Amazon Aurora is a MySQL-compatible relational database engine that combines the speed and availability of high-end commercial databases with the simplicity and cost-effectiveness of open source databases. Amazon Aurora is a disruptive technology in the database space, bringing a new architectural model and distributed system techniques to provide far higher performance, availability and durability than previously available using conventional monolithic database techniques. In this session, we will do a deep-dive into some of the key innovations behind Amazon Aurora, discuss best practices and configurations, and share customer experiences from the field. Learning Objectives: • Learn about the capabilities and features of Amazon Aurora and its new features • Learn about the benefits of Amazon Aurora and how it delivers 5x the performance and 1/10th the cost • Learn about the different use cases • Learn how to get started using Amazon Aurora

1. © 2016, Amazon Web Services, Inc. or its Affiliates. All rights reserved.
Steve Abraham – Solutions Architect
December 14, 2016
Deep Dive on Amazon Aurora
Including New Features

2. Agenda
 What is Aurora?
 Review of Aurora performance
 New performance enhancements
 Review of Aurora availability
 New availability enhancements
 Other recent and upcoming feature enhancements

3. Open source compatible relational database
Performance and availability of
commercial databases
Simplicity and cost-effectiveness of
open source databases
What is Amazon Aurora?

4. Performance

5. Write performance Read performance
Aurora scales with instance size for both read and write.
Aurora MySQL 5.6 MySQL 5.7
Scaling with instance sizes

6. Aurora vs. RDS MySQL – r3.4XL, MAZ
Aurora 3X faster on r3.4xlarge
Real-life data – gaming workload

7. Do fewer IOs
Minimize network packets
Cache prior results
Offload the database engine
Do Less Work
Process asynchronously
Reduce latency path
Use lock-free data structures
Batch operations together
Be More Efficient
Databases are all about i/o
Network-attached storage is all about packets/second
High-throughput processing is all about context switches
How did we achieve this?

8. BINLOG DATA DOUBLE-WRITELOG FRM FILES
TYPE OF WRITE
Mysql with replica
EBS mirrorEBS mirror
AZ 1 AZ 2
Amazon S3
EBS
Amazon Elastic
Block Store (EBS)
Primary
Instance
Replica
Instance
1
2
3
4
5
 Issue write to EBS – EBS issues to mirror, ack when both done
 Stage write to standby instance through DRBD
 Issue write to EBS on standby instance
Io flow
 Steps 1, 3, 4 are sequential and synchronous
 This amplifies both latency and jitter
 Many types of writes for each user operation
 Have to write data blocks twice to avoid torn writes
Observations
 780K transactions
 7,388K I/Os per million txns (excludes mirroring, standby)
 Average 7.4 I/Os per transaction
Performance
30 minute sysbench writeonly workload, 100GB dataset, RDS multiaz, 30K PIOPS
IO traffic in MySQL

9. AZ 1 AZ 3
Primary
Instance
Amazon S3
AZ 2
Replica
Instance
ASYNC
4/6 QUORUM
DISTRIBUTED
WRITES
Replica
Instance
BINLOG DATA DOUBLE-WRITELOG FRM FILES
TYPE OF WRITE
Amazon Aurora
 Boxcar redo log records – fully ordered by LSN
 Shuffle to appropriate segments – partially ordered
 Boxcar to storage nodes and issue writes
IO flow
 Only write redo log records; all steps asynchronous
 No data block writes (checkpoint, cache replacement)
 6X more log writes, but 9X less network traffic
 Tolerant of network and storage outlier latency
Observations
 27,378K transactions 35X MORE
 950K I/Os per 1M txns (6X amplification) 7.7X LESS
Performance
IO traffic in Aurora

