MySQL Cluster Whats New

1. MySQL Cluster
What's New & Best Prac0ces
Copyright 2016, oracle and/or its affiliates. All rights reserved

2. Safe Harbor Statement
The following is intended to outline our general product direcGon. It is intended for
informaGon purposes only, and may not be incorporated into any contract. It is not a
commitment to deliver any material, code, or funcGonality, and should not be relied upon
in making purchasing decisions. The development, release, and Gming of any features or
funcGonality described for Oracle’s products remains at the sole discreGon of Oracle.
Copyright 2016, oracle and/or its affiliates. All rights reserved 2

3. MySQL Cluster Overview
• Memory opGmized tables with durability
• Predictable Low-Latency, Bounded Access Time REAL-TIME
• Auto-Sharding, MulG-Master
• ACID Compliant, OLTP + Real-Time AnalyGcs
HIGH SCALE, READS +
WRITES
• Shared nothing, no Single Point of Failure
• Self Healing + On-Line OperaGons
99.999% AVAILABILITY
• Key/Value + Complex, RelaGonal Queries
• SQL + Memcached + JavaScript + Java + HTTP/REST & C++
SQL + NoSQL
• Open Source + Commercial EdiGons
• Commodity hardware + Management, Monitoring Tools
LOW TCO
Copyright 2016, oracle and/or its affiliates. All rights reserved 3

4. MySQL
Replica0on
MySQL
Fabric
DRBD
Solaris
Cluster,
Clusterware
or Oracle
VM
MySQL
Cluster
MySQL HA SoluGons
9 9 . 9 9 9 %
Copyright 2016, oracle and/or its affiliates. All rights reserved 4

5. Oracle MySQL HA & Scaling SoluGons
Copyright 2016, oracle and/or its affiliates. All rights reserved 5
MySQL Replica0on MySQL Fabric Oracle VM
Oracle
Clusterware
Solaris Cluster MySQL Cluster
App Auto-Failover MySQL Router ✔ ✔ ✔ ✔ ✔
Data Layer Auto-Failover MySQL Router ✔ ✔ ✔ ✔ ✔
Zero Data Loss MySQL 5.7 MySQL 5.7 ✔ ✔ ✔ ✔
Plaaorm Support All All All All Solaris All
Clustering Mode Master + Slaves Master + Slaves Ac0ve/Passive Ac0ve/Passive Ac0ve/Passive Mul0-Master
Failover Time Secs Secs Secs + Secs + Secs + < 1 Sec
Scale-out Reads ✔ ✖ ✖ ✖ ✔
Cross-shard operaGons N/A ✖ N/A N/A N/A ✔
Transparent rouGng MySQL Router For HA ✔ ✔ ✔ ✔
Shared Nothing ✔ ✔ ✖ ✖ ✖ ✔
Storage Engine InnoDB InnoDB InnoDB InnoDB InnoDB NDB
Single Vendor Support ✔ ✔ ✔ ✔ ✔ ✔

6. Who’s Using MySQL Cluster?
Copyright 2016, oracle and/or its affiliates. All rights reserved 6

7. Company Overview
PayPal allows any business or individual with an email
address to securely, conveniently, and cost-effecGvely
send and receive payments online.
Applica0on
PayPal built a cloud-based globally-distributed database
with 100 TB of user-related data based on MySQL Cluster.
A “must NOT lose data” system, delivering 99,999%
availability, transacGonal, with data available WW
anywhere in < 1 Sec.
Detects Fraud within 0.3 seconds.
Deployed over 5 data centers world-wide.
Why MySQL Cluster?
“You can achieve high performance and availability
without giving up rela8onal models and read
consistency.” -- Daniel AusGn, Chief Architect, PayPal
PayPal
Copyright 2016, oracle and/or its affiliates. All rights reserved 7

8. Company Overview
World's largest producer of casual games. Big Fish has
distributed more than 2.5 billion games to customers in 150
countries.
Applica0on
Serve real-Gme recommendaGons to Big Fish customers,
requiring high velocity data ingesGon, low latency access,
online scalability, 99.999% availability, and Enterprise SLAs.
Why MySQL Cluster CGE?
"The MMS plaAorm is a strategic project within Big Fish. We
couldn’t afford to take any chances and MySQL Cluster
provided us with a proven and trusted solu8on to meet the
demands of both our business and our users.“ -- Sean
Chighizola, Sr. Director of Database AdministraGon, Big Fish
Big Fish
Copyright 2016, oracle and/or its affiliates. All rights reserved 8

