Limit query results to 50 most recent posts

Use a single global table for all users

Use user_id as partition key and post_timestamp as clustering key

Perform a full table scan for each user’s feed

Use sensor ID and date as the partition key

Store each reading as a separate table

Normalize data into multiple tables

Store readings as wide rows with time as clustering column

Store all telemetry in one large document

Randomly shard by timestamp

Use device_id and timestamp as keys

Use aggregation pipelines with $match and $group stages

Denormalize only after reading

Rely on client-side aggregation

Create compound indexes on (category, month)

Indexes improve read performance at the cost of slower writes 

I.     Indexing reduces query speed

Indexing is not supported on nested fields 

Indexes can be created on nested fields using dot notation

Replication

Sharding 

Indexing 

Normalization

Storing one document in multiple indexes

Storing different documents in different collections

Dividing a document into multiple attributes

Denormalization is common for performance

Joins between column families are frequently used

Wide rows are often used for time-series data

Tunable consistency

ACID transactions across multiple column families

Replication across nodes

Partitioning based on keys

HBase

MongoDb

Apache Cassandra

ScyllaDB