ClickHouse Certified study plan

Explain how columnar storage and vectorized execution differ from row-store OLTP engines, and state which query shapes benefit

Identify MergeTree-family parts, granules, and primary index marks and describe how a scan prunes data

Distinguish the roles of primary key, sorting key, partition key, and sampling key in a MergeTree table

Read an EXPLAIN output for a simple aggregation and point out which stage reads the most rows

Create MergeTree, ReplacingMergeTree, and SummingMergeTree tables and load sample rows into each

Demonstrate how ReplacingMergeTree deduplicates only after merges and explain the FINAL clause trade-off

Choose a partition key and ORDER BY key for a given event table and justify the choice in writing

Describe how TTL rules and codecs on columns affect storage and read cost

Compare MergeTree, AggregatingMergeTree, and CollapsingMergeTree for a changelog-style dataset

Implement an AggregatingMergeTree rollup using -State and -Merge aggregate functions

Select a dictionary or a join engine for a small dimension table and explain the memory and consistency trade-offs

Write a schema for a star-style model that avoids high-cardinality partition keys

Write queries using arrayJoin, higher-order array functions, and JSON extraction on semi-structured columns

Use window functions and WITH expressions to compute session-level metrics

Apply PREWHERE, LIMIT BY, and SETTINGS to control scan and memory behavior

Write a query that uses a materialized view target instead of the raw table and verify identical results

Load data via clickhouse-client INSERT, CSV/JSONEachRow formats, and the HTTP interface

Configure a Kafka engine table with a materialized view into a MergeTree target and explain at-least-once semantics

Describe how ClickHouse Keeper or ZooKeeper supports replication and what happens during a replica catch-up

Set up a scheduled ingestion job using clickhouse-local or a small script

Profile a slow query with system.query_log and identify the top row-reading stages

Tune a query by changing ORDER BY key order and measure the effect on rows read

Explain when to use projections, skip indexes, and materialized views for the same access pattern

Diagnose excessive memory use in a GROUP BY and apply a fix such as spilling or pre-aggregation

Configure users, roles, and row policies and verify access with a restricted account

Describe backup and restore options and outline a recovery procedure for a single table

Explain how ClickHouse Cloud separates compute and storage and what that means for scaling and cost

Identify which system tables expose merge, replication, and disk health signals

Complete a timed practice set weighted by domain percentage and score each domain separately

Re-teach the two weakest domains from memory without notes and correct gaps

Rebuild the full pipeline from weeks 3 and 5 in a fresh environment to confirm retention

Review missed questions and write the underlying rule for each mistake