9. Company Overview
Global telecommunicaGons equipment company, focuses on
fixed, mobile, and converged networking hardware, IP
technologies, sopware, and services.
Applica0on
MySQL Cluster CGE is at the heart of Alcatel-Lucent’s
Subscriber Data Manager used to deliver converged IMS
services to mobile and fixed line users.
Why MySQL Cluster CGE?
• Delivers the performance, scalability, and availability
required by this mission criGcal real-Gme applicaGon.
• Real-Gme read & write access for tens of millions of
subscribers within a single system.
• Always-on service: no offline maintenance window for
services users depend on (voice, SMS, email, Web, social
media, …)
• Rapid delivery & low TCO
Alcatel-Lucent
Copyright 2016, oracle and/or its affiliates. All rights reserved 9

10. SQL
• Industry standard
• Joins & complex queries
• RelaGonal model
Memcached
• Simple to use API
• Key/value based
• Drivers for many languages
Mod-ndb
• RESTful
• JSON over HTTP
• Plugin for Apache
ClusterJ
• Simple to use Java API
• Web & Telco
• Object RelaGonal Mapping
• NaGve & fast access to data
ClusterJPA
• OpenJPA plugin
• Standards defined ORM
• Cross table Joins
JavaScript/Node.js
• NaGve JavaScript: client to
DB
• Blazing fast asynchronous
throughput
Choosing the Right Client API
Copyright 2016, oracle and/or its affiliates. All rights reserved 10

11. MySQL Cluster Auto-Installer
• Fast configuraGon
• Auto-discovery
• Workload opGmized
• Repeatable best pracGces
Specify
Workload
Auto-
Discover
Define
Topology Deploy
Copyright 2016, oracle and/or its affiliates. All rights reserved 11

12. Copyright 2016, oracle and/or its affiliates. All rights reserved 12

13. NoSQL Access to MySQL Cluster Data
Apps Apps Apps Apps Apps Apps Apps Apps Apps Apps Apps Apps
JPA
Cluster JPA
PHP Perl Python Ruby JDBC Cluster J JS Apache Memcached
MySQL JNI Node.JS mod_ndb ndb_eng
NDB API (C++)
MySQL Cluster Data Nodes
13 Copyright 2016, oracle and/or its affiliates. All rights reserved

14. MySQL Cluster 7.4 GA
Copyright 2016, oracle and/or its affiliates. All rights reserved 14

15. MySQL Cluster 7.4 GA
• 200 Million NoSQL
Reads/Sec
• 2.5M SQL Ops/Sec
• 50% Faster Reads
• 40% Faster Mixed R/
W
Performance
• AcGve-AcGve
Geographic
Redundancy
• MulG-DC Clusters
• Conflict DetecGon/
ResoluGon
AcGve-AcGve
• 5x Faster
Maintenance Ops
• Detailed ReporGng
Management
Copyright 2016, oracle and/or its affiliates. All rights reserved 15

16. • Memory opGmized tables
– Durable
– Mix with disk-based tables
• Massively concurrent OLTP
• Distributed joins for analyGcs
• Parallel index scans
• Parallel table scans (for non-
indexed searches)
16
MySQL Cluster 7.4 NoSQL Performance
200 Million NoSQL Reads/Second
Copyright 2016, oracle and/or its affiliates. All rights reserved
!"!!!!
!50,000,000!!
!100,000,000!!
!150,000,000!!
!200,000,000!!
!250,000,000!!
2! 4! 6! 8! 10! 12! 14! 16! 18! 20! 22! 24! 26! 28! 30! 32!
Reads&per&second&
Data&Nodes&
FlexAsync&Reads&

17. • Memory opGmized tables
– Durable
– Mix with disk-based tables
• Massively concurrent OLTP
• Distributed joins for analyGcs
• Parallel index scans
• Parallel table scans (for non-
indexed searches)
17
MySQL Cluster 7.4 SQL Performance
2.5M SQL Statements/Second
Copyright 2016, oracle and/or its affiliates. All rights reserved
!"!!!!
!500,000!!
!1,000,000!!
!1,500,000!!
!2,000,000!!
!2,500,000!!
!3,000,000!!
2! 4! 6! 8! 10! 12! 14! 16!
SQL$Statements/sec$
Data$Nodes$
DBT2$SQL$Statements$per$Second$

18. Performance Enhancements
MySQL Cluster 7.4
50% Read-Only Increase
Copyright 2016, oracle and/or its affiliates. All rights reserved 18
0"
1000"
2000"
3000"
4000"
5000"
6000"
7000"
8000"
9000"
64" 128" 192" 256" 320" 384" 448" 512"
Transac'ons)per)second)
Threads)
Sysbench)R/W)
7.4"
7.3"
7.2"
0"
1000"
2000"
3000"
4000"
5000"
6000"
7000"
8000"
9000"
10000"
64" 128" 192" 256" 320" 384" 448" 512"
Transac'on)per)second)
Threads)
Sysbench)RO)
7.4"
7.3"
7.2"
40% Read/Write Increase

