[cloud computing day with v-forum] going global on aws
DESCRIPTION
2013년 11월 5일 양재동 엘타워에서 개최된 '브이포럼과 함께하는 클라우드 컴퓨팅 데이' 의 첫 번째 세션 발표 자료 입니다. (연사 : AWS 김일호 책임)TRANSCRIPT
You begin your Startup in a garage, 친구집, PC방, 랩실, 토즈..….. J
…and build a fantastic app
people love it!
and everyone wants to use it…
503 Service Temporarily Unavailable
The server is temporarily unable to service your request due to
maintenance downtime or capacity problems. Please try again later.
and everyone wants to use it…
Now what?!
Which company…
…grew to 14 million users in just over a year
…over 90 million active users in January 2013
…reached 150 million photos & terabytes of data
…signed up 1.0 million users in 12 hours after launching an Android app
…with only 3 engineers?
HOW?
Startup in the Cloud
“Amazon Web Services is probably the most important thing that has happened to mobile and web app developers that the press just misses. Jeff Bezos has accidentally or maybe on purpose powered a whole generation of applications.”
Steve Blank
These Korean startups …
DURABILITY
SECURITY
SCALABILITY
RELIABILITY
AFFORDABILITY
are getting the same… as these companies:
“Cloud is like a fertilizer that creates Startups”
Eric Ries
01 02 03 04
Idea MVP Profitability Scale
01 02 03 04
Idea MVP Profitability Scale
Product?
Do we have a product? Can we ship?
How to develop faster? Better? Cheaper?
Market?
Do people buy? Do they pay?
Can we scale? Can we deal with spikes?
Profit?
Can we monetize? How to keep costs down? When can we break-even?
Can we get & stay profitable?
Infrastructure in Minutes
Lower Costs of Innovation
Shorter Development
Cycles
Experiment More, Develop Cheaper, Ship Faster
Add New Dev Environment
Add New Production in US
Add 1,000 Servers Remove 1,000 servers
Number of Instances 1,000
Instance Type M3 Extra Large
Availability Zone US-West-2b
Launch
aws.amazon.com/managementconsole
aws.amazon.com/code aws.amazon.com/architecture
Code, APIs, SDKs, Architectures
Amazon Web Services
Go global in minutes to reach a global audience Fast & Easy customer acquisition leveraging AWS content distribution
Application
AWS Global
Infrastructure Components
Regions An independent collection of AWS resources in a defined geography A solid foundation for meeting location-dependent privacy and compliance requirements
Availability Zones (AZ) Designed as independent failure zones Physically separated within a typical metropolitan region
Edge Locations To deliver content to end users with lower latency A global network of edge locations supports global DNS infrastructure (Route53) and Cloud Front CDN
On a global footprint
Region US-WEST (Oregon) EU-WEST (Ireland)
ASIA PAC (Tokyo)
ASIA PAC (Singapore)
US-WEST (N. California)
SOUTH AMERICA (Sao Paulo)
US-EAST (Virginia)
GOV CLOUD
ASIA PAC (Sydney)
Availability Zone
On a global footprint
US West Region (Oregon)
Availability Zone A
Availability Zone B
Singapore
Availability Zone A
Availability Zone B
EU West Region
Availability Zone A
Availability Zone C
Availability Zone B
Each Region has multiple Availability Zones. Availability Zones are distinct locations that are engineered to be insulated from failures in other Availability Zones and provide inexpensive, low latency network
connectivity to other Availability Zones in the same Region.
Edge Locations
Dallas(2)
St.Louis Miami
Jacksonville Los Angeles (2)
Palo Alto
Seattle
Ashburn(2)
Newark New York (2)
Dublin
London(2) Amsterdam
Stockholm
Frankfurt(2) Paris(2)
Singapore(2)
Hong Kong
Tokyo
Sao Paulo
South Bend
San Jose
Osaka Milan
India(2)
Sydney
On a global footprint
…and leverages many AWS services
Technical Architecture and Best Practices
Architecture
ELB – Elas%c Load Balancing (Single or Mul- AZ)
Monitoring CloudWatch (Alarms, Custom Metrics)
Auto Scaling (Regional)
EIP – Elas%c IP Address (Regional – mul- AZ)
EC2 Instances – Elas%c Compute Cloud On-‐Demand and Reserved Instances
[ AMI – Amazon Machine Images ]
EBS – Elas%c Block Store (Single AZ, Regional Snapshots)
AWS Architecture Fundamentals
AWS Architecture Fundamentals
LOAD BALANCE from the start • uninterrupted growth and evolu-on • ready for massive scale, high reliability • stateless and redundant array
Elas%c Load Balancing, Auto Scaling, and CloudWatch
Availability Zone
Availability Zone
AWS Architecture Fundamentals
6:00 AM
8:00 AM
8:00 AM
8:00 AM
7:00 PM
7:00 PM
7:00 PM
OFFLOAD sta%c content • reduce your compute demand and costs • improve end-‐user experience • increase reliability and durability
+
Amazon S3 + Amazon CloudFront
Sta%c content
Dynamic calls
ENTIRE SITE via CloudFront • minimize client-‐server chaHer (keep it at the edge) • reduce server-‐database traffic (cache the common calls) • speed up mobile app response (persistent connec-ons)
+
CloudFront for Dynamic Content
DATABASE smarts • enhance reliability, ensure data safety • op-mal performance, minimize load • scale out and share work • cost reduc-on
Amazon RDS, Amazon Elas%Cache, and Amazon DynamoDB
Availability Zone
Availability Zone
90% of calls