10. LOG RECORDS
Primary
Instance
INCOMING QUEUE
STORAGE NODE
S3 BACKUP
1
2
3
4
5
6
7
8
UPDATE
QUEUE
ACK
HOT
LOG
DATA
BLOCKS
POINT IN TIME
SNAPSHOT
GC
SCRUB
COALESCE
SORT
GROUP
PEER TO PEER GOSSIPPeer
Storage
Nodes
 All steps are asynchronous
 Only steps 1 and 2 are in foreground latency path
 Input queue is 46X less than MySQL (unamplified, per node)
 Favor latency-sensitive operations
 Use disk space to buffer against spikes in activity
Observations
IO Flow
1. Receive record and add to in-memory queue
2. Persist record and ACK
3. Organize records and identify gaps in log
4. Gossip with peers to fill in holes
5. Coalesce log records into new data block versions
6. Periodically stage log and new block versions to S3
7. Periodically garbage collect old versions
8. Periodically validate CRC codes on blocks
IO traffic in Aurora (Storage Node)

11. MySQL Master
30% Read
70% Write
MySQL Replica
30% New Reads
70% Write
SINGLE-THREADED
BINLOG APPLY
Data Volume Data Volume
Logical: Ship SQL statements to Replica
 Write workload similar on both instances
 Independent storage
 Can result in data drift between Master and Replica
Physical: Ship redo from Master to Replica
 Replica shares storage. No writes performed
 Cached pages have redo applied
 Advance read view when all commits seen
PAGE CACHE
UPDATE
Aurora Master
30% Read
70% Write
Aurora Replica
100% New Reads
Shared Multi-AZ Storage
IO traffic in Aurora Replicas
My SQL read scaling Amazon Aurora read scaling

12. “In MySQL, we saw replica lag spike to almost 12 minutes which is almost
absurd from an application’s perspective. With Aurora, the maximum read
replica lag across 4 replicas never exceeded 20 ms.”
Real-life data - read replica latency

13. Read
Write
Commit
Read
Read
T1
Commit (T1)
Commit (T2)
Commit (T3)
LSN 10
LSN 12
LSN 22
LSN 50
LSN 30
LSN 34
LSN 41
LSN 47
LSN 20
LSN 49
Commit (T4)
Commit (T5)
Commit (T6)
Commit (T7)
Commit (T8)
LSN GROWTH
Durable LSN at head-node
COMMIT QUEUE
Pending commits in LSN order
TIME
GROUP
COMMIT
Transactions
Read
Write
Commit
Read
Read
T1
Read
Write
Commit
Read
Read
Tn
Traditional Approach Amazon Aurora
 Maintain a buffer of log records to write out to disk
 Issue write when buffer full or time out waiting for writes
 First writer has latency penalty when write rate is low
 Request I/O with first write, fill buffer till write picked up
 Individual write durable when 4 of 6 storage nodes ACK
 Advance DB Durable point up to earliest pending ACK
Asynchronous group commits

14.  Re-entrant connections multiplexed to active threads
 Kernel-space epoll() inserts into latch-free event queue
 Dynamically size threads pool
 Gracefully handles 5000+ concurrent client sessions on r3.8xl
 Standard MySQL – one thread per connection
 Doesn’t scale with connection count
 MySQL EE – connections assigned to thread group
 Requires careful stall threshold tuning
Clientconnection
Clientconnection
LATCH FREE
TASK QUEUE
epoll()
MySQL thread model Aurora thread model
Adaptive thread pool

15. Scan
Delete
Scan
Delete
Insert
Scan Scan
Insert
Delete
Scan
Insert
Insert
MySQL lock manager Aurora lock manager
 Same locking semantics as MySQL
 Concurrent access to lock chains
 Multiple scanners allowed in an individual lock chains
 Lock-free deadlock detection
Needed to support many concurrent sessions, high update throughput
Aurora lock management

16. New Performance Enhancements

17. Catalog concurrency:
Improved data dictionary synchronization and cache eviction.
NUMA aware scheduler:
Aurora scheduler is now NUMA aware. Helps scale on
multi-socket instances.
Read views:
Aurora now uses a latch-free concurrent read-view algorithm
to construct read views.
0
100
200
300
400
500
600
700
MySQL 5.6 MySQL 5.7 Aurora 2015 Aurora 2016
In thousands of read requests/sec
* R3.8xlarge instance, <1GB dataset using Sysbench
25% Throughput gain
Cached read performance

