December 2025 Summaries

Advent of Code, a popular annual programming challenge, features algorithmic puzzles that are typically tackled using general-purpose languages. However, ClickHouse, a vectorized query engine, showcased its capabilities by solving the 2025 puzzles using only its SQL. By leveraging ClickHouse's extensive library of functions, including Recursive CTEs, array functions, and polygon functions, the team managed to overcome the inherent limitations of SQL in handling complex state management and data structures. The challenge imposed strict rules, such as using pure ClickHouse SQL without preprocessing inputs or using UDFs. Solutions demonstrated the database's ability to handle tasks traditionally thought to be impossible in SQL, transforming them into manageable data problems. Despite not being as computationally optimal as solutions in languages like Rust or Python, the ClickHouse queries delivered accurate results, underscoring the engine's versatility and power in complex problem-solving scenarios.

ClickStack's December update introduces a significant feature enhancement, focusing on the smarter integration of Materialized Views within the HyperDX UI to improve performance and scalability. Materialized Views, now directly creatable in ClickHouse, can be linked to a source and utilized by the Query Layer to enhance the speed and efficiency of charts, searches, and dashboards. This feature allows for automatic selection of the most efficient view for each query, leveraging the EXPLAIN ESTIMATE clause to minimize data scanning and deliver faster responses. Though currently in beta, the feature aims to lower operational costs and optimize performance, even at large scales, by computing only incremental changes to views as new data arrives. This advancement marks a step forward in user configurability and workflow optimization, with plans to introduce more automation based on user behavior in future updates.

As the year concludes, a reflective look at the demos built by the ClickHouse team highlights their role in showcasing the platform's capabilities in real-world applications. These demos, ranging from real-time market analytics with StockHouse to open analytics for RubyGems with ClickGems, and natural language querying with AgentHouse, demonstrate how ClickHouse enables fast, transparent, and scalable data processing. StockHouse exemplifies real-time data ingestion and visualization with minimal configuration, while ClickGems facilitates community-driven SQL exploration of large datasets. AgentHouse illustrates how ClickHouse can support agentic applications by allowing natural language interfaces to query data. Additionally, the ClickStack demo provides insight into observability practices by simulating a microservices-based e-commerce application with integrated telemetry. Throughout these projects, the emphasis is on storing raw data, utilizing materialized views for efficiency, and maintaining transparent query processes, thus allowing developers to adapt and build upon these systems. The openness and real-world data of these demos enable users to experience the rapid query responses and scalability that ClickHouse provides, complementing traditional benchmarks by offering a tangible sense of the platform's performance.

The text presents a curated list of favorite blogs and authors that focus on various aspects of technology, including performance optimizations, algorithms, computer science, hardware, and AI, among other topics. It highlights notable contributors such as Daniel Lemire, known for high-performance libraries and algorithms, and Ash Vardanyan, recognized for his work on performance optimizations and hardware. The list covers a wide array of interests, mentioning blogs dedicated to compilers, data compression, and distributed systems, while also acknowledging that some blogs have ceased updates. The author shares personal anecdotes and experiences with these resources, particularly in relation to the ClickHouse database, indicating the practical applications and contributions of these authors to the field. The text expresses a wish for the continuation of insightful technical writing from these authors despite some having moved on to other interests.

The December 2025 ClickHouse newsletter covers various updates, including a new Postgres extension for querying ClickHouse, cost-performance analyses of major cloud data warehouses, and a year-in-review for Postgres CDC. It highlights community member Zjan Carlo Turla's contributions to Canva's observability infrastructure and outlines upcoming global events, such as virtual training sessions and meetups in different regions. The newsletter also announces new self-paced learning paths, offering free ClickHouse certification until the end of 2025. Notable software updates include ClickHouse's 25.11 release, featuring enhancements like projections as secondary indices and optimized GROUP BY operations. It introduces pg_clickhouse, a PostgreSQL extension for enhanced query performance, and discusses multi-cloud cost analytics with ClickHouse. Additional insights are provided on optimizing ClickHouse for specific data types, setting up a local dbt environment, and using ClickHouse for security applications such as Tempesta FW's L7 DDoS mitigation.

