Part 8/12:
Despite the high throughput of batch data, Zepto identified the need for sub-second response times for critical business metrics. To address this, they built a near real-time architecture using ClickHouse, an OLAP database optimized for low-latency queries.
Why ClickHouse?
Sub-millisecond queries: Essential for real-time monitoring.
Flexible ingestion engines: Via Kafka, data is rapidly loaded into denormalized tables.
Robust table engines: Such as ReplacingMergeTree for upsert operations, SummingMergeTree for aggregation, and ReplicatedMergeTree for resilience.