectures follow the MACH principles:
- M — Microservices: Independent, modular functional blocks.
- A — API-first: All services exposed through documented, standardized APIs (REST, GraphQL).
- C — Cloud-native SaaS: Elastic, managed cloud infrastructure eliminating local server maintenance.
- H — Headless: Full separation between user frontends (web, mobile, retail kiosks) and backend transactional engines.
European case study — Carrefour (France, 2023–2024): Carrefour accelerated its composable transformation by decoupling consumer frontends from transactional core systems. Combining a Google Cloud Data Mesh with headless e-commerce modules, Carrefour assembles independent capabilities for rapid delivery, loyalty rewards, and warehouse logistics, rolling out store features without modifying central accounting systems.
European case study — Inditex / Zara (Spain, 2023–2024): Inditex powers global retail agility through its proprietary Inditex Open Platform (IOP). By linking in-store RFID tags with headless checkout and cloud logistics APIs, Zara synchronizes stock across 5,800+ stores with online storefronts, fulfilling orders from the nearest local store within 24 hours.
Public healthcare case study — NHS England (UK, 2023–2025): NHS England is rolling out its Federated Data Platform (FDP). Awarded in late 2023 and deployed across 2024 and 2025, the program uses open healthcare data standards (HL7 FHIR) and modular clinical micro-capabilities to interconnect regional hospital trusts and community care without replacing core electronic patient record (EPR) databases.
4. The IS decision framework: practical guide for project managers and architects
Choosing an information systems architecture is a strategic business decision that shapes team structure, capital costs, and delivery speed. IS project managers should apply four rules:
- Adopt Martin Fowler's "MonolithFirst" heuristic: Unless domain boundaries are stable and the team is already large, starting with microservices incurs the "microservice premium" (distributed tracing, network latency, and data reconciliation overhead) without delivering benefits. Begin with a well-structured modular monolith; extract autonomous services only when specific sub-domains face distinct scaling or deployment demands.
- Account for the fallacies of distributed computing: Formulated by L. Peter Deutsch and James Gosling at Sun Microsystems, these eight false assumptions disrupt distributed systems:
• The network is reliable.
• Latency is zero.
• Bandwidth is infinite.
• The network is secure.
• Topology does not change.
• There is one administrator.
• Transport cost is zero.
• The network is homogeneous.
Every inter-service call introduces failure probabilities that require circuit breakers, exponential backoff retries, idempotent consumers, and dead-letter queues. - Align architecture with team topologies: Managing 50 microservices fails with a single horizontal pool of developers. Microservices and composable systems require stream-aligned teams owning an end-to-end business flow, supported by platform teams offering self-service infrastructure.
- Enforce contract-driven API governance: Without automated tooling, distributed systems quickly drift into inconsistency. Adopt a contract-first approach (OpenAPI for REST, Protocol Buffers for gRPC, AsyncAPI for events) and automate backward compatibility testing in CI/CD pipelines to prevent breaking changes.
AI Transparency
This article was structured and drafted by Pierre Henriet using AI assistance for research synthesis and terminology alignment. All analyses, strategic conclusions, and professional insights are directly curated and validated by the author.
Sources and references:
• The Open Group (2022) — The TOGAF Standard, 10th Edition: Enterprise Architecture Methodology and BDAT Model
• Zachman, John A. (1987) — "A framework for information systems architecture", IBM Systems Journal, Vol. 26, No. 3
• Conway, Melvin E. (1968) — "How Do Committees Invent?", Datamation Magazine
• Deutsch, L. Peter and Gosling, James (1994–1997) — The Eight Fallacies of Distributed Computing (Sun Microsystems)
• Fowler, Martin (2014) — "Microservices: a definition of this new architectural term" and "MonolithFirst" (martinfowler.com)
• Evans, Eric (2003) — Domain-Driven Design: Tackling Complexity in the Heart of Software (Addison-Wesley)
• Hohpe, Gregor and Woolf, Bobby (2003) — Enterprise Integration Patterns: Designing, Building, and Deploying Messaging Solutions (Addison-Wesley)
• Skelton, Matthew and Pais, Manuel (2019) — Team Topologies: Organizing Business and Technology Teams for Fast Flow (IT Revolution)
• Gartner Research (2020/2023) — Future of Applications: Delivering the Composable Enterprise with Packaged Business Capabilities
• Adyen Engineering (2023/2024) — Architectural Principles: Why Adyen Operates on a Scalable Core Monolith
• UK Government Digital Service (2023/2024) — GOV.UK Architectural Foundations and One Login Architecture
• SNCF Connect and Tech (2023/2024) — High-Volume Railway Distribution and API Mediation Architecture
• France Travail (2024) — Réseau pour l'Emploi and Open API Architecture for National Public Employment
• European Commission (2024) — European Digital Identity Architecture and Reference Framework (eIDAS 2.0)
• Spotify Engineering (2023/2024) — Managing Microservices Complexity with Backstage and Platform Engineering
• BBC Research and Development (2023/2024) — Cloud-Native Serverless Microservices for BBC iPlayer Live Streaming
• Michelin Technology (2023/2025) — Event-Driven Architecture at Scale with Apache Kafka in Manufacturing
• European Central Bank (2023/2024) — TARGET Instant Payment Settlement (TIPS) Technical Architecture and Specifications
• Carrefour Tech (2023/2024) — Composable Commerce and Google Cloud Data Mesh Architecture
• NHS England (2023/2025) — Federated Data Platform (FDP) and HL7 FHIR Interoperability Architecture
• Inditex Tech (2023/2024) — Inditex Open Platform (IOP) Composable Architecture for Global Retail
Planning an information systems evolution?
Let's discuss your strategic alignment, compliance, or software deployment goals.
Comments (0)