Scaling your Application with AWS Relational Databases I AWS Dev Day 2018

© 2018, Amazon Web Services, Inc. or its Affiliates. All rights reserved.
Scaling your application with AWS
relational databases
Darin Briskman | April 2018
briskman@amazon.com

What is Amazon RDS?
• Managed relational database service in the
AWS cloud
• Multi-engine support: Amazon Aurora,
MySQL, MariaDB, PostgreSQL, Oracle, SQL
Server
• Automated provisioning, patching, scaling,
replicas, backup/restore
• Easily scales to handle growth
• High availability with Multi-AZ and Amazon
Aurora
Amazon RDS

Why use Amazon RDS?
• Lower TCO because we manage “the muck”
• Get more leverage from your teams
• Focus on the things that differentiate you
• Built-in high availability and cross-region
replication across multiple data centers
• Even a small startup can leverage multiple data
centers to design highly available apps with over
99.95% availability

Commercial Open Source Cloud Native
Which RDS engine should I use?
Amazon EBS-based Storage Aurora Storage System
MySQL Compatible
PostgreSQL Compatible

T2 Family
• Burstable instances
• 1 vCPU/1 GB RAM > 8
vCPU 32 GB RAM
• Moderate networking
performance
• Good for smaller or variable
workloads
• Monitor CPU credit metrics
in Amazon CloudWatch
• T2.micro is eligible for free
tier
M3/M4 Family
• General-purpose instances
• 2 vCPU/8 GiB RAM > 64
vCPU 256 GiB RAM
• High-performance
networking
• Good for running CPU
intensive workloads (e.g.,
WordPress)
R3/R4 Family
• Memory-optimized instances
• 2 vCPU/16 GiB RAM > 64
vCPU 488 GiB RAM
• High-performance networking
• Good for query-intensive
workloads or high connection
counts
Which instance type should I choose?

General purpose (GP2)
• SSD storage
• Maximum of 16 TB!
• Leverages Amazon EBS
Elastic Volumes
• IOPS determined by
volume size
• Minimum of 100 IOPS
(below 33.33GiB)
• Bursts to 3,000 IOPS
(applicable below 1.3 TB)
• Baseline of 10,000 IOPS
(at 3.3 TB and above)
• Affordable performance
Provisioned IOPS (IO1)
• SSD storage
• Maximum of 16 TB!
• Leverages Amazon EBS
Elastic Volumes
• Maximum of 40 K IOPS
(20 K on SQL Server)
• Delivers within 10% of the
IOPS performance 99.9%
of the time
• High performance and
consistency
Magnetic
• Magnetic storage
• Maximum of 1 TB
• Supported for legacy
databases
Which storage type should I choose?

How do I decide between GP2 and IO1?
W h y a m I n o t s e e i n g 3 0 K I O P S ?
• GP2 is a great choice, but be aware of burst credits on volumes < 1 TB
• Hitting credit-depletion results in IOPS drop—latency and queue depth metrics
will spike until credits are replenished
• Monitor BurstBalance to see percent of burst-bucket I/O credits available
• Monitor read/write IOPS to see if average IOPS is greater than the baseline
• Think of GP2 burst rate and PIOPS
stated rate as maximum I/O rates

How do I scale my database instance?
W i l l t h e r e b e d o w n t i m e ?
• Scale compute/memory vertically up or down
• Handle higher load to grow over time
• Lower usage to control costs
• New host is attached to existing storage with minimal
downtime
• Scale up Amazon EBS storage (now up to 16 TB!)
• Amazon EBS engines now support Elastic Volumes for fast
scaling (now including SQL Server)
• No downtime for storage scaling
• Initial scaling operation may take longer, because storage is
reconfigured on older instances
• Can re-provision IOPS on the fly

How do I ensure database high
availability?Multi-AZ provides enterprise-grade fault-tolerance
solution for production
databases
Automatic failover
Synchronous replication
Inexpensive and enabled with one click

What happens during a Multi-AZ failover?
H o w l o n g d o e s i t t a k e ?
• Each host manages set of Amazon
EBS volumes with a full copy of
the data
• Instances are monitored by an
external observer to maintain
consensus over quorum
• Failover initiated by automation or
through the Amazon RDS API
• Redirection to the new primary
instance is provided through DNS
DNS
Primary Standby
EC2
#1
EC2
#2
Availability Zone #1 Availability Zone #2
APP
EBS
#1
EBS
#2
PrimarySecondar
y