19. • Asynchronous replicaGon between
MySQL Clusters
• AcGve-AcGve MulG-Homed
– Update anywhere
– Conflict detecGon
• ApplicaGon noGfied through excepGon tables
• Can opt to have conflicts resolved
automaGcally
– AutomaGc conflict resoluGon
• ConflicGng transacGon and dependent ones
are rolled-back
• No changes to applicaGon / schema
AcGve-AcGve Geo-ReplicaGon
Copyright 2016, oracle and/or its affiliates. All rights reserved 19

20. What Is a Conflict?
John.balance == $100
John.balance -= $40
John.balance == $60
John.balance == $200
John.balance == $100
John.balance += $100
John.balance == $200
John.balance == $60
Spend $40 Add $100
$60
$200
Copyright 2016, oracle and/or its affiliates. All rights reserved 20

21. • NDB$EPOCH2 and
NDB$EPOCH2_TRANS introduced
• Detects conflicGng inserts/updates/
deletes/reads
• EnGre transacGons (and dependent
ones) rolled back
• Rolling back of transacGons that
read conflicted data
• Improved NDB ExcepGons table
format
– Non-PK columns, operaGon type,
transacGon id, before and aper values
• Online conflict role change
21
Handling of Conflicts – Extensions in MySQL Cluster 7.4
Copyright 2016, oracle and/or its affiliates. All rights reserved

22. • Primary stores logical Gmestamp (GCI)
against updated row(s)
– Window for conflict opens
• GCI is replicated with updated row(s) to
Secondary
• The same GCI is replicated back
(reflected) from Secondary to Primary
aper it has been applied
– Closing window for conflict
• Primary checks every event originaGng
from the Secondary to check for conflicts
DetecGng Conflicts - Reflected GCI
John.balance == $100
John.balance -= $40
John.balance == $60
John.balance == $200
John.balance == $100
John.balance += $100
John.balance == $200
John.balance == $60
Spend $40 Add $100
$60
$200
Primary Secondary
Copyright 2016, oracle and/or its affiliates. All rights reserved 22

23. How to Use Conflict DetecGon and ResoluGon
Decide which
tables need
protecGng
For each table,
specify what to
do on conflicts
Record event in
excepGon table
ApplicaGon or
DBA can act on
content
Rollback the
conflicGng row
Rollback the
conflicGng
transacGon
Copyright 2016, oracle and/or its affiliates. All rights reserved 23

24. MySQL Cluster 7.4 Restart Improvements
• DuraGon of long-running maintenance acGviGes dominated by Data Node
restart Gmes
• MySQL Cluster 7.4 = 5.5x faster restarts!
• Benefits any kind of node restart
– Upgrades, add-node, …
• Benefits both ”manual” and MySQL Cluster Manager operaGons
• Accomplish 5x as much work during a single maintenance window
Make Data Node Restarts Fast!
Copyright 2016, oracle and/or its affiliates. All rights reserved 24

25. MySQL Cluster 7.4 Restart Improvements
• Verbose logging
– Task start and compleGon
– Data volume info
– Parallelism & wait Gme info
• NDBINFO for recent node restarts
• More documentaGon of concrete stages
• Goal: make proper analysis of a slow restart easy
– Determine cause; Detect pa}erns; Understand the impact of indexes, local
checkpoints, etc.
Observability Improvements
Copyright 2016, oracle and/or its affiliates. All rights reserved 25

26. MySQL Cluster 7.4 – Enhanced Memory ReporGng
• ndbinfo.memory_per_fragment memory usage informaGon for
each fragment replica, for each table, and each index
• Allocated memory, and how much of that is actually in use
• Exposes
– FragmentaGon of fixed and variable sized fragment pages
– Accurate Data and Index Memory use
– Comparison of Primary and Backup fragment usage
– ParGGon/Shard distribuGon and effecGveness
Copyright 2016, oracle and/or its affiliates. All rights reserved 26

27. Enhanced Memory ReporGng
mysql> CREATE TABLE simples (id INT NOT NULL AUTO_INCREMENT PRIMARY KEY) ENGINE=NDB;
mysql> SELECT node_id AS node, fragment_num AS frag, fixed_elem_alloc_bytes AS alloc_bytes,
fixed_elem_free_bytes AS free_bytes, fixed_elem_free_rows AS spare_rows FROM memory_per_fragment
WHERE fq_name LIKE '%simples%';
+------+------+-------------+------------+------------+
| node | frag | alloc_bytes | free_bytes | spare_rows |
+------+------+-------------+------------+------------+
| 1 | 0 | 131072 | 5504 | 172 |
| 1 | 2 | 131072 | 1280 | 40 |
| 2 | 0 | 131072 | 5504 | 172 |
| 2 | 2 | 131072 | 1280 | 40 |
| 3 | 1 | 131072 | 3104 | 97 |
| 3 | 3 | 131072 | 4256 | 133 |
| 4 | 1 | 131072 | 3104 | 97 |
| 4 | 3 | 131072 | 4256 | 133 |
+------+------+-------------+------------+------------+
See How Much Memory a Table is Using
Copyright 2016, oracle and/or its affiliates. All rights reserved 27