18. Smart scheduler:
Aurora scheduler now dynamically assigns threads
between IO heavy and CPU heavy workloads.
Smart selector:
Aurora reduces read latency by selecting the copy of
data on a storage node with best performance
Logical read ahead (LRA):
We avoid read IO waits by prefetching pages based
on their order in the btree.
0
20
40
60
80
100
120
MySQL 5.6 MySQL 5.7 Aurora 2015 Aurora 2016
In thousands of requests/sec
* R3.8xlarge instance, 1TB dataset using Sysbench
Non-cached read performance
10% Throughput gain

19. Scan
Delete
Scan
Delete
Insert
Scan Scan
Insert
Delete
Scan
Insert
Insert
Hot row contention
MySQL lock manager Aurora lock manager
Highly contended workloads had high memory and CPU
 1.9 (Nov) – lock compression (bitmap for hot locks)
 1.9 – replace spinlocks with blocking futex – up to 12x reduction in CPU, 3x improvement in throughput
 December – use dynamic programming to release locks: from O(totalLocks * waitLocks) to O(totalLocks)
Throughput on Percona TPC-C 100 improved 29x (from 1,452 txns/min to 42,181 txns/min)

20. MySQL 5.6 MySQL 5.7 Aurora Improvement
500 connections 6,093 25,289 73,955 2.92x
5000 connections 1,671 2,592 42,181 16.3x
Percona TPC-C – 10GB
* Numbers are in tpmC, measured using release 1.10 on an R3.8xlarge, MySQL numbers using RDS and EBS with 30K PIOPS
MySQL 5.6 MySQL 5.7 Aurora Improvement
500 connections 3,231 11,868 70,663 5.95x
5000 connections 5,575 13,005 30,221 2.32x
Percona TPC-C – 100GB
Hot row contention

21.  Accelerates batch inserts sorted by
primary key – works by caching the cursor
position in an index traversal.
 Dynamically turns itself on or off based on
data pattern
 Avoids contention in acquiring latches
while navigating down the tree.
 Bi-directional, works across all insert
statements
 LOAD INFILE, INSERT INTO SELECT, INSERT
INTO REPLACE and, Multi-value inserts.
Index
R4 R5R2 R3R0 R1 R6 R7 R8
Index
Root
Index
R4 R5R2 R3R0 R1 R6 R7 R8
Index
Root
MySQL: Traverses B-tree starting from root for all inserts
Aurora: Inserts avoids index traversal;
Insert performance

22.  MySQL 5.6 leverages Linux read ahead – but
this requires consecutive block addresses in
the btree. It inserts entries top down into the
new btree, causing splits and lots of logging.
 Aurora’s scan pre-fetches blocks based on
position in tree, not block address.
 Aurora builds the leaf blocks and then the
branches of the tree.
 No splits during the build.
 Each page touched only once.
 One log record per page.
2-4X better than MySQL 5.6 or MySQL 5.7
0
2
4
6
8
10
12
r3.large on 10GB
dataset
r3.8xlarge on
10GB dataset
r3.8xlarge on
100GB dataset
Hours RDS MySQL 5.6 RDS MySQL 5.7 Aurora 2016
Faster index build

23. Need to store and reason about spatial data
 E.g. “Find all people within 1 mile of a hospital”
 Spatial data is multi-dimensional
 B-Tree indexes are one-dimensional
Aurora supports spatial data types (point/polygon)
 GEOMETRY data types inherited from MySQL 5.6
 This spatial data cannot be indexed
Two possible approaches:
 Specialized access method for spatial data (e.g., R-Tree)
 Map spatial objects to one-dimensional space & store in
B-Tree - space-filling curve using a grid approximation
A
B
A A
A A
A A A
B
B
B
B
B
A COVERS B
B COVEREDBY A
A CONTAINS B
B INSIDE A
A TOUCH B
B TOUCH A
A OVERLAPBDYINTERSECT B
B OVERLAPBDYINTERSECT A
A OVERLAPBDYDISJOINT B
B OVERLAPBDYDISJOINT A
A EQUAL B
B EQUAL A
A DISJOINT B
B DISJOINT A
A COVERS B
B ON A
Why Spatial Index

