Introducing the Micro Command Macro Query (MCMQ) Pattern

Elevating Command and Query Management in Microservices

Mehmed Ali Çalışkan
14 min readFeb 13, 2024


Introduction: Addressing the Gap in Event Sourcing and CQRS

The adoption of Event Sourcing and Command Query Responsibility Segregation (CQRS) patterns marks a significant evolution in handling operations within applications, especially in the realm of microservices. These patterns excel in cleanly separating the responsibilities of commands — such as create, update, and delete operations — from queries and in maintaining these as discrete event sequences. However, they stop short of addressing the intricate challenges associated with data distribution across microservices. This gap is most evident in the aggregation and presentation of the latest data state, a crucial aspect for query purposes that remains underserved by current methodologies.

Traditional approaches to this challenge have struggled to provide a holistic solution for the effective aggregation of distributed data, essential for achieving a comprehensive and up-to-date view of data across services. Enter the Micro Command Macro Query (MCMQ) pattern: a pioneering architectural pattern conceived to fill this void. MCMQ extends the principles of CQRS in microservices by introducing a refined strategy for the storage and aggregation of data. It proposes an innovative approach to manage data in distributed systems, facilitating the construction and access of the most recent data state for query purposes with unprecedented efficiency. This article delves into the MCMQ pattern, illustrating its capacity to refine command and query management in microservice architectures and presenting a solid framework to address a previously unmet need.

CQRS Pattern Explained

The Command Query Responsibility Segregation (CQRS) pattern is a foundational principle in software architecture that emphasizes the separation of operations that modify data (“commands”) from operations that retrieve data (“queries”). This separation is typically achieved by using distinct databases: one optimized for write operations (create, update, delete) and another for read operations (queries). Such an approach not only simplifies the design and optimization of each database for its specific workload but also facilitates the implementation of these responsibilities as separate services. This segregation aligns naturally with the principles of microservices architecture, where the division of responsibilities and scalability are paramount. Consequently, CQRS is particularly popular in microservices due to its alignment with the architecture’s emphasis on responsibility segregation and distributed systems design.

Ensuring Consistency in CQRS Implementations

In the context of the Command Query Responsibility Segregation (CQRS) pattern, the write data store serves as the source of truth (SoT). To ensure data consistency, the read data store is synchronized with the write data store as closely as possible. This synchronization is facilitated by various replication technologies, which bridge the gap between the two data stores. These technologies are designed to replicate data in real-time or near-real-time, ensuring that the read data store reflects the most current state of the write data store. This approach allows for the efficient segregation of command and query operations while maintaining data integrity across the system.

A typical CQRS architecture

Data Replication via Event-Driven Architecture in CQRS

In the Command Query Responsibility Segregation (CQRS) architecture, synchronizing the source of truth (SOT) with the read data store is a critical challenge. A prevalent method to achieve this is through an event-driven mechanism. This process begins when the write service executes a successful write operation. Subsequently, it publishes an event detailing this change. The read service, which is subscribed to these events, listens for such notifications. Upon receiving an event, the read service applies the same state change to its database, thus mirroring the update that occurred in the write data store.

To facilitate the publication and subscription of events, third-party message brokers such as Apache Kafka, RabbitMQ, and others are commonly employed. These tools provide robust, scalable, and reliable messaging systems that ensure events are efficiently transmitted from producers (write services) to consumers (read services). This approach not only maintains data consistency across different stores but also leverages the benefits of decoupling and asynchronous communication, hallmark features of a microservices architecture.

CQRS Data synchronization Through Event Driven Mechanism

Event Sourcing Pattern: Enhancing Data Integrity and History

The Event Sourcing pattern reimagines data storage not merely as a snapshot of the latest state but as a chronological series of state change events. These events, encompassing create, update, and delete operations, are stored sequentially, each capturing a distinct state transformation. To ascertain the current state of an entity, one must sequentially “replay” these events, thereby reconstructing the entity’s state over time. To optimize performance, event stores may also incorporate snapshots of data states at specific intervals, facilitating quicker access to the current state without needing to process the entire event history.

Synergy with CQRS

Event Sourcing naturally complements the Command Query Responsibility Segregation (CQRS) pattern, which divides data operations into distinct write (command) and read (query) responsibilities. This division is inherently aligned with an event-driven mechanism, making Event Sourcing an ideal fit for the write side of a CQRS-based architecture. By adopting Event Sourcing, the write data store evolves into an event store, capturing the full history of state changes. This approach offers dual benefits: it leverages the separation of concerns advocated by CQRS and provides a comprehensive history of data changes, enhancing traceability and data integrity.