Why would I use Read Replicas?
• Relieve pressure on your source
database with additional read capacity
• Bring data close to your applications in
different regions
• Promote a Read Replica to a master for
faster recovery in the event of disaster
• Upgrade a Read Replica to a new
engine version
• Supported for MySQL, MariaDB, and
PostgreSQL

© 2018, Amazon Web Services, Inc. or its Affiliates. All rights reserved.© 2018, Amazon Web Services, Inc. or its Affiliates. All rights reserved.
• Synchronous replication—highly
durable
• Only primary instance is active at any
point in time
• Backups can be taken from secondary
• Always in two Availability Zones within
a Region
• Database engine version upgrades
happen on primary
• Automatic failover when a problem is
detected
• Asynchronous replication—highly
scalable
• All replicas are active and can be used
for read scaling
• No backups configured by default
• Can be within an Availability Zone,
cross-AZ, or cross-region
• Database engine version upgrades
independently from source instance
• Can be manually promoted to a
standalone database
R e a d R e p l i c a sM u l t i - A Z
When should I use Multi-AZ?
When should I use Read Replicas?

How does Amazon RDS manage backups?
• Two options – automated backups
and manual snapshots
• Amazon RDS backups leverage
Amazon EBS snapshots stored in
Amazon S3
• Transaction logs are stored every
5 minutes in Amazon S3 to support
point-in-time recovery (PITR)
• No performance penalty for
backups
• Snapshots can be copied across
regions or shared with other
accounts
Availability Zone #2
Availability Zone #1
Region #1 Region #2
Transaction Logs
DB
App
RDS
Host
RDS
Host
EBS
Vol
EBS
Vol
EBS Snapshot
Amazon
S3

• Specify backup retention window per
instance (7-day default)
• Kept until outside of window (35-day
maximum) or instance is deleted
• Supports PITR
• Good for disaster recovery
• Manually created through AWS
console, AWS CLI, or Amazon RDS
API
• Kept until you delete them
• Restores to saved snapshot
• Use for checkpoint before making
large changes, non-production/test
environments, final copy before
deleting a database
M a n u a l s n a p s h o t sA u t o m a t e d b a c k u p s
When should I use automated backups?
When should I use snapshots?

How do I restore a backup?
W h y d o e s i t t a k e s o l o n g ?
• Restoring creates an entirely new database instance
• Define the instance configuration just like a
new instance
• Will get the default parameter, security, and
option groups
• New volumes are hydrated from Amazon S3
• While the volume is usable immediately, full
performance requires the volume to warm up
until fully instantiated
• Migrate to a DB instance class with high I/O
capacity
• Maximize I/O during restore process

How do I secure my Amazon RDS
database? • Amazon RDS is designed to be secure by default
• Network isolation with Amazon Virtual Private
Cloud (Amazon VPC)
• AWS Identity and Access Management (IAM)-
based resource-level permission controls
• Encryption at rest using AWS KMS (all engines) or
Oracle/Microsoft TDE
• Use SSL protection for data in transit

What does Amazon VPC provide?
• Places your instance in a private subnet,
making it secure from public routes on the
Internet
• Database instance IP firewall protection lets
you securely control network configuration
• Turn off Public Accessibility in DB instance
settings to restrict access outside Amazon VPC
• Use ClassicLink to network with non-VPC
resources
AWS Direct
Connect
VPN
connection
VPC
peering
Internet
gateway
Routing
rules
Availability Zone
AWS Region
10.1.0.0/16
10.1.1.0/24

How do I grant access to my database?
• Use IAM to control who can perform actions on RDS resources
• Do not use AWS root credentials to manage Amazon RDS resources—you should
create an IAM user for everyone, including yourself
• Can use AWS Multi-Factor Authentication (MFA) to provide extra level of
protection
DBA and Ops
Amazon RDS
Controlled with IAM
Users and DBAApplications
Your database
Controlled with database grants

How do I encrypt my database?
• Use AWS KMS-based encryption in the AWS
console
• No performance penalty for encrypting data
• Centralized access and audit of key activity
• Best practices
• Encryption cannot be removed from DB
instances
• If source is encrypted, Read Replicas must
be encrypted
• Add encryption to an unencrypted DB
instance by encrypting a snapshot copy
Data keys
Amazon
RDS
instance 3
Amazon
RDS
instance 2
Amazon
RDS
instance 1
Customer master key(s)

How do I monitor my Amazon RDS
database? Amazon CloudWatch metrics
• CPU/Storage/Memory
• Swap usage
• I/O (read and write)
• Latency (read and write)
• Throughput (read and write)
• Replica lag
Amazon CloudWatch Alarms
• Similar to on-premises monitoring tools
Enhanced monitoring for Amazon RDS
• Access to over 50 CPU, memory, file
system, and disk I/O metrics
• Low as 1-second intervals
Integration with third-party monitoring tools

• Introducing Amazon RDS Performance
Insights
• Measures DB Load: Average Active
Sessions (AAS)
• Identifies database bottlenecks (Top
SQL)
• Easy
• Powerful
• Identifies source of bottlenecks
• Enables problem discovery
• Adjustable time frame
• Hour, day, week, and longer
• Coming soon for Amazon EBS-based
Amazon RDS engines
How do I improve database performance?