24. Z-index used in Aurora
Challenges with R-Trees
 Keeping it efficient while balanced
 Rectangles should not overlap or cover empty space
 Degenerates over time
 Re-indexing is expensive
R-Tree used in MySQL 5.7
Z-index (dimensionally ordered space filling curve)
 Uses regular B-Tree for storing and indexing
 Removes sensitivity to resolution parameter
 Adapts to granularity of actual data without user declaration
 Eg GeoWave (National Geospatial-Intelligence Agency)
Spatial Indexes in Aurora

25. . . . . . . . . . . . . .
. . . . . . . . . . . . .
. . . . . . . . . . . . .
. . . . . . . . . . .. .
. . . . . . . . . . . . .
* r3.8xlarge using Sysbench on <1GB dataset
* Write Only: 4000 clients, Select Only: 2000 clients, ST_EQUALS
0
20000
40000
60000
80000
100000
120000
140000
Select-only (reads/sec) Write-only (writes/sec)
Aurora
MySQL 5.7
Spatial Index Benchmarks
Sysbench – points and polygons

26. Availability
“Performance only matters if your database is up”

27.  Storage volume automatically grows up to 64 TB
 Quorum system for read/write; latency tolerant
 Peer to peer gossip replication to fill in holes
 Continuous backup to S3 (built for 11 9s durability)
 Continuous monitoring of nodes and disks for repair
 10GB segments as unit of repair or hotspot rebalance
 Quorum membership changes do not stall writes
AZ 1 AZ 2 AZ 3
Amazon S3
Storage Durability

28.  Aurora cluster contains primary node
and up to fifteen replicas
 Failing database nodes are
automatically detected and replaced
 Failing database processes are
automatically detected and recycled
 Customer application may scale-out
read traffic across replicas
 Replica automatically promoted on
persistent outage
AZ 1 AZ 3AZ 2
Primary
Node
Primary
Node
Primary
Node
Primary
Node
Primary
Node
Secondary
Node
Primary
Node
Primary
Node
Secondary
Node
More Replicas

29. Segment snapshot Log records
Recovery point
Segment 1
Segment 2
Segment 3
Time
Continuous Backup
 Take periodic snapshot of each segment in parallel; stream the redo logs to Amazon S3
 Backup happens continuously without performance or availability impact
 At restore, retrieve the appropriate segment snapshots and log streams to storage nodes
 Apply log streams to segment snapshots in parallel and asynchronously

30. Traditional Databases
 Have to replay logs since the last checkpoint
 Typically 5 minutes between checkpoints
 Single-threaded in MySQL; requires a large
number of disk accesses
Amazon Aurora
 Underlying storage replays redo records
on demand as part of a disk read
 Parallel, distributed, asynchronous
 No replay for startup
Checkpointed Data Redo Log
Crash at T0 requires
a re-application of the
SQL in the redo log since
last checkpoint
T0 T0
Crash at T0 will result in redo logs being
applied to each segment on demand, in
parallel, asynchronously
Instant Crash Recovery

31.  We moved the cache out of the
database process
 Cache remains warm in the event
of database restart
 Lets you resume fully loaded
operations much faster
 Instant crash recovery + survivable
cache = quick and easy recovery
from DB failures
SQL
Transactions
Caching
SQL
Transactions
Caching
SQL
Transactions
Caching
Caching process is outside the DB process
and remains warm across a database restart
Survivable Caches

32. App
RunningFailure Detection DNS Propagation
Recovery Recovery
DB
Failure
MYSQL
App
Running
Failure Detection DNS Propagation
Recovery
DB
Failure
AURORA WITH MARIADB DRIVER
15-20 sec
3-20 sec
Faster Fail-Over

33. 0.00%
5.00%
10.00%
15.00%
20.00%
25.00%
30.00%
35.00%
1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 33 35
0 - 5s – 30% of fail-overs
0.00%
5.00%
10.00%
15.00%
20.00%
25.00%
30.00%
35.00%
40.00%
45.00%
50.00%
1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 33 35
5 - 10s – 40% of fail-overs
0%
10%
20%
30%
40%
50%
60%
1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 33 35
10 - 20s – 25% of fail-overs
0%
5%
10%
15%
20%
1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 33 35
20 - 30s – 5% of fail-overs
Database fail-over time