28. Enhanced Memory ReporGng
+------+------+-------------+------------+------------+
| node | frag | alloc_bytes | free_bytes | spare_rows |
+------+------+-------------+------------+------------+
| 1 | 0 | 131072 | 5504 | 172 |
| 1 | 2 | 131072 | 1280 | 40 |
| 2 | 0 | 131072 | 5504 | 172 |
| 2 | 2 | 131072 | 1280 | 40 |
| 3 | 1 | 131072 | 3104 | 97 |
| 3 | 3 | 131072 | 4256 | 133 |
| 4 | 1 | 131072 | 3104 | 97 |
| 4 | 3 | 131072 | 4256 | 133 |
+------+------+-------------+------------+------------+
mysql> DELETE FROM clusterdb.simples LIMIT 1;
+------+------+-------------+------------+------------+
| node | frag | alloc_bytes | free_bytes | spare_rows |
+------+------+-------------+------------+------------+
| 1 | 0 | 131072 | 5504 | 172 |
| 1 | 2 | 131072 | 1312 | 41 |
| 2 | 0 | 131072 | 5504 | 172 |
| 2 | 2 | 131072 | 1312 | 41 |
| 3 | 1 | 131072 | 3104 | 97 |
| 3 | 3 | 131072 | 4288 | 134 |
| 4 | 1 | 131072 | 3104 | 97 |
| 4 | 3 | 131072 | 4288 | 134 |
+------+------+-------------+------------+------------+
See How Memory is Made Available A_er Dele0ng Rows
Copyright 2016, oracle and/or its affiliates. All rights reserved 28

29. Enhanced Memory ReporGng
mysql> CREATE TABLE simples (id INT NOT NULL AUTO_INCREMENT, species VARCHAR(20) DEFAULT "Human", PRIMARY KEY(id,
species)) engine=ndb PARTITION BY KEY(species);
// Add some data
mysql> SELECT node_id AS node, fragment_num AS frag, fixed_elem_alloc_bytes AS alloc_bytes, fixed_elem_free_bytes
AS free_bytes, fixed_elem_free_rows AS spare_rows FROM ndbinfo.memory_per_fragment WHERE fq_name LIKE
'%simples%';
+------+------+-------------+------------+------------+
| node | frag | alloc_bytes | free_bytes | spare_rows |
+------+------+-------------+------------+------------+
| 1 | 0 | 0 | 0 | 0 |
| 1 | 2 | 196608 | 11732 | 419 |
| 2 | 0 | 0 | 0 | 0 |
| 2 | 2 | 196608 | 11732 | 419 |
| 3 | 1 | 0 | 0 | 0 |
| 3 | 3 | 0 | 0 | 0 |
| 4 | 1 | 0 | 0 | 0 |
| 4 | 3 | 0 | 0 | 0 |
+------+------+-------------+------------+------------+
Check Par00on/Shard Distribu0on and Effec0veness
Copyright 2016, oracle and/or its affiliates. All rights reserved 29

30. MySQL Cluster 7.4 – Enhanced AcGvity ReporGng
• ndbinfo.operations_per_fragment acGvity counters for each
fragment replica, for each table, and each index
• PK based and scan accesses – requests, bytes, rows, …
• Exposes
– How SQL statements map to individual tables and indices
– Query execuGon, use of indexes etc.
– LDM and node imbalances
– Hotspots and scan overloads
Copyright 2016, oracle and/or its affiliates. All rights reserved 30

31. MySQL Cluster 7.5
• Based on MySQL 5.7
– NaGve JSON datatype
– Generated Columns
• FuncGon Based Indexes
– Cost Based OpGmizer
– New SpaGal engine
– New hint syntax with addiGonal hints
– Memory instrumentaGon
– Secure by default
– Much more!
Copyright 2016, oracle and/or its affiliates. All rights reserved 31
NEW!

32. MySQL Cluster 7.5
• ConGnue Improving Supportability
– Overload controls via thread idle wait Gme
– Improve event buffer memory allocaGon
– Be}er and more descripGve error messages
– Dump command improvements
– Improved Lock reporGng
• ConGnue Improving Performance
– Faster backup & restore
Copyright 2016, oracle and/or its affiliates. All rights reserved 32
NEW!

33. MySQL Cluster 7.5
• ConGnue Improving Availability
– Convert ndb_binlog_index to ACID storage
– Make use of dynamic columns everywhere
• Support for online defragmentaGon and space reclamaGon
• Support for row transformaGons – encrypGon and compression
• ConGnue Improving Quality
– Refactor binlog injector in order to address many known issues
– Increase the test base
– More system correctness and fault injecGon tests
Copyright 2016, oracle and/or its affiliates. All rights reserved 33
NEW!

