Back to roadmap
Module 7 · NoSQL, Search, Graph, ObjectDay 06125 min

Wide-Column Stores

Cassandra/HBase: write-optimized, partitioned, big.

← PreviousNext →
Day 061

Wide-Column Stores

Partition
datastore
Cluster cols
datastore
SSTable
datastore
Signal path
Wide-column row organized by partition+cluster
Partition
datastore
flow
Cluster cols
datastore
Cluster cols
datastore
flow
SSTable
datastore
Memory hook

Wide-Column Stores: cassandra/hbase

Mental model

match the datastore to the access pattern

Design lens

Excellent writes; reads need partition awareness.

Recall anchors
Cassandra/HBasePartition + ClusterLSM/Compaction

Why it matters

Wide-column stores group columns into rows, organized by partition (where the row lives) and clustering keys (sort within partition). LSM-tree storage absorbs huge write throughput.

Deep dive

Cassandra: leaderless replication, tunable consistency, AP.

Schema is per-table but rows can have varying columns.

Compaction merges SSTables; tune to avoid surprise spikes.

Demo / scenario

Time-series event ingestion.

  1. Partition by (device_id, day).
  2. Cluster by ts DESC.
  3. Bulk writes hit memtable then flush to SSTable.
  4. Reads scan partition by ts range.

Tradeoffs

  • Excellent writes; reads need partition awareness.
  • Tombstones from deletes accumulate.
  • Tunable consistency adds operational complexity.

Diagram

Partition
Cluster cols
SSTable
Wide-column row organized by partition+cluster.

Mind map

Check yourself

Loading quiz…

Sources & further reading

← PreviousNext →