34. New Availability Enhancements

35. You also incur availability disruptions when you
1. Patch your database software
2. Modify your database schema
3. Perform large scale database reorganizations
4. DBA errors requiring database restores
Availability is about more than HW failures

36. Networking
state
Application
state
Storage Service
App
state
Net
state
App
state
Net
state
BeforeZDP
New DB
Engine
Old DB
Engine
New DB
Engine
Old DB
Engine
WithZDP
User sessions terminate
during patching
User sessions remain
active through patching
Storage Service
Zero downtime patching

37. We have to go to our current patching model when we
can’t park connections:
 Long running queries
 Open transactions
 Bin-log enabled
 Parameter changes pending
 Temporary tables open
 Locked Tables
 SSL connections open
 Read replicas instances
We are working on addressing the above.
Zero downtime patching – current constraints

38. Create a copy of a database without
duplicate storage costs
 Creation of a clone is nearly instantaneous – we
don’t copy data
 Data copy happens only on write – when
original and cloned volume data differ
Typical use cases:
 Clone a production DB to run tests
 Reorganize a database
 Save a point in time snapshot for analysis
without impacting production system.
Production database
Clone Clone
Clone
Dev/test
applications
Benchmarks
Production
applications
Production
applications
Database cloning

39. Page
1
Page
2
Page
3
Page
4
Source Database
Page
1
Page
3
Page
2
Page
4
Cloned database
Shared Distributed Storage System: physical pages
Both databases reference same pages on the shared
distributed storage system
Page
1
Page
2
Page
3
Page
4
How does it work?

40. Page
1
Page
2
Page
3
Page
4
Page
5
Page
1
Page
3
Page
5
Page
2
Page
4
Page
6
Page
1
Page
2
Page
3
Page
4
Page
5
Page
6
As databases diverge, new pages are added appropriately to each
database while still referencing pages common to both databases
Page
2
Page
3
Page
5
Shared Distributed Storage System: physical pages
Source Database Cloned database
How does it work? (contd.)

41.  Full Table copy; rebuilds all indexes – can take hours
or days to complete.
 Needs temporary space for DML operations
 DDL operation impacts DML throughput
 Table lock applied to apply DML changes
Index
LeafLeafLeaf Leaf
Index
Root
table name operation column-name time-stamp
Table 1
Table 2
Table 3
add-col
add-col
add-col
column-abc
column-qpr
column-xyz
t1
t2
t3
 We add an entry to the metadata table and use schema
versioning to decode the blo.
 Added a modify-on-write primitive to upgrade the block
to the latest schema when it is modified.
 Currently support add NULLable column at end of table
 Priority is to support other add column, drop/reortder,
modify datatypes.
MySQL Amazon Aurora
Online DDL: Aurora vs. MySQL

42. On r3.large
On r3.8xlarge
Aurora MySQL 5.6 MySQL 5.7
10GB table 0.27 sec 3,960 sec 1,600 sec
50GB table 0.25 sec 23,400 sec 5,040 sec
100GB table 0.26 sec 53,460 sec 9,720 sec
Aurora MySQL 5.6 MySQL 5.7
10GB table 0.06 sec 900 sec 1,080 sec
50GB table 0.08 sec 4,680 sec 5,040 sec
100GB table 0.15 sec 14,400 sec 9,720 sec
Online DDL performance

43. t0 t1 t2
t0 t1
t2
t3 t4
t3
t4
Rewind to t1
Rewind to t3
Invisible Invisible
What is online point in time restore?
Online point in time restore is a quick way to bring the database to a particular point in time
without having to restore from backups
 Rewinding the database to quickly recover from unintentional DML/DDL operations
 Rewind multiple times to determine the desired point in time in the database state. For example quickly iterate
over schema changes without having to restore multiple times

44. Online PiTR
Online PiTR operation changes the state of
the current DB
Current DB is available within seconds, even
for multi terabyte DBs
No additional storage cost as current DB is
restored to prior point in time
Multiple iterative online PiTRs are practical
Rewind has to be within the allowed rewind
period based on purchased rewind storage
Cross-region online PiTR is not supported
Offline PiTR
PiTR creates a new DB at desired point in time
from the backup of the current DB
New DB instance takes hours to restore for multi
terabyte DBs
Each restored DB is billed for its own storage
Multiple iterative offline PiTRs is time consuming
Offline PiTR has to be within the configured
backup window or from snapshots
Aurora supports cross-region PiTR
Online vs Offline Point in Time Restore (PiTR)