Can I know when service events happen?
• Amazon RDS uses Amazon SNS to
receive notification when an event occurs
• Notifications can be in any form supported
by Amazon SNS (email, a text message, or
a call to an HTTP endpoint)
• Six different source types (DB instance,
DB parameter group, DB security group,
DB snapshot, DB cluster, DB cluster
snapshot)
• 17 different event categories (availability,
backup, deletion, configuration change,
etc.)

• Any maintenance that causes downtime (typically only a few times per year) will be scheduled
in your maintenance window
• Operating system or Amazon RDS software patches are usually performed without restarting
databases
• Database engine upgrades require downtime
• Minor version upgrades—automatic or manually applied
• Major version upgrades—manually applied
• Version deprecations—three- to six-month notification before scheduled upgrades
• View upcoming maintenance
events in your AWS Personal
Health Dashboard
How do you maintain my database?
H o w c a n I p l a n f o r i t ?

How am I charged for Amazon RDS?
• Database instance (instance hours)
• Combination of Region + instance type + database engine
+ license (optional)
• Database storage (GB-mo)
• Can be either provisioned (Amazon EBS) or consumed
(Amazon Aurora)
• Provisioned IOPS (IOPS-Mo) for IO1 storage type
• Database I/O requests (IOs) for Amazon Aurora and
Amazon EBS magnetic-storage types
• Backup storage (GB-mo)
• Size of backups and snapshots stored in Amazon S3
• No charge for backup storage up to 100% of total
database storage
• Data transfer (GB-mo)
• Uses AWS regional data-transfer pricing
Amazon RDS Free Tier

How can I save money on my database?
• Amazon RDS Reserved Instances (RIs) provide a discount over on-demand prices
• Region, instance family, and engine of on-demand usage must match to apply
benefit
• Amazon RDS RIs offer size flexibility for open-source and Oracle BYOL engines
• By default, RIs are shared between all
accounts in consolidated billing
• Use the RI utilization report to determine
how your RIs are being used
• Support for RI coverage report
coming soon

Can I stop my database when it’s not in
use?
• Stop and start a running database instance from the
console or AWS CLI
• Available for single-AZ DB instances
• While instance is stopped, you only pay for storage
• Backup retention window is maintained while
stopped
• Instances are restarted after 7 days
• Pending maintenance operations are applied
• Instances can be stopped again if desired

What is Amazon Aurora ……
D atabas e r eimagined for the c loud
 Speed and availability of high-end commercial databases
 Simplicity and cost-effectiveness of open source databases
 Drop-in compatibility with MySQL and PostgreSQL
 Simple pay as you go pricing
Delivered as a managed service

Re-imagining relational database
Automate administrative tasks – fully managed service
1
2
3
Scale-out and distributed design
Service-oriented architecture leveraging AWS services

Scale-out, distributed architecture
Master Replica
Availability
Zone 1
Shared storage volume
Availability
Zone 2
Availability
Zone 3
Storage nodes with SSDs
 Purpose-built log-structured distributed
storage system designed for databases
 Storage volume is striped across
hundreds of storage nodes distributed
over 3 different availability zones
 Six copies of data, two copies in each
availability zone to protect against
AZ+1 failures
 Plan to apply same principles to other
layers of the stack
SQL
Transactions
Caching
SQL
Transactions
Caching
SQL
Transactions
Caching

Amazon Aurora is fast …
5x more throughput than MySQL
3x more throughput than PostgreSQL

WRITE PERFORMANCE READ PERFORMANCE
MySQL SysBench results
R3.8XL: 32 cores / 244 GB RAM
5X faster than RDS MySQL 5.6 & 5.7
Five times higher throughput than stock MySQL
based on industry standard benchmarks.
0
25,000
50,000
75,000
100,000
125,000
150,000
0
100,000
200,000
300,000
400,000
500,000
600,000
700,000
Aurora MySQL 5.6 MySQL 5.7