34. MySQL Cluster 7.5
• ConGnue Improving Capacity
– Enable individual fragments larger than 16GB (now up to 128TB)
• Improved Read Scale-Out
– The introducGon of “read from backup” for tables
• Localized reads on every node in a node group
– The introducGon of “global” tables (fully replicated)
• Localized reads on every node in the Cluster
• Support for Increased Redundancy
– Support for up to 4 nodes per node group (1-3 replicas per shard)
Copyright 2016, oracle and/or its affiliates. All rights reserved 34
NEW!

35. MySQL Cluster Best PracGces
Copyright 2016, oracle and/or its affiliates. All rights reserved 35

36. When to Consider MySQL Cluster
l UpGme requirements
l Cost per minute of downGme?
l Failure versus maintenance?
l Scalability demands
l Sharding for write scaling?
l Latency demands
l Cost of each millisecond?
l ApplicaGon agility
l Developer languages and frameworks?
l SQL and NoSQL?
Copyright 2016, oracle and/or its affiliates. All rights reserved 36

37. General Usage Considerations
• MySQL Cluster is designed for
– Short transactions
– Single key reads and writes
– Many parallel transactions
• Utilize simple access patterns for high running transactions
– Use efficient scans and batching interfaces
– Push down joins yield huge performance improvements for JOIN operations
– MySQL Cluster 7.4 speeds up table scans
• Storage Engine is configurable for each table…InnoDB or NDB
MySQL Cluster EvaluaGon Guide
h}p://mysql.com/why-mysql/white-papers/mysql_cluster_eval_guide.php
Copyright 2016, oracle and/or its affiliates. All rights reserved 37

38. D1 D2 D3 D4
API
----------
----------
----------
TC TC TC TC
1
2
Accessing Data: PK Lookups
• Single TransacGon
Coordinator (TC), no
addiGonal data
distribuGon handling
• First statement decides TC
• Keep transacGons short!
Copyright 2016, oracle and/or its affiliates. All rights reserved 38

39. D1 D2 D3 D4
API
----------
----------
----------
TC TC TC TC
1
2
3
Accessing data: Unique Secondary Key Lookups
• Secondary keys
implemented as hidden/
system tables
• Hidden tables have new
secondary key as PK and
base table’s PK as value
• Data may reside on same
node or other node
Copyright 2016, oracle and/or its affiliates. All rights reserved 39

40. D1 D2 D3 D4
API
----------
----------
----------
TC TC TC TC
1 3
2
Accessing data: Range Scans and Joins
• TC is chosen using round-
robin
• Data nodes return data
directly to API
• Flow:
– Choose data node
– Send request to all LDM
– Send data to API
Copyright 2016, oracle and/or its affiliates. All rights reserved 40

41. SELECT SUM(population) FROM towns WHERE
country=“UK”;
SELECT SUM(population) FROM towns WHERE
town=“Boston”;
• ParGGon selected using hash on
specified ParGGon Key
– Primary Key by default
– User can override in table definiGon
• MySQL Server (or NDB API) will
a}empt to send transacGon to the
correct data node
– If all data for the transacGon are in the
same parGGon, less messaging ->
be}er scalability
• Aim to have all rows for high-
running queries in same parGGon
DistribuGon Aware Apps
Copyright 2016, oracle and/or its affiliates. All rights reserved 41

42. • Extend parGGon awareness over
mulGple tables
• Same rule – aim to have all data for
instance of high running
transacGons in the same parGGon
– ALTER TABLE service_ids
PARTITION BY KEY(sub_id);
DistribuGon Aware – MulGple Tables
Copyright 2016, oracle and/or its affiliates. All rights reserved 42

43. Faster Joins – Pushdown Joins
• AcGvated when ndb_join_pushdown is on (default)
• Rules for a join to be pushed down:
1. Joined columns use idenGcal types
2. No reference to BLOB or TEXT columns
3. No explicit lock
4. Child tables in the Join must be accessed using ref, eq_ref, or const
5. Tables not explicitly parGGoned by [LINEAR] HASH, LIST, or RANGE
6. Query execuGon plan doesn’t use mysqld’s internal ‘join buffer'
7. If root of join is of the eq_ref or const type, child tables must be joined by eq_ref
• Run ANALYZE TABLE <table_name> on each table once
• Use EXPLAIN to see what components are being pushed down:
– For example: Extra: Child of 'd' in pushed join@1
Copyright 2016, oracle and/or its affiliates. All rights reserved 43

