Scribd
Upload
138 views33 pages

Mysql Cap

The document discusses CAP theorem and different MySQL high availability (HA) solutions in relation to CAP. It summarizes that CAP theorem states that a distributed system can only satisfy two of three properties: consistency, availability, and partition tolerance. It then provides overviews of various MySQL HA solutions including shared storage, DRBD, master-slave replication, multi-master replication, and MySQL Cluster and discusses which CAP properties they satisfy. It also provides technical details on MySQL Cluster architecture including node types, partitions, fragments, node groups, and how it handles splits and failures to maintain availability and partition tolerance.

Uploaded by

mnomikhan
Copyright
© Attribution Non-Commercial (BY-NC)
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
138 views33 pages

Mysql Cap

The document discusses CAP theorem and different MySQL high availability (HA) solutions in relation to CAP. It summarizes that CAP theorem states that a distributed system can only satisfy two of three properties: consistency, availability, and partition tolerance. It then provides overviews of various MySQL HA solutions including shared storage, DRBD, master-slave replication, multi-master replication, and MySQL Cluster and discusses which CAP properties they satisfy. It also provides technical details on MySQL Cluster architecture including node types, partitions, fragments, node groups, and how it handles splits and failures to maintain availability and partition tolerance.

Uploaded by

mnomikhan
Copyright
© Attribution Non-Commercial (BY-NC)
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 33

谭俊青@MySQL实验室

(http://www.mysqlab.net)
上海爱可生信息技术有限公司
2010-04-03
CAP Theorem and NoSQL

Consistency
Availability
Partition Tolerance
CAP Theorem (2/3)
CAP Theorem (3/3)
MySQL HA Solution
 MySQL + Shared-Storage
 MySQL + DRBD (CP)
 Master + Slave (AP)
 Master + Slave(SemiSyncReplication) (CP/AP)
 Multi-Master (AP)
 MySQL Cluster (CAP? CP/AP)
MySQL + Shared-Storage
 Active/Passive
Web/App
Server

Active Server Cluster Management Passive Server


= Virtual IP =
10.10.10.10

Cluster Agent Cluster Agent

= Private IP = = Private IP =
10.10.10.20 10.10.10.21

SAN
MySQL + DRBD
Web/App
Server

Active Server Linux Heartbeat Passive Server


= Virtual IP =
10.10.10.10

= Private IP = = Private IP =
10.10.10.20 10.10.10.21

Primary DRBD Secondary DRBD


DRBD
Master + Slave
 Multi-Master
Web/App
 SemiSyncReplication Server

Master Server Linux Heartbeat Slave Server


= Virtual IP =
10.10.10.10

= Private IP = = Private IP =
10.10.10.20 10.10.10.21
Replication
MySQL Cluster
MySQL Server or NDB API
Web/App Web/App for all Writes & Reads
Server Server

MySQL MySQL
Server Server

NDB API
Management
Memory NDB Server
& Data Storage Engine Data
Disk Node Node
Management
Server
MySQL General Architecture (1/3)
MySQL General Architecture (2/3)
MySQL General Architecture (3/3)
What is a Cluster
 Shared-nothing vs Shared-disk
- In a shared-nothing , which MySQL Cluster is, each node
has its own complete set of hardware
- In a shared-disk architecture, there is a central storage
location that all of the nodes will access and make use of
 MySQL Cluster Hardware
- MySQL Cluster does not require and special hardware,
such as SAN or NAS
- Each node can run on commodity type hardware
- Designed to allow many maintenance operations to be
completed in a online fashion
MySQL Cluster Architecture
Node Types
 Data Nodes
-- Storage Nodes
 API Nodes
-- Mediators between the end process
and the data nodes
 Management Node
-- Manages the configuration and control
of the MySQL Cluster
Partition
MySQL Cluster Architecture
Fragments
Node Groups
Split Brain (1/3)
Split Brain (2/3)
Split Brain (2/3)
Arbitration

First Node to ask will continue while the other will be


shut down
More Data Nodes
Asynchronous (AP of CAP)
Synchronous (CP of CAP)
Phase One: Commit-Request
Phase Two: Successful Commit
Phase Two: Failure(Abort, Rollback)
MySQL Cluster Benchmarks (1/2)
 For 4 Node Cluster, MySQL Cluster 7 achieved 251,000
Transactions per minute which is more than 4X
improvement over the MySQL Cluster 6.3 release.
 For 2 Node Cluster, MySQL Cluster 7 achieved 143,000
Transactions per minute which is more than 4X
improvement over the MySQL Cluster 6.3 release.
MySQL Cluster Benchmarks (2/2)
Questions / Discussion

You might also like