45. Segment snapshot Log records
Rewind Point
Segment 1
Segment 2
Segment 3
Time
Storage Segments
How does it work?
 Take periodic snapshots within each segment in parallel and store them locally
 At rewind time, each segment picks the previous local snapshot and applies the log streams to the snapshot to
produce the desired state of the DB

46. t0 t1 t2
t0 t1
t2
t3 t4
t3
t4
Rewind to t1
Rewind to t3
Invisible Invisible
How does it work? (contd.)
Logs within the log stream are made visible or invisible based on the branch within the LSN
tree to provide a consistent view for the DB
 The first rewind performed at t2 to rewind the DB to t1 makes the logs in purple color invisible
 The second rewind performed at time t4 to rewind the DB to t3 makes the logs in red and purple invisible

47. Removing Blockers

48.  Amazon Aurora PostgreSQL compatibility
now in preview
 Same underlying scale out, 3 AZ, 6 copy,
fault tolerant, self healing, expanding
database optimized storage tier
 Integrated with a PostgreSQL 9.6
compatible database
Logging + Storage
SQL
Transactions
Caching
Amazon
S3
My applications require PostgreSQL

49. T2 RI discounts
Up to 34% with a 1-year RI
Up to 57% with a 3-year RI
vCPU Mem Hourly Price
db.t2.medium 2 4 $0.082
db.r3.large 2 15.25 $0.29
db.r3.xlarge 4 30.5 $0.58
db.r3.2xlarge 8 61 $1.16
db.r3.4xlarge 16 122 $2.32
db.r3.8xlarge 32 244 $4.64
T2.Small coming in Q1. 2017
*Prices are for Virginia
An R3.Large is too expensive for my use case

50.  Amazon Aurora gives each database
instance IP firewall protection
 Aurora offers transparent encryption at rest
and SSL protection for data in transit
 Amazon VPC lets you isolate and control
network configuration and connect securely
to your IT infrastructure
 AWS Identity and Access Management
provides resource-level permission controls
*New* *New*
My databases need to meet certifications

51. MariaDB server audit plugin Aurora native audit support
We can sustain over 500K events/sec
Create event string
DDL
DML
Query
DCL
Connect
Write
to File
DDL
DML
Query
DCL
Connect
Create event string
Create event string
Create event string
Create event string
Create event string
Latch-free
queue
Write to File
Write to File
Write to File
MySQL 5.7 Aurora
Audit Off 95K 615K 6.47x
Audit On 33K 525K 15.9x
Sysbench Select-only Workload on 8xlarge Instance
Aurora Auditing

52. Lambda
S3
IAM
CloudWatch
Generate Lambda event from Aurora stored procedures.
Load data from S3, store snapshots and backups in S3.
Use IAM roles to manage database access control.
Upload systems metrics and audit logs to CloudWatch.
*NEW*
Q1
AWS ecosystem

53. Business Intelligence Data Integration Query and Monitoring
“We ran our compatibility test suites against
Amazon Aurora and everything just worked."
- Dan Jewett, VP, Product Management at Tableau
MySQL 5.6 / InnoDB compatible
 No application compatibility issues
reported in last 18 months
 MySQL ISV applications run pretty
much as is
Working on 5.7 compatibility
 Running a bit slower than expected
– hope to make it available soon
 Back ported 81 fixes from different
MySQL releases;
MySQL compatibility

54. Available now Available in Dec Available in Q1
Performance
Availability
Security
Ecosystem
PCI/DSS
HIPPA/BAA
Fast online schema change
Managed MySQL to Aurora
replication
Cross-region snapshot copy
Online Point in Time Restore
Database cloning
Zero-downtime patching
Spatial indexingLock compression
Replace spinlocks with blocking futex
Faster index build
Aurora auditing
IAM Integration
Copy-on-write volume
T2.Medium T2.Small
Cloudwatch for metrics, audit
Timeline

55. Thank you!
Steve Abraham – abrsteve@amazon.com