HBase Fundamentals, Architecture, and Operations: Data Model, Table Design, Shell, and Java API

The 18 HBase notes across the two repositories are highly homologous (9 in Personal + 9 in Fuwari, essentially one-to-one counterparts differing only in language and length). This page converges them into four parts — data model, architecture, table design, and operations — removing duplicated chapters and keeping reusable concepts.

Core Mechanism

1. Data model: logical vs physical separation

• Logical: Table → RowKey → ColumnFamily:ColumnQualifier → Timestamp → Value, sparse and versioned.
• Physical: split Region vertically by ColumnFamily; regions are distributed across RegionServers by RowKey ranges; HStore corresponds to a family, StoreFile/HFile is persisted, MemStore buffers in memory, WAL is write-ahead.
• The sources Hbase逻辑模型与物理模型详解.md (HBase Logical vs Physical Deep Dive) and HBase-Logical-vs-Physical explain how the same row is physically scattered by family.

2. Architecture components

3. Table design and operations

• RowKey: uniform hashing, avoid monotonic incremental hotspots; keep family count small (1–2), qualifiers can be many.
• Operations: HBase Shell (create/put/get/scan/disable/drop), Java API (Admin/Table/Connection) corresponding to HBase-Shell-Administration-Guide.

Applicability

• Wide, sparse tables, write-heavy, need prefix scans by RowKey and version retention.
• You already operate HDFS/ZK, or use managed HBase in the cloud.

Not Applicable and Risks

• Strong transactions, JOINs, secondary indexes, and full-text search are not HBase strengths.
• Poor RowKey design causes Region hotspots; too many families amplify Flush/Compaction.
• Cluster scripts in personal notes (hadoop-hbase-spark) were only validated locally, not in production.

Minimum Verification

1. Create a table via Shell, write multiple versions, verify timestamps and family splitting via scan.
2. Create/delete tables and batch Put/Get via Java API, confirming connection close and error handling.
3. Adjust RowKey prefixes and observe Region splits and hotspots.

Evidence and Uncertainty

• Source facts: ingest-hbase-foundation contains 18 source notes; duplicates differ only in language.
• Synthesis: This page converges scattered chapters into model—architecture—design.
• Unconfirmed: HBase version (not pinned in sources), ZK dependency evolution, and integration with Phoenix/secondary indexes are unverified.

Related Pages

• spark-bigdata-ecosystem
• database-schema-drift
• mysql-storage-and-deadlock

Sources and related documents

• ingest-hbase-foundation
• Spark and Big Data Ecosystem: Hadoop, HBase, Spark Runtime Modes, and Clusters
• Database Schema Drift: Locating with the history, schema, and query Views
• MySQL Storage Engines and Deadlock Detection