The Cloud: What’s Different?
Fajri Abdillah
[email protected]
Every business required a system that can satisfied the customer
Happy customer == More money
Traditional (owned and operated) IT Infrastructure
Business Point of View
1. Higher investment in Hardware
2. Too much server is costly
3. Too few server is making Our Customer unhappy
4. Higher Total Cost Of Ownership (TCO)
Technical Point of View
1. More time for capacity planning
2. Harder to keep up with Customer Data
(Harder to Scale)
3. More "Server" team
4. Slower in delivering software
Cloud
Business Point of View
1. No investment in Hardware
(Pay as you go, pay per use)
2. Too much server is less costly
(Add more server when needed)
3. No more too few server
(Autoscaling to the rescue)
4. Lower Total Cost Of Ownership (TCO Calculator)
Technical Point of View
1. Less time for capacity planning
2. Easier to keep up with Customer Data
(Easier to scale)
3. Less "Server" team
4. Faster in delivering software
Airbnb in 4 years (2010 - 2013)
2 Billion rows stored in RDS
From 24 EC2 instances to 1000 EC2 instances
From 300GB of Photos to 50TB of Photos in S3
5 person operations team
Serverless Architectures for Web Applications
What is serverless? [1][2][3]
Native architecture of the cloud that enables you to focus on business not infrastructure
Build and run applications and services without thinking about servers
No need to thinking about : server provisioning, patching, maintenance and capacity provisioning
Why use serverless?
Build modern applications with increased agility and lower TCO
Built-in high availability and fault tolerance, security is included
Developers can focus on code not infrastructure, more vacation time
Who use serverless?
Better Australian Census Website (Original)
Cost $9,000,000 with $400,000 on load testing alone
Load tested at a 150% of expected usage levels
"It crashed pretty much straight away" - Austin Wilshire
Better Australian Census Website (Serverless)
Cost $500 and under 24 hours of total work
Built by 2 University student that totally new to AWS
Tested with goad.io to achieve up to 100k requests per second
"cool: QUT Students built an Australian Census system for $500 that does not #CensusFail"
Werner Vogels [1] [2]
Mobile Analytics as a Service
11 million messages in SQS
How big is 10 million messages?
100 KB * 10 = 1,000 KB = 1 MB
100 KB * 100 = 10,000 KB = 10 MB
100 KB * 1,000 = 100,000 KB = 100 MB
100 KB * 10,000 = 1,000,000 KB = 1,000 MB = 1 GB
100 KB * 100,000 = 10,000,000 KB = 10,000 MB = 10 GB
100 KB * 1,000,000 = 100,000,000 KB
= 100,000 MB = 100 GB
100 KB * 10,000,000 = 1,000,000,000 KB
= 1,000,000 MB = 1,000 GB
= 1 TB
Messages in Flight
Processing ~2000 Messages / second
100 KB * 2,000 = 200,000 KB = 200 MB / second
AWS SQS
Amazon Simple Queue Service (Amazon SQS) offers a secure, durable, and available hosted queue that lets you integrate and decouple distributed software systems and components.
Redis based Queue
SQS based Queue
SQS - Power Features
Serverless - No need to think about server, battle tested since 2006
Unlimited Throughput - Standard queues support a nearly unlimited number of transactions per second
Security - Control who can send messages to and receive messages from an Amazon SQS queue
Durability - Make use of dead letter queue, no data is missing, up to 14 days data retention.
Event Source - Lambda can process message in SQS
SQS - Limitations
Single Consumer - once a consumer processed a message then message is removed
Tricky pricing - Billed per transaction, chunked per 64 KB, but still cheap
Limited payload size - 256 KB, can use S3 for bigger payload, up to 2 GB, but more expensive
Delay queue - Maximum 15 minutes
Maximum batch message - 10 messages or 256 KB
SQS - Pricing
| Price per 1 Million Requests after Free Tier (Monthly) | |
|---|---|
| Standard Queue | $0.40 ($0.00000040 per request) |
| FIFO Queue | $0.50 ($0.00000050 per request) |