This document summarizes a presentation on using non-relational databases (NOSQL) for data storage in the cloud. It discusses typical motivations for cloud solutions like low costs and scalability. It then covers issues with using relational databases to scale horizontally and vertically. Various NOSQL database models are introduced, including key-value, document, columnar, and graph databases. Tradeoffs between consistency, availability, and partition tolerance are discussed in relation to Brewer's CAP theorem. Code examples are provided for different NOSQL databases.
Axa Assurance Maroc - Insurer Innovation Award 2024
NOSQL in the Cloud
1. 19. und 20. Oktober 2010 in Nürnbergwww.prioconference.de
NOSQLDatenspeicherung in der Cloud
Sergey Shishkin
sergey.shishkin@mt-ag.com
http://shishkin.org
@sshishkin
2.
3. Setting Expectations
• Motivators for cloud solutions
• Distributed database issues
• NOSQL solutions
• Code examples
• It’s not a deep dive session
7. Commodity Infrastructure
• Infrastructure-, Platform as a Service
• Raise the level of abstraction
– Computational power
instead of CPU, GHz and L2 Cache
– Persistent storage
instead of SATA, RAID and SCSI
– Communication channels
instead of Ethernet, DHCP and TCP
• Everything will be in the cloud
8.
9. Who can benefit from Cloud?
• Web 2.0 start-ups
• Social media
17. Data persistence is a bottleneck
• ACID Transactions
• Locks are expensive
• RDBMS power of choice
– Optimized for massive writes
– Or for ad-hoc structured queries
– What a surprise!
• How does RDBMS scale?
21. Horizontal Scale
• Master-Slave Replication
– Still bad for massive writes
• Master-Master Replication
– Resolving conflicts is hard in a normalized model
• Sharding/Partitioning
– Give up unique constraints, foreign keys, joins
22.
23. What was the point of all that again?
Pretty much nothing left over of RDBMS.
24. Consistency
All clients always
have the same
view of the data
Availability
Each client can
always read and
write
Partition
Tolerance
The system works
well despite
physical network
partitions
Brewer’s CAP-
Theorem
32. Data Storage Ecosystem
• NOSQL = Not Only SQL
• RDBMS still has its strengths
– Unique constraints
– But they don’t scale horizontally
• You don’t need a hammer when you have the
right tools ;)