ClickHouse Cloud's launch in Microsoft Azure UAE North marks a pivotal advancement in providing real-time analytics services while adhering to data residency requirements within the United Arab Emirates. This expansion supports the UAE's burgeoning digital landscape, particularly in fintech, e-commerce, and smart city initiatives, by offering a high-performance analytics infrastructure compliant with local data protection regulations. The service enables enterprises to process massive data volumes with reduced latency and ensures compliance with regulations such as the UAE's Federal Decree-Law No. 45 of 2021. Key benefits of ClickHouse Cloud include simplicity in service creation, scalability, security, and seamless integration with the Azure ecosystem, appealing to sectors like financial services, retail, IoT, and gaming. The platform's architecture allows for efficient, real-time analytics, offering a competitive edge in a region where rapid digital transformation is underway.

ClickHouse Inc. has announced the availability of "ClickHouse Cloud" in the Google Cloud Tokyo region, allowing Japanese companies to perform real-time data analysis with low latency and high reliability while keeping their data within the country. This full-managed service, based on the open-source ClickHouse, offers a highly efficient query engine capable of processing billions of rows in milliseconds and is designed to reduce infrastructure management burdens. The launch is part of the company's strategy to enhance its analytics platform offerings tailored to the Japanese market and support next-generation data-driven applications across various industries, including AI, IoT, digital marketing, and financial services. With the introduction of ClickHouse Cloud, businesses can optimize application performance, enhance AI/ML workflows, and maintain data sovereignty, all while integrating seamlessly with Google Cloud services. This service is accessible through the Google Cloud Marketplace, facilitating easy integration with existing Google Cloud environments.

In the process of developing pg_clickhouse, an extension that facilitates querying ClickHouse from PostgreSQL, a significant optimization was achieved by altering how session settings are parsed and managed. Initially, the settings were parsed individually for each query, causing overhead. The solution involved pre-parsing these settings into a key/value data structure at the time of their assignment, which reduced the need for repeated parsing. The implementation journey involved several attempts, each refining the approach to memory allocation and data handling within PostgreSQL's GUC (Grand Unified Configuration) API. The final solution, inspired by PostgreSQL's handling of time zones, uses a single memory block for storage, simplifying memory management and reducing error potential. This process not only enhanced the functionality of pg_clickhouse but also provided valuable insights into the intricacies of C programming and PostgreSQL extension development.

The author reflects on the significance of benchmarks in their work over the past year, highlighting how benchmarks have provided clarity and challenged existing narratives in the field of data analytics, particularly with ClickHouse. Benchmarks like JSONBench and FastFormats have demonstrated ClickHouse's superior performance and storage efficiency compared to traditional systems, leading to practical changes in how data is managed and processed. The benchmarks explored various aspects, such as input formats, object storage caching, and update mechanisms, revealing insights into system efficiency, latency, and cost-effectiveness. This emphasis on performance has shown that an engine's ability to move and process data efficiently is crucial, with benchmarks serving as a tool to validate architectural decisions and drive real-world improvements. The author notes that these benchmarks not only corrected misconceptions but also influenced the development of analytical systems, underscoring the importance of measuring and publishing results to guide future innovations.

In 2025, ClickHouse introduced 119 new functions, with significant additions in each version update throughout the year. Among these are various functions like `mapContainsValueLike`, which checks if a map contains a value matching a specific pattern, and `perimeterCartesian`, which calculates the perimeter of a geometry in a Cartesian coordinate system. The `HMAC` function, also introduced, offers cryptographic message authentication. Functions such as `argAndMax` and `argAndMin` allow for advanced querying capabilities, such as retrieving the maximum or minimum priced property along with its location. Financial computations are enhanced with `financialInternalRateOfReturn`, which calculates annual investment returns. Other notable functions include `stringBytesUniq` for counting unique bytes in a string, `toInterval` for creating interval values from numbers, and `timeSeriesRange`, which generates a range of timestamps. These functions collectively improve the data processing, querying, and analytical capabilities of ClickHouse, providing users with diverse tools to handle complex data tasks efficiently.

In ClickHouse 25.11, a significant optimization was introduced to enhance the performance of aggregations on 8-bit and 16-bit keys by parallelizing the merge phase for FixedHashMap-based aggregations. This innovation, implemented by Jianfei Hu, aimed at speeding up the aggregation process by addressing the discrepancies in query performance due to differences in data type handling, such as UInt16 and UInt64, during group by operations. Hu's exploration involved overcoming challenges related to concurrency, memory management, and the limitations of existing aggregation mechanisms, which initially used a one-dimensional array for hash maps, hindering parallel processing. The solution involved a novel approach to distribute merge work among different threads without race conditions, although it faced issues such as memory corruption and performance degradation for trivial functions due to the overhead of parallel merging. These challenges were addressed by refining memory allocation practices and extracting min/max indices before parallel merging to limit iteration ranges, thereby enhancing the efficiency and reliability of the aggregation process in ClickHouse.

