ClickHouse Indonesia User Group

If anyone from the community is interested in sharing a talk at future events, complete this CFP form and we’ll be in touch.
_____________________________________

🎤 Session Details: Stream, Ingest, Monitor: Scaling OTT Analytics with ClickHouse
As OTT platforms grow, managing telemetry data streams from active viewers becomes a significant engineering challenge. Processing millions of events per minute—ranging from playback initializations and user interactions to performance indicators like buffering—requires a robust pipeline architecture to prevent infrastructure cost inflation or system performance degradation.
In this session, we will discuss how ClickHouse serves as a core component in our OTT analytics data pipeline architecture. The discussion will be broken down into three key pillars aligned with our title: how large-scale telemetry data is streamed, efficiently ingested in high volumes, and processed to monitor performance metrics in near real-time. We will share the end-to-end architecture used to handle millions of events per minute directly from edge devices.
Speaker: Rafif Abdus Salam, Senior Data Engineer @ Vidio
Rafif Abdus Salam is a Data Engineer at Vidio. He is passionate about building scalable data pipelines, driving data governance, and enabling AI from data to solve complex business challenges. His work centers on managing high-velocity streaming data and building efficient architectures for large-scale analytics.

🎤 Session Details: Why Real-Time Makes Batch Data Modeling Harder Than Expected: Lessons from a Personal Data Engineering Project
Organizations are increasingly adopting real-time analytics to enable faster decision-making and more responsive applications. However, many existing data platforms were built for batch processing, where data is collected, transformed on a schedule, and stored as stable historical snapshots. Enabling real-time capabilities therefore requires not only infrastructure changes but also a fundamental shift in data modeling.
Traditional batch models assume datasets are complete, consistent, and immutable after processing. In contrast, real-time systems continuously process streaming events that may arrive late, out of order, or be corrected after ingestion. As a result, analytical outputs must be updated continuously rather than generated as fixed batch results.
This transition introduces several modeling challenges, including handling late-arriving events, maintaining consistency between fact and dimension data over time, and ensuring reproducible analytics despite continuous updates. To address these issues, data models must support event-time processing, incremental updates, upserts, versioned records, and slowly changing dimensions instead of relying solely on static star schemas.
Modern lakehouse technologies such as Delta Lake, Apache Iceberg, Apache Hudi, and Apache Spark help unify batch and streaming workloads through features including ACID transactions, schema evolution, and incremental processing. However, these technologies do not remove the need for careful data model design, as correctness still depends on effectively managing state, time, and consistency.
This work presents the implementation of a real-time lakehouse architecture to support both batch and streaming workloads within a unified platform, illustrating that adopting real-time analytics requires rethinking traditional data modeling approaches rather than simply introducing new technologies.
Speaker: Yunata Gunawan, Data Engineer @ Global IT consulting firm (withheld for privacy)
Data Engineer working in a consulting environment, focused on building reliable data pipelines and analytics infrastructure. Passionate about data architecture, system design, and scalable backend systems, with a goal of becoming a Solution Architect.

🎤 Session Details: Document Analytics at Columnar Speed: The Internals of ClickHouse's JSON Type
JSON has always been the awkward guest in analytical databases: flexible to write, painful to query. This talk goes under the hood of ClickHouse's production-ready JSON data type to show how it stores every JSON path as a true columnar subcolumn. We'll start with the foundations. The JSON type is built on two new general-purpose types: Variant, which efficiently stores values of different data types within the same column without coercing them into a common type, and Dynamic, which handles values whose types aren't known in advance. Then we'll make it concrete with real numbers. Using the JSONBench benchmark. Finally, we'll cover practical guidance: when the JSON type is the right choice versus explicit columns or Tuple types. Plus when you add others indexing capability like fulltext search
Speaker: Andi Pangeran, Head Of Engineering @ amartha.com