Data at Scale
Phase 15 is a comprehensive masterclass in architecting data infrastructure at planetary scale.
All Topics in Phase 15
0 of 11 completedExamine database processing paradigms: Row-oriented ACID transactional stores (PostgreSQL, MySQL) versus Columnar analytical execution engines (ClickHouse, Snowflake, BigQuery), storage formats, compression algorithms, and SIMD vectorization.
Explore data transformation architectures: Extract-Transform-Load (legacy ETL) vs Extract-Load-Transform (modern ELT with dbt, Snowflake, and BigQuery), Change Data Capture (CDC), and the Medallion data architecture.
Unify open data storage architecture: The Apache Parquet columnar file format, Apache Iceberg and Delta Lake table formats, transactional metadata layers (Snapshots, Manifests), ACID commits on object storage, and time-travel querying.
Understand large-scale distributed computation: Jeffrey Dean and Sanjay Ghemawat’s MapReduce architecture, the Map phase, Shuffle & Sort network mechanics, the Reduce phase, and the evolution to Apache Spark DAG memory execution.
Architect petabyte-scale distributed file storage: Master/NameNode in-memory metadata management, ChunkServer/DataNode block storage, 128MB large block sizing physics, 3x rack-aware pipelined replication, and the Small Files Problem.
Optimize analytical query pruning: Directory-level partition pruning, multi-dimensional Clustering Keys (Z-Ordering, Hilbert curves), Min/Max metadata skipping, avoiding partition explosion, and query cost reduction in Snowflake and BigQuery.
Architect high-throughput, sub-second analytical streaming pipelines: Apache Kafka event ingestion, ClickHouse MergeTree storage engines, streaming Materialized Views, SIMD vectorization, and Kappa architecture.
Architect large-scale personalized recommendation engines: Collaborative Filtering, Matrix Factorization, Two-Tower Deep Neural Networks, Vector Search (HNSW / IVF-PQ), Real-Time Feature Stores, and Multi-Stage Ranking Pipelines.
Architect large-scale distributed search engines: Web crawling frontiers and politeness policies, Inverted Index data structures (Finite State Transducers, Postings Lists), Okapi BM25 relevance ranking, Lucene immutable segments, and Elasticsearch distributed sharding.
Architect large-scale multi-channel notification engines: Push notifications (Apple APNs / Google FCM), SMS (Twilio), Email (AWS SES / SendGrid), in-app WebSockets, user preference matrices, rate-limiting spam guards, and dead-letter queues.