EMQ, a leader in the industrial IoT sector, leverages ClickHouse Cloud to provide real-time analytics for over 1,000 enterprise clients, connecting over 250 million devices worldwide. Starting as an open-source MQTT broker, EMQX has evolved into a scalable platform for IoT, IIoT, and connected vehicle applications, supporting a diverse client base including Geely, HPE, VMware, and Ericsson. The integration of ClickHouse enhances EMQ's ability to deliver high-throughput data ingestion and sub-second query processing, enabling real-time insights and flexible analytics. EMQ's platform, which facilitates seamless connectivity from edge devices to cloud dashboards, supports AI-driven observability tools that allow operators to troubleshoot using natural language. The use of ClickHouse MCP further extends AI capabilities into industrial environments, addressing challenges such as managing diverse edge devices, tracking message latency, and handling multiple databases. This integration not only simplifies data querying, monitoring, and manipulation but also supports the development of intelligent, responsive factory operations by combining messaging, storage, and analytics into a unified, AI-ready stack.

EMQ leverages ClickHouse Cloud to enhance real-time analytics for its MQTT-based industrial IoT platform, serving over 1,000 enterprise customers and connecting more than 250 million devices worldwide. Through a ClickHouse-based data pipeline, EMQ connects edge devices to cloud dashboards, facilitating high-throughput data ingestion and sub-second query capabilities. EMQ's flagship product, EMQX, is a scalable MQTT broker designed for low-latency, high-concurrency environments, enabling efficient data delivery from edge to cloud. The integration with ClickHouse supports real-time insights and flexible data analysis by utilizing AI-powered observability tools to troubleshoot issues and query live data with natural language. This setup allows for seamless data flow from industrial devices to analytics platforms, ultimately empowering operators to run smarter, more responsive operations and paving the way for future intelligent, data-driven factories.

ClickHouse has been awarded the AWS Financial Services Competency, underscoring its commitment to supporting global financial institutions by transforming data into swift, actionable insights. This achievement highlights ClickHouse's focus on meeting the rigorous demands of financial data management, such as real-time market data streaming, fraud detection, regulatory compliance, and large-scale customer analytics. ClickHouse's architecture, featuring columnar storage, parallel execution, and robust time-series operations, promises high performance and efficiency, as demonstrated by Capital One's significant improvements in dashboard load times and cost reductions. Prominent financial entities like Deutsche Bank, RBC, and fintech firms such as Longbridge and Juspay leverage ClickHouse for their mission-critical analytics and operational needs, benefiting from its speed and scalability in handling tasks ranging from market data analysis to fraud prevention. With ongoing innovation in its offerings, ClickHouse aims to empower organizations within the financial sector to leverage their data as a competitive advantage, offering a cloud-based model with flexible pricing options for easy integration and scalability.

At the P99 conference, a ClickHouse developer discussed the company's exploration of integrating Rust into its primarily C++ codebase, highlighting the potential shift from C++ to Rust due to its memory and thread safety benefits. While C++ has been dominant in database management systems, its issues with memory safety, segmentation faults, and race conditions prompted the consideration of Rust as an alternative. The approach taken by ClickHouse was iterative rather than a full rewrite, starting with introducing Rust into their CMake build system via the Corrosion project and testing it with non-critical tasks like implementing a new SQL function. This integration saw successes, such as contributions to the ClickHouse client and the inclusion of a library for Delta Lake, though challenges arose, including reproducibility of builds, cross-compilation issues, and managing dependencies. Despite these hurdles, Rust's modularity and the interest from developers in using it offer a promising avenue for future development, encouraging contributions to ClickHouse in Rust.