Index lookups
User details
Stats / metrics
Network traversal
Mul%-‐Region Latency Based Rou%ng Route 53 now powering CloudFront, EC2 and ELB
Application instances in several AWS Regions
Users automatically routed to lowest latency Region
Applica%on
A/B TEST experiment, evolve • inexpensive idea valida-on • seamless switch over and versioning • rapid dev / test agility
AWS CloudForma%on
Lower Costs
$5.2B retail business 7,800 employees
A whole lot of servers…
2003
Every Day… AWS adds the equivalent server capacity to power Amazon when it was a global,
$5.2B enterprise in 2003
Economies of Scale Lower variable expense than companies can achieve themselves
Reduced Prices
More Customers
More AWS Usage
More Infra-structure
Economies of Scale
Lower Costs
On-Demand Pay for compute capacity by the hour with no long-term commitments For spiky workloads, or to define needs
Cost Optimization using different purchase models
Reserved Make a low, one-time payment and receive a significant discount on the hourly charge For committed utilization
Spot Bid for unused capacity, charged at a Spot Price which fluctuates based on supply and demand For time-insensitive or transient workloads
Free Tier Get Started on AWS with free usage & no commitment For POCs and getting started
70:30
Reduce Costs & Grow revenue
Pay only what for you use, with no commitment and lock in, so No Up-Front Capital Expense
Leveraging our large scale, we have reduced our prices 26 times in the last years, leading to Low Costs
AWS removes undifferentiated heavy lifting – allowing you to focus on your business and Generate Revenue
Traditional IT capacity
Elas%c capacity
Capacity
Time Your IT needs
On and Off Fast Growth
Variable peaks Predictable peaks
Elas%c capacity
Elas%c capacity
On and Off Fast Growth
Predictable peaks Variable peaks
WASTE
CUSTOMER DISSATISFACTION
Elastic cloud capacity
Traditional IT capacity
Your IT needs
Time
Capacity
Elas%c capacity
Elas%c capacity
Fast Growth On and Off
Predictable peaks Variable peaks
When you turn off your cloud resources, you actually stop paying for them
On-‐demand Instances
• Pay as you go
• Starts from $0.02/Hour
Reserved Instances
• One %me low upfront fee + Pay as you go
• $23 (1 year) $0.01/Hour
Spot Instances
• Requested Bid Price and Pay as you go
• Varying…$0.005/Hour at 9am today
Heavy U%liza%on RI
Medium U%liza%on RI
Light U%liza%on RI
1 and 3 year terms
Save more money by using appropriate purchase models
Reserved Instance Pricing
Make a low, one-time payment and receive a significant discount on the hourly charge
For committed utilization
• Light Utilization RI
• Medium Utilization RI
• High Utilization RI
• 1-year
• 3-year
2 Terms 3 Versions
Reserved Instance Pricing
Reserved Instance Pricing
UJlizaJon RI opJon Savings over On-‐Demand
<10% On-‐Demand
10% -‐ 40% Light U%liza%on RI Up to 56%
40% -‐ 75% Medium U%liza%on RI Up to 66%
>75% Heavy U%liza%on RI Up to 71%
Reserved Instance Pricing
Save more money by using Spot Instances
Reserved Hourly Price > Spot Price < On-‐Demand Price
Spot Use cases Use Case Types of ApplicaJons
Batch Processing Generic background processing (scale out compu%ng)
Hadoop Hadoop/MapReduce processing type jobs (e.g. Search, Big Data, etc.)
ScienJfic CompuJng Scien%fic trials/simula%ons/analysis in chemistry, physics, and biology
Video and Image Processing/Rendering
Transform videos into specific formats
TesJng Provide tes%ng of sokware, web sites, etc
Web/Data Crawling Analyzing data and processing it
Financial Hedgefund analy%cs, energy trading, etc
HPC U%lize HPC servers to do embarrassingly parallel jobs
Cheap Compute Backend servers for Facebook games
Typical Spot Bidding Strategies
0%
2%
4%
6%
8%
10%
12%
14%
16%
18%
20%
Perc
enta
ge o
f the
Dis
trib
utio
n
Bid Price as Percentage of the On-Demand Price
Bid Distribution
Op%mizing Video Transcoding Workloads
" Free Offering • Op%mize for reducing cost • Acceptable Delay Limits
ImplementaJon
• Set Persistent Requests • Use on-‐demand Instances, if delay
Maximum Bid Price < On-‐demand Rate Get your set reduced price for your workload
" Premium Offering § Op%mized for Faster response %mes § No Delays
ImplementaJon
§ Invest in RIs § Use on-‐demand for Elas%city Maximum Bid Price >= On-‐demand Rate Get Instant Capacity for higher price
ConJnuous opJmizaJon in your architecture results in
recurring savings as early as your next month’s bill
Implement CloudWatch
Metrics and Actions
Use the built-in metrics such as CPU Utilization or create
Custom Metrics
Set thresholds with notifications to SNS or HTTP calls to react
Study Trusted Advisor Reports Underutilization
All AWS Accounts are notified through the designated email
Use an email alias to share the notifications and reports
Right Size Right Type
Auto Scaling
Re-visit and Re-evaluate design decisions based on performance
Consider variations – more small
instances vs. fewer large instances
Vary reaction times and thresholds
for Auto Scaling actions to minimize “churn” in capacity
Ac%onable Sugges%ons
Questions?