All 35 articles, sorted alphabetically
Cloudera Operational DB (COD)
Managed HBase on Cloudera CDP.
Read article →HBase Alerting Best Practices
Latency, compaction backlog, RIT.
Read article →HBase Backup + Restore Full/Incremental
Snapshot + WAL replay.
Read article →HBase Balancer
Move regions for even distribution.
Read article →HBase Block Cache
How the block cache accelerates HFile reads, why off-heap BucketCache scales beyond JVM heap, and how to size cache versus memstore.
Read article →HBase Bloom Filters
How bloom filters let HBase skip HFiles during point lookups, the ROW vs ROWCOL trade-off, and how to size filters for the false positive rate you wan…
Read article →HBase BucketCache -- caching without GC pain
Deep-dive on HBase BucketCache: the on-heap-cache GC-pressure problem, the block cache, LruBlockCache vs BucketCache, off-heap/SSD storage, fixed-size…
Read article →HBase Bulk Load
How bulk load bypasses WAL and memstore by writing HFiles directly, why it is orders of magnitude faster for large ingests, and when to use it vs clie…
Read article →HBase Compaction
The compaction process that merges HFiles, why minor and major compactions differ, and how to schedule them to avoid impacting production traffic.
Read article →HBase Coprocessor Architecture in Depth
A 2500-word walkthrough of HBase coprocessors: Observer + Endpoint, access control, secondary index, aggregation, loading, isolation, and safety.
Read article →HBase Coprocessors
How coprocessors let you run user code inside the RegionServer for triggers, secondary indexes, and custom RPCs. Powerful and dangerous.
Read article →HBase HFile Format
The internal structure of HFile: data blocks, indices, bloom filters, and trailer. How the format enables fast point lookups and range scans over sort…
Read article →HBase HMaster
How the HMaster coordinates region assignment, executes DDL, runs the load balancer, and manages RegionServer lifecycle. Not on the read/write hot pat…
Read article →HBase hotspotting
Deep-dive on HBase hotspotting: range-based region distribution, monotonic-key hotspots, symptoms of uneven load, mitigations (salting, hashing, field…
Read article →HBase MemStore and Flushes
How the memstore accumulates writes in-memory, what triggers a flush to an HFile on HDFS, and how flushes interact with compaction and the block cache…
Read article →HBase write path architecture
Deep-dive on HBase's write path: WAL append and sync durability semantics, MemStore skiplists and MVCC visibility, flush trig…
Read article →HBase hbase:meta architecture
How a row key finds its RegionServer: the hbase:meta catalog table, the ZooKeeper bootstrap pointer, the death of -ROOT-, aggressive client-side caching, and meta hotspotting…
Read article →HBase MOB -- storing medium objects efficiently
Deep-dive on HBase MOB (Medium OBject storage): the medium-object write-amplification problem, separate MOB files, decoupled compaction, cell referenc…
Read article →HBase Region Normalizer architecture
Deep-dive on the HBase Region Normalizer: why region shape sets the load ceiling, how the Master chore computes split and merge plans against the tabl…
Read article →HBase Overview
The HBase data model (row key, column families, cells, versions), how it complements HDFS's sequential nature with random access,…
Read article →Apache Phoenix
How Phoenix provides a full SQL interface over HBase, using coprocessors for server-side aggregation, joins, and secondary indexes.
Read article →HBase Procedure v2 architecture
Deep-dive on HBase Procedure v2 and AssignmentManager v2: the ProcedureExecutor, procedure store and WALs, parent-child trees, locks, TransitRegionSta…
Read article →HBase quota throttling architecture
Deep-dive on HBase quota throttling: throttle quotas (req/s, bytes/s, size caps) scoped to user, table, and namespace; token-bucket enforcement at the…
Read article →HBase Region
Contiguous row range. Unit of distribution.
Read article →HBase region replicas architecture
Deep-dive on HBase region replicas: read-only secondary copies of a region that serve stale, timeline-consistent reads for high availability and low p…
Read article →HBase RegionServer
How the RegionServer serves reads and writes, why the WAL is the durability foundation, and how memstore, block cache, and HFiles interact per region.
Read article →HBase Regions and Splits
How HBase automatically splits regions when they grow past a size threshold, how pre-splitting avoids initial hotspots, and when to disable automatic …
Read article →HBase Replication
How HBase replicates data between clusters by shipping WAL edits asynchronously. Master-slave, master-master, and cyclic setups; lag management.
Read article →HBase RegionServer Groups architecture
Deep-dive on HBase RSGroups: the hbase:rsgroup metadata table and ZooKeeper mirror, RSGroupAdminEndpoint, the group-aware stochastic balancer, group-s…
Read article →HBase Snapshots
How HBase snapshots create instant point-in-time views of a table via HFile references, and how to use them for backup, cloning, and restore.
Read article →HBase StochasticLoadBalancer architecture
Deep-dive on the HBase StochasticLoadBalancer: cluster state snapshots, the weighted cost ensemble, candidate generators, the hill-climbing loop under…
Read article →HBase TTL and MAX_VERSIONS
How TTL expires cells automatically, how MAX_VERSIONS bounds historical retention, and how both interact with compaction to reclaim storage.
Read article →HBase Write-Ahead Log
How HBase uses the write-ahead log to make writes durable before acknowledgment, how sync policies balance latency versus durability, and how WAL repl…
Read article →HBase WAL Durability Levels: Architecture Deep-Dive
How HBase's WAL durability levels — SKIP, ASYNC, SYNC, and FSYNC — trade write speed for crash safety per table or per write,…
Read article →HBase WAL splitting architecture
Deep-dive on HBase WAL splitting: why one shared WAL per RegionServer is fast to write but must be demultiplexed per region to recover, ZooKeeper fail…
Read article →