44. Global Checkpoints and Logging
• Global Checkpoint Protocol/Group
Commit (GCP)
– REDO log, synchronized between the
Data Nodes
– Writes transacGons that have been
recorded in the REDO log buffer to
REDO log on disk
– Frequency controlled by
TimebetweenGlobalCheckpoints
• Default is 2000ms
– Size of the REDO log is set using
NumOfFragmentLogFiles
Copyright 2016, oracle and/or its affiliates. All rights reserved 44

45. Local Checkpoints and Logging
• Local Checkpoint Protocol (LCP)
– Flushes the Data Nodes’ data to disk.
Aper 2 LCPs the REDO log is trimmed
– Frequency controlled by
TimebetweenLocalCheckpoints
• Specifies the amount of data that can change
before flushing to disk
• Not actually Gme! Base-2 logarithm of the
number of 4-byte words
• Example: Default value of 20 bytes means
4*2^20 = 4MB of data changes
Copyright 2016, oracle and/or its affiliates. All rights reserved 45

46. Checkpoints and Logging
Local & Redo
• LCP and REDO Log are used to bring
the cluster back online
– Following system failure or planned
shutdown
1. Data Nodes are restored using the
latest LCP
2. REDO logs are applied unGl the
latest GCP
Copyright 2016, oracle and/or its affiliates. All rights reserved 46

47. MySQL Cluster Hardware Requirements
• TCP/IP or Infiniband
– Minimum 100Mbps, 1Gbps or higher recommended
– MySQL Cluster dedicated network: separate Data Nodes from MySQL Servers
• Homogeneous environments for each layer: API nodes & Data Nodes
• Machines
– Linux, Solaris, Windows, OS X
• h}ps://www.mysql.com/support/supportedplaaorms/cluster.html
– Min. 1Gb RAM
– CPU, RAM, DISK: “the more and faster, the be}er”
Copyright 2016, oracle and/or its affiliates. All rights reserved 47

48. MySQL Cluster Hardware SelecGon
• 4 - 24 x86_64 bit CPU threads (vCPUs)
• Minimum 4GB of RAM
– Memory not as criGcal at this layer
– Requirements influenced by connecGons and buffers
• 2 Network Interface Cards (NICs)
• 2 Power Supply Units (PSUs)
• Linux, Solaris, Windows, or OS X operaGng system
MySQL Servers
Copyright 2016, oracle and/or its affiliates. All rights reserved 48

49. MySQL Cluster Hardware SelecGon
• Up to 64 x86_64 bit CPU threads (vCPUs)
– High frequency: enables faster processing of messages
• CalculaGng RAM per Data Node (in-memory database)
– Database Size * # Replicas * 1.25 / # data nodes
– 50GB database * 2 replicas * 1.25 / 2 data nodes = 64GB of RAM
– Non-indexed columns can be stored on disk, reduces RAM requirements
• 2 NICs and 2 PSUs
• Deep-dive into network best-pracGces in Reference Architecture Guide
• Use MaxNoOfExecutionThreads or ThreadConfig to configure how threads
should be allocated
Data Nodes
Copyright 2016, oracle and/or its affiliates. All rights reserved 49

50. Entry-Level Mid-Range High-End
1 x SATA 7200 RPM
• For read-mostly
• No redundancy
(but other data node is
the mirror)
1 x SAS 10K RPM or SSD
• Heavy duty (many MB/s)
• No redundancy
(but other data node is
the mirror)
4 x SAS 10-15K RPM or SSDs
• Heavy duty (many MB/s)
• Disk redundancy (RAID1+0 / 10),
hot swappable
LCP / REDOLOG
MySQL Cluster Hardware SelecGon
• REDO, LCP, BACKUP – wri}en sequenGally in small chunks (256KB)
• If possible, use ODirect=1
Copyright 2016, oracle and/or its affiliates. All rights reserved 50
Disk Subsystem for Checkpoin0ng
LCP / REDOLOG LCP / REDOLOG

51. Recommended Minimum High-End RecommendaGon
2 x SAS 10K RPM
or 2 x SSD
TABLESPACE
LCP
REDO LOG
UNDO LOG
UNDOLOG
LCP
(REDO LOG / UNDO LOG)
TABLESPACE 1
TABLESPACE 2
4 x SAS 10-15K RPM or SSD
(REDO LOG) (REDO LOG)
MySQL Cluster Hardware SelecGon:
• Use High-End RecommendaGon for heavy read/write workloads
• Having TABLESPACE on separate disk is good for read performance
• Enable WRITE_CACHE on block storage devices
Copyright 2016, oracle and/or its affiliates. All rights reserved 51
Disk Based Tables

52. Zone B Zone A
LocaGon of Nodes
• Minority of data nodes lost
– Isolated minority will shutdown
• Majority pool (> 50% of the data
nodes and one from each Node
Group)
– Remains viable
– Exclude failed nodes from cluster
operaGons
Copyright 2016, oracle and/or its affiliates. All rights reserved 52
Handling Network Par00ons
Node
Group 1
Node
Group 2

