Skip to content

DE Mastery GitHub

Start

Curriculum
Roadmap
How to study with this course

Tracks

FND Foundations of distributed data

Overview
01 Why distributed?
02 Partitioning & shuffles
03 Replication & consensus
04 Time
05 Delivery semantics
06 Storage formats
07 Back-of-envelope estimation

SPK Apache Spark

Overview
01 Spark architecture
02 Lazy evaluation and the DAG
03 The DataFrame API
04 Spark SQL and data sources
05 Catalyst and the physical plan
06 Joins and aggregations at scale
07 Partitioning, bucketing, and the small-file problem
08 Memory, caching, and Adaptive Query Execution
09 UDFs and the Pandas API
10 Structured Streaming I — the micro-batch engine and exactly-once
11 Structured Streaming II — watermarks, state, and streaming SCD2
12 Testing and production Spark
BYO Build your own mini columnar query engine

LKH Lakehouse & open table formats

Overview
01 The Iceberg metadata tree
02 Commits & concurrency
03 Catalogs, branching & write-audit-publish
04 Row-level changes — SCD2, CoW & MoR
05 Table layout & maintenance
06 Delta Lake
07 Hudi & the format landscape
08 Engines on the lake
BYO Build your own mini table format

JVM Java for data engineers

Overview
01 Java for Python developers
02 Collections and streams
03 Build and dependencies
04 Serialization
05 Concurrency and the JVM
06 Testing and debugging

KFK Apache Kafka

Overview
01 The log — events, state, retention and replay
02 Topics, partitions, replication — leaders, ISR, acks, durability
03 Producers — batching, idempotence, ordering under retries
04 Consumers & groups — manual commits, at-least-once, no loss across crashes
05 Transactions & exactly-once — read-process-write, read_committed, fencing
06 Schema Registry & contracts
07 Kafka Connect & CDC
08 Kafka Streams
09 Event & topic design
10 Testing & observability (apps)
11 Cluster configuration & partition reassignment
12 Security — SASL/SCRAM authentication & ACL authorization
13 Deployment architecture — multi-region DR
14 Operating & troubleshooting a cluster
BYO Build your own mini Kafka

FLK Apache Flink

Overview
01 Philosophy & architecture
02 DataStream API
03 Time & windows
04 Flink SQL & dynamic tables
05 Joins — regular, interval, temporal (point-in-time) & lookup
06 State, checkpoints & exactly-once — surviving a TaskManager crash
07 Process functions — timers, keyed state & broadcast rules
08 Table API & PyFlink — one pipeline, two languages
09 Streaming patterns — dedup, top-N, split/merge, MATCH_RECOGNIZE, OVER
10 CDC & lake sinks — Debezium changelog to Iceberg, commit on checkpoint, schema evolution
11 Performance & state at scale — backpressure, parallelism, redistribution, skew
12 Deploy & operate — upgrade a stateful job via savepoint without losing state
BYO Build your own mini stream processor

STH Streamhouse

Overview
01 Architecture history — Lambda → Kappa → lakehouse → streamhouse
02 Apache Paimon — primary-key tables, merge engines, partial-update
03 Apache Fluss — PK tables, lake tiering, streaming vs batch consistency
04 Iceberg as streaming storage — commit interval vs small files
05 Streaming dimensional modeling — SCD2 dim + fct with temporal (as-of) join
06 Streaming data quality — in-flight DQ, quarantine / DLQ, completeness
07 Serving layer — OLAP on the streamhouse
08 Latency tiers & cost — seconds, minutes, hours

Foundations of distributed data · FND-02

Partitioning & shuffles

3 h
requires FND-01

From your world. Redshift DISTKEY/DISTSTYLE choices; why a bad distribution key makes every join a network shuffle

Lesson Exercises 1 Quiz Interview drills Explain it back

Practice quiz with instant feedback. For a scored run that counts toward your local progress, use uv run de quiz FND-02.

1Two tables are both hash-partitioned on customer_id across the same number of partitions. A join on customer_id will: ARun locally per partition with no shuffle (collocated join) BAlways require a full shuffle of both tables CRequire broadcasting the larger table DFail unless the keys are sorted

Equal keys hash to the same partition on both tables. So each partition joins its local slice, with no network redistribution. This is Redshift's collocated DISTKEY join.

Source: https://docs.aws.amazon.com/redshift/latest/dg/c_choosing_dist_sort.html

2You hash-partition orders on company_id, but one of the three companies has 90% of the rows. What happens? ASevere skew: one partition holds ~90% of the rows and its task is a straggler BThe data is perfectly balanced because hashing is uniform CThe join becomes a broadcast join automatically DNothing — hashing fixes skew

Hashing spreads distinct keys, not row volume. A skewed key with low cardinality puts most rows in one partition. Use a high-cardinality key, or salt the hot value.

Source: https://spark.apache.org/docs/latest/sql-performance-tuning.html

3What is the best practical signal that a partitioning is skewed? AA high max/mean ratio of per-partition row counts (or bytes) BThe query plan contains the word 'shuffle' CThe number of partitions is a prime number DThe output is unsorted

Skew shows up as an uneven distribution of partition sizes. If max/mean >> 1, one task does far more work and becomes a straggler.

Source: https://spark.apache.org/docs/latest/sql-performance-tuning.html

4Why is a shuffle expensive? AIt serializes records to disk and moves them across the network between every mapper and reducer BIt recompiles the query plan CIt always sorts the entire dataset twice DIt requires a broadcast of all data to all nodes

The map side spills partitioned records. The reduce side pulls its slice from every source. Serialization, disk, and network all happen. These are the three slowest operations.

Source: https://research.google/pubs/pub62/

5Which are valid fixes for a single hot key causing skew? (select all) (select all that apply) ASalt the hot key (append a random suffix), aggregate, then re-aggregate BChoose a higher-cardinality partition key if the logic allows CBroadcast the small side if the join's other table is small DIncrease the modulus P so that hash(key) % P changes for the hot key

Salting, a better key, and broadcasting all help. A larger P does NOT help a single hot key. hash(k) % P is still one value, so all its rows still go to the same partition.

Source: https://spark.apache.org/docs/latest/sql-performance-tuning.html

6The best way for range partitioning to pick boundaries is: ASampling the actual data to choose balanced split points BAlways using evenly spaced numeric boundaries CHashing the key first DUsing the alphabetical order of column names

Range partitioners sample the data, so each range gets a similar number of rows. Evenly spaced boundaries cause skew when the data is not uniform.

Source: https://spark.apache.org/docs/latest/sql-performance-tuning.html

PreviousFND-01 · Why distributed? NextFND-03 · Replication & consensus

DE Mastery · source on GitHub · Found an error? Report it on GitHub