In 2025, ClickHouse introduced significant enhancements across multiple areas, marking a year of extensive development with 277 new features, 319 performance optimizations, and 1,051 bug fixes. Notable advancements include the introduction of lightweight updates in version 25.7, which allow for efficient data modifications using patch parts, and the expansion of data lake capabilities with support for various catalog systems like REST, Polaris, and Unity. The platform also saw the progression of text and vector index features, enabling faster and more precise querying capabilities, with vector search achieving general availability in 25.8. Additionally, ClickHouse implemented a query condition cache in version 25.3 to improve query performance by caching conditions in the WHERE clause, and introduced automatic global join reordering in 25.9 to optimize complex join operations. The introduction of lazy materialization in 25.4 further optimized resource usage, significantly speeding up data processing by delaying the reading of column data until necessary. The growing community of contributors was instrumental in these developments, reflecting ClickHouse's increasing popularity and robust community support.

pg_clickhouse v0.1.0 is a newly released PostgreSQL extension designed to facilitate the migration of analytical queries from PostgreSQL to ClickHouse, addressing the challenges faced during such migrations. This extension allows queries to be executed on ClickHouse directly from PostgreSQL without modification, by pushing down query execution to ClickHouse and providing a foundation for ongoing query evolution. The development of pg_clickhouse involved building on existing foreign data wrappers like clickhouse_fdw but added improvements such as universal pushdown for analytics queries, support for advanced aggregations, and compatibility with newer PostgreSQL and ClickHouse versions. The extension also features SEMI-JOIN pushdown and enhanced query performance, particularly for decision support workloads like TPC-H. Future developments aim to extend pushdown coverage, optimize query planning, and support additional PostgreSQL functions and ClickHouse data types. Users are encouraged to install pg_clickhouse from GitHub or PGXN releases and to provide feedback on any pushdown issues encountered.

ClickHouse Cloud introduces a new feature for ingesting data from Amazon S3, allowing files to be processed in any order using S3 Event Notifications via Amazon SQS, thus overcoming the previous requirement of lexicographical order. This advancement enhances the functionality of ClickPipes by providing a more robust and efficient method for handling unordered data, backfills, retries, and late-arriving files. The new mode employs a more dynamic approach by listening for ObjectCreated events in an SQS queue, which facilitates the immediate processing of new files without the need for constant polling. This method ensures exactly-once data ingestion semantics, preventing duplicates even in case of reprocessing. Setting up this system involves configuring an SQS queue, creating an IAM role, and establishing ClickPipe connections to S3, with the potential for further automation through ClickHouse's Terraform provider. This feature is currently exclusive to Amazon S3, with plans to extend support to Google Cloud Storage and Azure Blob Storage, promising enhanced flexibility and performance for data integration pipelines.

The BlueSky social network's API has led to a surge in popularity, resulting in a high-volume stream of JSON events per second. To address the challenges of processing this data, a Medallion architecture is proposed, consisting of three layers: Bronze, Silver, and Gold. The Bronze layer processes raw data from the BlueSky API, filtering out malformed or incorrect timestamps and duplicates. It uses ClickHouse's new JSON type to efficiently store semi-structured data. The Silver layer transforms the data further, performing more filtering, standardizing schemas, and ensuring all duplicates are removed. It introduces deduplication windows to minimize delays. The Gold layer provides a final, query-ready dataset for analytics. Throughout the architecture, materialized views and incremental updates enable efficient querying and data processing. The proposed system tackles common challenges in handling semi-structured data, such as malformed timestamps, inconsistencies in structure, and high duplication rates.

At a recent AI Agents Hackathon hosted at Datadog's New York office, ClickHouse emerged as the sole database sponsor, becoming the preferred choice for data-intensive projects. The event drew around 135 participants, including software engineers, students, and indie hackers, all tasked with building autonomous AI agents using at least three sponsor technologies. ClickHouse-powered projects like VitalSignal, which provides personalized global disease alerts, and RedBot, a chatbot security tool, secured top honors for their innovative use of real-time data processing. Throughout the hackathon, ClickHouse demonstrated its capability to handle diverse data types and facilitate rapid query execution, proving to be a reliable backbone for agentic AI applications. The hackathon underscored the synergy between agentic AI and real-time analytics, highlighting how in-person mentorship and strategic use of sponsor technologies can drive successful project outcomes.