53. Zone B Zone A
LocaGon of Nodes
• Network connecGvity lost between
zones
• Each pool has 50% of data nodes and
1 from each Node Group
– Each viable, but Zones will diverge if they
both live (split-brain)
– So who lives on and who dies?
– We need arbitraGon
Copyright 2016, oracle and/or its affiliates. All rights reserved 53
Handling Network Par00ons
Node
Group 1
Node
Group 2

54. Zone B Zone A
LocaGon of Nodes
• Arbitrator decides
– Contacted by data nodes from each zone
– Tells those in one zone to shutdown
– Tells those in other zone to stay up
• Vital that arbitrator shares no single
point of failure with either availability
zone
• Default arbitrator is an ndb_mgmd
node but can also be a mysqld
node
Copyright 2016, oracle and/or its affiliates. All rights reserved 54
Avoiding Split-Brain During Network Par00ons
Node
Group 1
Node
Group 2
Zone C
Arbitrator

55. Spli…ng MySQL Cluster Over MulGple Sites
• The greater the latency between sites, the higher the impact on
performance
• Target latency should be <= 10 ms; 20 ms acceptable
– Test at the network layer with 1000 byte packets under load
• Bandwidth requirements dependent on traffic but aim for 1 Gbps+ (100
Mbps for a low traffic Cluster)
• Simplest WAN topology possible (fewer points of failure/failover latency)
• Typical WAN failover Gmes should be short enough not to trigger STONITH
acGviGes within Cluster
Copyright 2016, oracle and/or its affiliates. All rights reserved 55
WAN Engineering

56. • Asynchronous replicaGon between
MySQL Clusters
• AcGve-AcGve MulG-Homed
– Update anywhere
– Conflict detecGon
• ApplicaGon noGfied through excepGon tables
• Can opt to have conflicts resolved
automaGcally
– AutomaGc conflict resoluGon
• ConflicGng transacGon and dependent ones
are rolled-back
• No changes to applicaGon / schema
AcGve-AcGve Geo-ReplicaGon (Reminder)
Copyright 2016, oracle and/or its affiliates. All rights reserved 56

57. MySQL Cluster Geographic ReplicaGon
Copyright 2016, oracle and/or its affiliates. All rights reserved 57
Implementa0on

58. MySQL Cluster Geographic ReplicaGon
Copyright 2016, oracle and/or its affiliates. All rights reserved 58
Scenarios for replica0ng between clusters

59. • ParGGon customers across mulGple
clusters, distributed by region to
opGmize low latency access
• Each sub-cluster is replicated for
High Availability
• AcGve/passive or AcGve/AcGve
59
Geographic ReplicaGon for Low-Latency Local Access
Copyright 2016, oracle and/or its affiliates. All rights reserved
Cluster 4
Cluster 1
Cluster 1B
Cluster 1C
Cluster 2
Cluster 2B
Cluster 2C
Cluster 3
Cluster 3B
Cluster 3C
Cluster 4C
Cluster 4B

60. MySQL Cluster Backups
• Backup of NDB tables
– Online – can have ongoing transacGons
– Consistent – only commi}ed data and changes
are backed up
• ndb_mgm -e "START BACKUP"
– Copy backup files from data nodes to safe
persistent locaGon
– Non-NDB tables must be backed up separately
– MySQL system tables are stored in MyISAM or
InnoDB
• You want to backup (for each MySQL
Server)
– mysql database/schema
– Triggers, rouGnes, events, ...
• Use mysqldump
– mysqldump mysql > mysql.sql
– mysqldump --no-data -A >
schemas.sql
– mysqldump --no-data --no-create-
info -R > routines.sql
• Copy my.cnf & config.ini files
Copyright 2016, oracle and/or its affiliates. All rights reserved 60

61. MySQL Cluster Backups
mcm> backup cluster mycluster;
mcm> list backups mycluster;
mcm> restore cluster -I <backup-id> mycluster
Using MySQL Cluster Manager
Copyright 2016, oracle and/or its affiliates. All rights reserved 61

62. Features to Use With CauGon
• Large BLOBs
• Disk based tables
• Enforced Foreign Keys
• Large range scans and joins
• Spli…ng a single MySQL Cluster over mulGple data centers
• AcGve-AcGve Geographic ReplicaGon
– Fine to use but best if conflicts can be avoided on the applicaGon side
Copyright 2016, oracle and/or its affiliates. All rights reserved 62
Poten0al Impacts to Performance, Capacity, & Stability

63. Machine 1
Typical Architectures
Copyright 2016, oracle and/or its affiliates. All rights reserved 63
Enterprise - Minimal HA Configura0on
MySQL
Server
Data Node
App
Machine 2
MySQL
Server
Data Node
App
Machine 3
Mgmt
Node