Do fewer I/Os
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
How did we achieve this?
DATABASES ARE ALL ABOUT I/O
NETWORK-ATTACHED STORAGE IS ALL ABOUT PACKETS/SECOND
HIGH-THROUGHPUT PROCESSING IS ALL ABOUT CONTEXT SWITCHES

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
AZ 1 AZ 3
Primary
Instance
Amazon S3
AZ 2
Replica
Instance
ASYNC
4/6 QUORUM
DISTRIBUT
ED WRITES
Replica
Instance
AMAZON AURORA
780K transactions
7,388K I/Os per million txns (excludes mirroring, standby)
Average 7.4 I/Os per transaction
MySQL I/O profile for 30 min Sysbench run
27,378K transactions 35X MORE
0.95 I/Os per transaction (6X amplification) 7.7X LESS
Aurora IO profile for 30 min Sysbench run
Aurora I/O profile

Scan
Delete
Aurora lock management
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
Needed to support many concurrent sessions, high update throughput
 Multiple scanners in individual lock chains
 Lock-free deadlock detection

New performance enhancements
► Smart selector
► Logical read ahead
► Read views
► Hash joins
► Parallel query
Read performance
Write performance
Meta-data access
► NUMA aware scheduler
► Latch-free lock manager
► Instant schema update
► B-Tree concurrency
► Catalog concurrency
► Faster index build

What about availability?
“Performance only matters if your database is up”

Six copies across three availability zones
4 out 6 write quorum; 3 out of 6 read quorum
Peer-to-peer replication for repairs
Volume striped across hundreds of storage nodes
SQL
Transaction
AZ 1 AZ 2 AZ 3
Caching
SQL
Transaction
AZ 1 AZ 2 AZ 3
Caching
Read and write availabilityRead availability
6-way replicated storage
Sur vives c atas tr ophic failur es

Amazon Aurora Recovers Up to 97%Faster
3 GiB Redo
Recovered in 19
seconds
10 GiB Redo
Recovered in 50
seconds
30 GiB Redo
Recovered in 123
seconds
0
20
40
60
80
100
120
140
160
0 20,000 40,000 60,000 80,000 100,000 120,000 140,000
RecoveryTimeinSeconds(lessisbetter)
Writes / Second (more is better)
RECOVERY TIME FROM CRASH UNDER LOAD
Bubble size represents redo log, which must be recovered
As PostgreSQL
throughput goes up,
so does log size and
crash recovery time
Amazon Aurora has no redo.
Recovered in 3 seconds while
maintaining significantly greater
throughput.

Up to 15 promotable read replicas
Master
Read
Replica
Read
Replica
Read
Replica
Shared distributed storage volume
Reader end-point
► Up to 15 promotable read replicas across multiple availability zones
► Re-do log based replication leads to low replica lag – typically < 10ms
► Reader end-point with load balancing and auto-scaling

Cross-region read replicas
Fas ter dis as ter r ec over y and enhanc ed data loc ality
Promote read-replica to a master
for faster recovery in the event of
disaster
Bring data close to your
customer’s applications in
different regions
Promote to a master for easy
migration

Availability is about more than HW failures
You also incur availability disruptions when you
1. Patch your database software – Zero Down Time Patch
2. Perform large scale database reorganizations – Fast Cloning
3. DBA errors requiring database restores – Online Point-in-time Restore

Database backtrack
Backtrack brings the database to a point in time without requiring restore from backups
• Backtracking from an unintentional DML or DDL operation
• Backtrack is not destructive. You can backtrack multiple times to find the right point in time
t0 t1 t2
t0 t1
t2
t3 t4
t3
t4
Rewind to t1
Rewind to t3
Invisible Invisible

Performance Insights
Dashboard showing
Load on Database
• Easy
• Powerful
Identifies source of bottlenecks
• Top SQL
Adjustable time frame
• Hour, day, week , month
• Up to 35 days of data
Max CPU

Fast database cloning
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

DevTest environments
Master Replica
Clone
Master Replica
Production Account Development Account
Cross-account snapshot
Test on prod data
• Create a clone of multi-terabyte DB clusters in minutes. Use for diagnosis and pre-
prod testing
• Restore a production DB cluster in a development accounts for development, staging,
or partnering with other teams

Thank you!
… and thanks to our sponsors!

Scaling your Application with AWS Relational Databases I AWS Dev Day 2018

More Related Content

Similar to Scaling your Application with AWS Relational Databases I AWS Dev Day 2018 (20)

More from AWS Germany (20)

Recently uploaded (20)

Scaling your Application with AWS Relational Databases I AWS Dev Day 2018