Comparison between databases

less than 1 minute read

Published:

BigTable vs DynamoDB

DynamoDBBigTable
ArchitectureDecentralized
Every node has same set of responsibilites		Centralized
Master handles metadata, tablet server handle reads/writes
DataModelKey-ValueMultidimensional Sorted-Map
PartitioningConsistent Hashing
Each node is assigned to a random position on the ring		Tablets
Each table is broken into a contiguous range of rows called tablets
ReplicationSloppy Quorum
Each data item is replicated to 'N' number of nodes		GFS Chunk replication
Data is stored in GFS. Files in GFS are broken into chunks, and these chunks are replicated to different servers
CAPAPCP
OperationsBy KeyBy Key-Range
StoragePlug-inSSTables in GFS
Memberships and failure detectionGossip based protocolHandshakes initiated by the master

Cassandra is a distributed, decentralized, and highly available NoSQL database. Its architecture is based on Dynamo and BigTable. Cassandra can be described as a BigTable-like datastore running on a Dynamo-like infrastructure. Cassandra is also a wide-column store and utilizes the storage model of BigTable, i.e., SSTables and MemTables.

Leave a Comment

You May Also Enjoy

Algorithms Index

less than 1 minute read

Published:

This is the central index for the C++ algorithms series.

Segment Tree and Fenwick Tree

4 minute read

Published:

Fenwick Tree (Binary Indexed Tree)

Fenwick tree supports point updates and prefix/range sums efficiently. Core idea: each index stores partial sum of a power-of-two range. How it works: move index with idx += idx & -idx for update and reverse for query.

String Algorithms

4 minute read

Published:

KMP Prefix Function

Prefix function (pi) stores longest proper prefix that is also suffix for each prefix. Core idea: reuse previous match length instead of restarting from zero. How it works: on mismatch, jump using pi[j-1]; on match, increase j.

Dynamic Programming

2 minute read

Published:

1D DP: Fibonacci

DP stores answers to subproblems so each state is solved once. Core idea: define state and transition. How it works: dp[i] = dp[i-1] + dp[i-2].