64. Machine 4 Machine 3
Machine 1
Typical Architectures
Copyright 2016, oracle and/or its affiliates. All rights reserved 64
Enterprise - Small
MySQL
Server
Data Node
App
Machine 2
MySQL
Server
Data Node
App
Mgmt
Node
Mgmt
Node

65. Machine 4 Machine 3
Machine 1
Typical Architectures
Copyright 2016, oracle and/or its affiliates. All rights reserved 65
Enterprise - Scale-out
MySQL
Server
Data Node
App
Machine 2
MySQL
Server
Data Node
App
Mgmt
Node
Mgmt
Node
Machine 8 Machine 7
Machine 5
25/10/17
MySQL
Server
Data Node
App
Machine 6
MySQL
Server
Data Node
App
Machine 9
MySQL
Server
App
MySQL
Server
App

66. Typical Architectures
Copyright 2016, oracle and/or its affiliates. All rights reserved 66
Telco/Gaming
Data Node
App
Data Node
App
Mgmt
Node
Mgmt
Node
25/10/17
Data Node
App
MySQL
Server
Data Node
App
MySQL
Server

67. Typical Architectures
Copyright 2016, oracle and/or its affiliates. All rights reserved 67
Telco/Gaming
Data Node
App
Data Node
App
Mgmt
Node
Mgmt
Node
25/10/17
Data Node
App
MySQL
Server
Data Node
App
MySQL
Server
Data Node
App
Data Node
App

68. ‘Typical’ Architectures
Copyright 2016, oracle and/or its affiliates. All rights reserved 68
Telco/Gaming - scale
25/10/17

69. Site 2 Site 1
‘Typical’ Architectures
Copyright 2016, oracle and/or its affiliates. All rights reserved 69
Telco/Gaming – Geographic Replica0on
25/10/17

70. MySQL Cluster Support
Issue Detected
SR Created
Support Works on SR
If Bug
Bug Report Opened
BPS (priority) Set Based on SR Fields
Bug Fix Created
Hot Fix? Maintenance Release
If Not Bug
Support Responds to Customer Via SR
Copyright 2016, oracle and/or its affiliates. All rights reserved 70

71. MySQL
Cluster
CGE
MySQL
Cluster
Manager
MySQL
Enterprise
Scalability
MySQL
Enterprise
Audit
MySQL
Enterprise
Security
Oracle
Premier
LifeGme
Support
Oracle
Product
CerGficaGons
MySQL
Enterprise
Monitor
MySQL
Workbench
Copyright 2016, oracle and/or its affiliates. All rights reserved 71

72. Enhancing DevOps Agility, Reducing DownGme
Automated Management
• Start/stop node or
enGre cluster
• On-Line scaling
• On-Line reconfiguraGon
• On-Line upgrades
• On-Line backup &
restore
• Import running Cluster
Self-Healing
• Node monitoring
• Auto-recovery
extended to SQL and
mgmt nodes
HA OperaGons
• Cluster-wide
configuraGon
consistency
• Persistent
configuraGons
• HA agents
Copyright 2016, oracle and/or its affiliates. All rights reserved 72

73. Without MySQL Cluster Manager
• 1 x preliminary check of cluster state
• 8 x ssh commands per server
• 8 x per-process stop commands
• 4 x scp of config files (2 x mgmd & 2 x mysqld)
• 8 x per-process start commands
• 8 x checks for started and re-joined processes
• 8 x process compleGon verificaGons
• 1 x verify compleGon of the whole cluster
• Excludes manual ediGng of each configuraGon file
• Total: 46 commands
– 2.5 hours of aNended opera8on
With MySQL Cluster Manager
mcm> upgrade cluster
--package=7.4 mycluster;
• Total: 1 Command
– UnaNended Opera8on
Sopware Upgrade – Example of MCM Benefits
Copyright 2016, oracle and/or its affiliates. All rights reserved 73

74. MySQL Cluster Manager 1.3 GA
“Unmanaged” producGon Cluster
mcm> create cluster --import
mcm> import config [--dryrun]
mcm> import cluster [--dryrun]
Cluster now managed by MCM
Import a running Cluster into MCM
Copyright 2016, oracle and/or its affiliates. All rights reserved 74

75. MySQL Cluster Manager 1.4 GA
Autotune Your Cluster
Copyright 2016, oracle and/or its affiliates. All rights reserved 75
mcm> autotune --writeload=high realtime mycluster;
+-----------------------------------------------------+
| Command result |
+-----------------------------------------------------+
| Cluster successfully autotuned to template realtime |
+-----------------------------------------------------+
1 row in set (2 min 58.09 sec)

76. MySQL Enterprise Monitor
76 Copyright 2016, oracle and/or its affiliates. All rights reserved

77. MySQL Enterprise Monitor
77 Copyright 2016, oracle and/or its affiliates. All rights reserved