Benchmarking RDBMS vs NoSQL
RDBMS
RDBMS not fast enough although there is materiallized view, join is the problem. The need of read and write data, scalability, dinamika struktur yang berbeda/fleksibelitas struktur(vertikal/horizontal)/bagaimana kita mengextend pengembangan kapasitas.:
Vertikal/scale up: MEngembangkan kemampuan server di fisik yang sama
Horizontal/scale out: Banyak server yang bekerja bersama(ex. per region).
RDBMS traditional is good for vertical scaling, sedangkan NoSQL is good for horizontal scaling. This is fundamental for NoSQL to be chosen by scalability for modern app.
RDBMS Schema ketat, dalam organization startup can be bottleneck for development velocity
NoSQL
Schemaless, no need structure.
- document can be different structure
- add field can be anytime
- there is no downtime
- fast development
BASE(Almost opposite of ACID):
- Basically available, nodes go down but whole system shouldnt be affected
- Soft state(scalable), state ssytem exchange overtime
- Eventual Consistency, given enough time
CAP:
quartz/content/Semsester 4/Data Modelling/cap.png
NoSQL(Collection(Documents(Fields)))
4 Categories:
1. Key Values Store
Store data as key-value
caching, session fastest O(1)
Digunakan ketika memerlukan akses data yang sangat cepat dengan access pattern
Very good for situation that only need one primary key access to get data
2. Column Familiy
banyak baris dimana satu baris mengandung banyak kolom, like sparse matrixx
setiap baris dapat memeiliki banyak kolom berbeda(unlimited columns per row)
time series, logs fast for analytics
3. Documents
CMS, e-commerce fast for queries
schema less ofc, nested objects, arrays support, rich query language
json
mongoDB
Agregat dan Denormalisasi:
- embedding(denormalisasi),
- referencing(rnormalisasi)
4. Graph
social networks, reccomendations fast for traversal
Jangan terpaku pada satu kategori, gunakan polyglot persistence,
postgre transactional
redis caching
elasticsearch search
neo4j recommendation