Big Data Fundamentals and Alibaba Cloud Big Data Products
Start with the shared vocabulary: volume, velocity, variety, veracity, and value describe why distributed systems exist, while batch versus streaming, OLTP versus OLAP, and storage-compute coupling versus separation describe how they are built. On Alibaba Cloud the core product map is what questions test. MaxCompute is the managed data warehouse for large-scale batch SQL and MapReduce-style computation. DataWorks is the development, scheduling, and governance console that orchestrates work across engines. E-MapReduce (EMR) provides open-source Hadoop/Spark/Hive clusters. Hologres is the real-time interactive analytics engine compatible with PostgreSQL. AnalyticDB serves high-concurrency OLAP queries. DataHub and Log Service (SLS) ingest streaming and log data, OSS provides object storage, and Flink on Alibaba Cloud handles stateful stream processing. DataWorks Data Integration moves data between these stores.
Traps: candidates confuse Hologres with AnalyticDB (both are OLAP, but Hologres emphasizes real-time serving and PostgreSQL compatibility and federated queries over MaxCompute); they assume MaxCompute is a cluster they must size (it is serverless and capacity is purchased as compute units); and they mix up DataHub (streaming ingestion) with DataWorks Data Integration (batch/scheduled synchronization). Another frequent trap is treating OSS as a database - it is object storage with no SQL engine of its own.
Rule of thumb: choose by access pattern - scheduled bulk SQL on petabytes → MaxCompute; sub-second point queries and serving → Hologres; open-source engine control → EMR; continuous event streams → DataHub plus Flink.
Worked example: a retailer lands clickstream events, needs minute-level dashboards, and nightly petabyte aggregation. Answer: DataHub ingests events, Flink aggregates in-stream, results land in Hologres for dashboards, raw events archive to OSS, and MaxCompute runs the nightly batch joins orchestrated by DataWorks.