When a command (create, update, delete) is issued to the write service, it is transformed into a corresponding state change event and stored in the event store. Simultaneously, this event is published to a message broker, ensuring that the change is propagated system-wide. The read service, listening for these events via the message broker, updates the read data store accordingly, ensuring it reflects the latest state. Please note that a data row in the read database is equal to a current snapshot of the data item as stored in the event store. Thus, the read data store functions as a conventional database, providing quick access to the current state, while the event store serves as the source of truth, maintaining a detailed history of all data changes.

Practical Implications

This architectural approach enhances system resilience and auditability by maintaining a detailed change log. It also facilitates complex event processing and analysis, offering insights into the evolution of data over time. Moreover, by decoupling the read and write responsibilities through CQRS and utilizing Event Sourcing for the write operations, systems gain the flexibility to scale and evolve independently, catering to diverse and changing business requirements.

When Event Sourcing pattern is used, an Event Store is located instead of a legacy write db.

Challenges of Implementing CQRS in Microservices Architectures

While the Command Query Responsibility Segregation (CQRS) pattern, in tandem with Event Sourcing, offers significant advantages for both monolithic and microservices-based applications, its implementation in a microservices environment introduces specific challenges. One of the primary issues arises from the distributed nature of microservices: data managed by separate services often need to be related, aggregated, or merged for presentation purposes or to fulfill business logic requirements. For instance, a Basket service might require confirmation that stock quantities, managed by a separate Warehouse service, are sufficient before allowing a product checkout. This necessity underscores a critical challenge: each service, operating within its own bounded context with a dedicated CQRS and Event Sourcing mechanism, may excel in isolation but struggle to communicate and share data with other services.

Consider the complexity multiplied across an ecosystem of 10 or more such services, each encapsulated within its own CQRS and Event Sourcing sandbox. The result is a highly decoupled system where services efficiently manage their internal states and histories but are hindered in their ability to interact or share data seamlessly. While I have previously outlined best practices for data sharing between microservices in another article [link], this discussion aims to introduce a novel pattern. This pattern builds on the foundational principles of Event Sourcing and CQRS but reimagines the architecture to facilitate seamless data sharing and interaction among microservices, thereby addressing one of the most pressing challenges in distributed systems design.

Evaluating the Backend for Frontend (BFF) Pattern as a Solution

The Backend for Frontend (BFF) pattern is specifically designed to address the challenges of real-time data aggregation on demand. By aggregating data from multiple backend services and preparing it for the frontend, the BFF serves as a tailored gateway, optimizing data delivery for specific user interface requirements. This approach is particularly effective for populating different sections of an application’s UI with data sourced from distinct microservices.

However, the BFF pattern encounters performance limitations in scenarios involving dynamic and complex data collation. Consider a use case where the BFF needs to fetch thousands of products from a Product service in a paginated manner and subsequently query a Warehouse service for stock quantities of each product. This process becomes even more complicated if the data from the two services needs to be correlated, such as sorting products based on their stock quantities. Achieving this requires multiple network calls and complex data merging logic at the BFF layer, leading to significant performance bottlenecks.

Bulk Join : Aggregating large sets of data that need to be joined or combined from different services can overwhelm the BFF, as it must handle substantial data volumes and complex merging logic.

Remote Criteria: Executing queries that depend on criteria stored across different services poses a challenge. For instance, filtering or sorting data based on attributes managed by another service requires the BFF to implement intricate coordination and data processing.

These scenarios reveal the BFF pattern’s struggles with efficiently handling bulk data joins and queries involving remote criteria, pointing to the need for more scalable and performant solutions in complex data aggregation cases.

The CQRS pattern together with BFF pattern. BFF pattern is good to aggregate data from read services but has important challenges.

Introducing the Micro Command and Macro Query (MCMQ) Pattern for Microservices

Microservices architecture is pivotal for distributing request loads across separate services and databases, and maintaining isolated codebases for optimal development and maintenance cycles. The transition from monolithic architectures to microservices is driven by the need for development flexibility and performance optimization, alongside the desire to prevent code and database structures from devolving into unmanageable complexities as new features are introduced. A critical aspect of microservices is the separation of databases to ensure clear, segregated data designs. However, data aggregation and sharing between services must be carefully managed to avoid inadvertently reverting to a monolithic design through excessive coupling.