The November edition of "What's New in ClickStack" introduces several enhancements to the open-source observability stack for ClickHouse, including the beta release of Service Maps, which provides a high-level overview of service interactions to identify traffic patterns and failures. The update integrates Service Maps throughout the ClickStack experience to visualize request flows without interrupting investigative processes. It also introduces root span filtering to streamline trace searches by focusing on top-level spans, reducing data scanned and improving query latency. Additionally, new features such as highlighted attributes allow users to configure, view, and interact with specific data points directly within search results and trace views. The integration with incident.io offers seamless alert routing, and the inclusion of Lucene search capabilities within the trace waterfall view enhances data exploration. Furthermore, period comparison for line charts enables easy performance analysis by overlaying current data with previous periods. These innovations aim to improve usability, efficiency, and insight discovery for ClickStack users.

ClickHouse version 25.11 introduces a range of enhancements, including 24 new features, 27 performance optimizations, and 97 bug fixes, with significant contributions from a diverse group of new contributors. The release introduces ACME (Let's Encrypt) integration for automatic TLS certificate provisioning and expands parallel processing capabilities for improved GROUP BY operations on small integer keys, optimizing query performance. Additionally, the update features projections as secondary indices, which enhance query efficiency by reducing storage overhead while maintaining primary index benefits. The release also includes optimizations for DISTINCT queries using projections, and introduces new functions such as argAndMax and argAndMin for retrieving associated maximum and minimum values. New functionalities like fractional LIMIT and OFFSET allow for more granular data retrieval, and EXECUTE AS user functionality facilitates query execution on behalf of another user, enhancing access rights management. These updates collectively aim to improve both the speed and flexibility of data processing and querying within the ClickHouse platform.

Over the past year, ClickHouse has successfully integrated PeerDB into its cloud platform to power the Postgres CDC connector, resulting in significant growth and customer adoption. The primary goal was to enable seamless synchronization between Postgres and ClickHouse, allowing customers to efficiently handle transactional data and analytics. The acquisition of PeerDB in 2024 facilitated this integration, with PeerDB remaining open-source to foster community contributions. Since its launch, the Postgres CDC connector has seen a significant increase in usage, with over 400 companies now utilizing it to replicate vast amounts of data monthly. Two main use cases have emerged: real-time customer-facing analytics and data warehousing, with the connector proving essential in addressing the limitations of Postgres for large-scale analytical workloads. The integration has also led to the development of several key features to enhance reliability, usability, and performance, such as avoiding replication connection reconnections and implementing user-facing alerts. However, challenges remain, including data modeling overhead and the need for further platform maturity, which are being actively addressed to ensure a robust and enterprise-grade experience for users. As data volumes continue to grow, particularly with AI-driven workloads, ClickHouse plans to further optimize its offerings and strengthen the relationship between Postgres and ClickHouse for real-time applications.

clickhouse.build is an open-source, command-line interface tool designed to integrate ClickHouse into existing Postgres-backed TypeScript applications, enhancing analytical workloads by leveraging the strengths of both databases. It functions by scanning the codebase for analytical queries, identifying necessary tables, and automatically syncing these tables to ClickHouse Cloud using the ClickPipes API, while maintaining backward compatibility through a feature flag. Developed in collaboration with AWS Prototyping and Cloud Engineering, the tool utilizes multiple agents, including a scanner, data migrator, and code migrator, to facilitate the transition to a Postgres+ClickHouse architecture. Each agent plays a specific role in identifying queries, migrating relevant data, and updating code, while a QA sub-agent ensures the quality of changes. By using clickhouse.build, users can quickly achieve a proof of concept, allowing them to evaluate their application's performance with this dual-database backend. The project is open-source, inviting contributions and feedback from the community, with aspirations to expand support to other programming languages and databases in the future.

Pump, likened to the "Costco of cloud" by Forbes, aggregates cloud usage from its customers to secure cost-effective deals from providers like AWS, and has developed Pump View, a cost visibility hub, to enhance transparency in cloud billing. This platform addresses the limitations of AWS Cost Explorer by offering detailed insights into expenses by region, usage type, or custom tags, powered by ClickHouse to handle the massive scale of data involved. The architecture, initially considered for Postgres, leverages ClickHouse for its ability to decouple reads from writes and handle bulk data ingestion, dealing with challenges such as semi-sparse records, ingestion volume, atomic updates, and interval queries. Pump uses a hybrid system where ClickHouse manages large-scale data processing, while Postgres ensures transactional consistency. Upcoming enhancements include integrating Azure as a new data provider and migrating more functionalities into ClickHouse to further optimize performance and scalability.