MCMQ Pattern Overview

The Micro Command and Macro Query (MCMQ) pattern is designed to uphold the segregation intrinsic to microservices architecture while facilitating data aggregation for the presentation layer. This pattern draws a parallel with Command Query Responsibility Segregation (CQRS) but introduces a novel component: the view store, in contrast to the traditional read database.

- View Store Concept: The view store functions analogously to relational database views but diverges in its management of a static, rather than dynamic, view. This static view is a JSON document that is continuously updated in response to state change events from the write services. Unlike dynamic views that are query-intensive and performance-costing, view stores offer a performant solution for maintaining updated data representations.

- Architectural Integration: Positioned within the context of a command microservice, view stores transcend the micro-management paradigm to act as macro modules. These modules consolidate data relevant to presentation and other microservices’ business logic needs, promoting efficient data accessibility and interaction across services.

- Flexibility and Adaptability: Whether underpinned by event sourcing or traditional databases, the write services in the MCMQ pattern do not necessitate a direct read database counterpart. Data can be presented as-is from the write database or integrated with additional data for comprehensive views. The static view design is primarily driven by user experience (UX) requirements, with the architecture capable of adapting to presentation layer changes by generating new view combinations from the source of truth — the write database.

The MCMQ pattern offers a strategic approach to leveraging the benefits of microservices while addressing the challenges of data aggregation and sharing. By integrating the view store concept, MCMQ facilitates a scalable, flexible framework for data management that supports dynamic presentation needs without compromising service segregation or data integrity.

Understanding the View Store in Data Architecture

A view store is a specialized database designed to aggregate data from multiple collections or sources, serving as a centralized repository for aggregated, precomputed views. It is particularly suited for technologies that prioritize scalability and rapid read access over relational data complexities, making document-based NoSQL platforms like Elasticsearch and MongoDB ideal choices for implementing view stores.

- Technology Basis: The choice of technology for a view store depends on the specific needs for scalability and read performance. Elasticsearch, with its focus on scalable search capabilities, and MongoDB, known for its flexible document model, are both effective platforms for view stores. In Elasticsearch, a static view is represented as an index, while in MongoDB, it takes the form of a collection.

- Static Views and Interactions: Unlike relational databases that rely heavily on SQL joins for data interaction, view stores manage static views without inter-view interactions. This approach aligns well with platforms like Elasticsearch, which excel in handling non-relational data at scale. A view store can encompass multiple static views, each optimized for specific query patterns or presentation needs.

- Architecture and Scaling: Theoretically, a single installation of Elasticsearch with multiple clusters can suffice for large-scale applications due to its inherent scalability. However, distributing static views across separate view stores can also be a viable strategy, primarily influenced by system scaling considerations rather than architectural constraints. The absence of inter-view relations allows for flexible scaling and distribution strategies.

- Indexer Service: The creation and maintenance of static views are managed by a component known as the Indexer. This service, acting as a macro module, is tasked with generating and updating static views based on predefined business logic. Depending on the application’s scale and complexity, an Indexer service may oversee a single view, multiple views, or all views within the application. While centralizing this functionality within one Indexer service might seem contrary to microservice architecture principles, it does not constitute an anti-pattern. Since each view encapsulates isolated business logic and responds to specific state change events, the organizational structure of Indexer services (singular or multiple) can be tailored to suit operational demands and the load characteristics emanating from the BFF (Backend for Frontend) service.

The MCMQ pattern

Optimizing View Store Usage for Efficient Data Presentation

The creation and updating of views within a view store are dictated by presentation-layer business logic, often necessitating the aggregation of complex data derived from multiple events. While the Backend for Frontend (BFF) pattern still plays a pivotal role in dynamically aggregating data from these views for user interfaces, it’s crucial to manage view store utilization judiciously to address storage costs effectively.

- Focused Aggregation for Efficiency: Views should primarily be designed to consolidate data essential for overcoming specific challenges, such as facilitating bulk joins or enabling advanced query capabilities (remote criteria) as previously discussed. This approach ensures that views are optimized for scenarios requiring aggregated data for bulk operations or sophisticated filtering and sorting, rather than indiscriminately merging data.

- Strategic Data Inclusion: While the primary intent of a view is to support operations that inherently benefit from pre-aggregated data, the decision to include additional aggregated data not directly related to bulk requests or query optimizations should be made with careful consideration of storage implications. Incorporating such data into views can enhance performance for individual document requests, offering high-speed access to enriched data sets.

- Balancing Performance and Storage Costs: The decision to enrich views with additional aggregated data, beyond the immediate needs of bulk operations or query enhancements, hinges on a cost-benefit analysis of storage expenses against performance gains. If storage costs are deemed manageable, and the performance improvement for accessing single documents is justified, then enriching views can be a viable strategy. This approach leverages the high performance of static views to deliver a comprehensive data experience, even in scenarios not characterized by bulk interactions or complex query requirements.

Innovating View Schema Design for Static Views

A view schema represents the structural and logical blueprint for static views, extending beyond traditional SQL view definitions to accommodate a wide array of data aggregation strategies. These strategies encompass one-to-one, one-to-many relationships, grouping, summarizing, calculating, simplifying, transposing, and conditional joining among others. Given the complexity and versatility required for these operations, a detailed and clear aggregation code is paramount for effectively realizing a view schema. However, the potential complexity of such schemas invites innovation in their definition and implementation.

- Towards a Schema-based Decorative Language: The intricate nature of static view aggregation presents an opportunity to devise a simple, schema-based decorative language. This language would enable developers to define static views with clarity and precision, specifying aggregation logic in a way that is both comprehensible and executable by view store technologies.

- View Store Technologies with Integrated Schema Processing: Envisioning view store platforms that not only store data akin to Elasticsearch but also dynamically update it in response to schema declarations represents a significant advancement. By subscribing to message broker events, these platforms could automatically adjust stored views according to newly declared view schemas. This capability would dramatically reduce the time and complexity involved in deploying new presentations, allowing for the rapid instantiation of static views based on emerging business needs.

- Redefining the Role of Indexer Services: In a scenario where view store platforms inherently understand and process view schemas, the traditional role of separate indexer services may be reimagined. If the view store itself can act as the indexer — interpreting schema declarations and adjusting views in real-time — the architecture simplifies, consolidating functionality and streamlining data management processes.

This approach not only enhances the flexibility and responsiveness of data presentation layers but also signifies a shift towards more intelligent and autonomous data systems. By enabling quick adaptation to new presentation requirements through declarative schema definitions, view store technology promises to revolutionize how static views are created, updated, and managed, potentially obviating the need for dedicated indexer services in certain contexts.

A new View Store Technology Vision

Conclusion: The Transformative Advantages of the MCMQ Pattern

The Micro Command and Macro Query (MCMQ) pattern represents a significant advancement in microservices architecture, adeptly addressing complex challenges associated with data management and presentation. By innovatively redefining the dynamics between microservices, static views, and data aggregation, MCMQ delivers a holistic framework that bolsters scalability, flexibility, and operational performance. The pattern’s standout advantages, particularly in solving the bulk join and remote criteria challenges, underscore its value:

- Efficient Resolution of Bulk Join Challenges: MCMQ adeptly navigates the complexities of aggregating large datasets from disparate services, a task that traditionally imposes significant performance and complexity burdens. Through the strategic use of static views and view stores, the pattern facilitates efficient data compilation, enabling seamless bulk joins without the overhead associated with traditional relational database operations.

- Elegant Handling of Remote Criteria: The pattern excels in addressing remote criteria challenges, where query operations depend on data managed by different services. MCMQ’s architecture allows for data aggregation based on sophisticated filtering and sorting criteria, bridging the gap between distributed data sources and presenting a unified, coherent dataset to end-users.

- Enhanced Data Aggregation: Beyond solving specific technical challenges, MCMQ streamlines the aggregation of data from multiple sources, offering a unified view essential for complex data presentations. This efficiency in data retrieval and aggregation significantly reduces the real-time processing load, ensuring high-performance data delivery.

- Scalability and Flexibility: Rooted in microservices principles, MCMQ inherently supports scalable and adaptable system designs, allowing services and views to evolve independently. This capacity for dynamic adaptation ensures that systems can grow and adapt without extensive rearchitecting.

- Optimized Performance: By segregating command and query operations and leveraging static views, MCMQ minimizes performance bottlenecks, enhancing system responsiveness and efficiency in processing high-demand queries.

- Simplified Data Management: The introduction of view schemas and the potential automation of view updates through view store technologies streamline data management processes, reducing the complexity and maintenance efforts traditionally required.

- Comprehensive Data Insights: Through event sourcing, MCMQ maintains a detailed historical record of data changes, enhancing data traceability and providing valuable insights for audit and analysis purposes.

Incorporating the MCMQ pattern into microservices architecture offers a robust solution to the enduring challenges of bulk join and remote criteria, ensuring systems are not only more efficient and manageable but also poised for future technological advancements. By embracing MCMQ, organizations can achieve a delicate balance between operational efficiency and technological agility, making their architectures resilient, responsive, and ready for the challenges and opportunities of the digital future.