20. ?
No upfront
capital
Only pay for what you use No need to
guess capacity
Agility, speed &
innovation
Remove
undifferentiated
heavy lifting
Go global
in minutes
21.
22. SIMPLICITY
“A complex system that works is invariably found to have
evolved from a simple system that worked. A complex system
designed from scratch never works and cannot be patched
up to make it work. You have to start over with a working
simple system. ”
Gall’s law
27. Private Docker
repositories
Mount persistent
volumes at launch
Across a cluster
of EC2 instances
Launch and
terminate
Docker containers
Amazon EC2 Container Service
A Fully Managed Service for Docker Containers
31. The Best Place to Run Containers in Production
EC2 CONTAINER
SERVICE
EC2 CONTAINER
REGISTRY
32. The Best Place to Run Containers in Production
EC2 CONTAINER
SERVICE
EC2 CONTAINER
REGISTRY
AZ-AW ARE
SERVICE SCHEDULER
33. The Best Place to Run Containers in Production
EC2 CONTAINER
SERVICE
EC2 CONTAINER
REGISTRY
AZ-AW ARE
SERVICE SCHEDULER
DOCKER INTEGRATION
WITH COMPOSE
34. The Best Place to Run Containers in Production
EC2 CONTAINER
SERVICE
EC2 CONTAINER
REGISTRY
AZ-AW ARE
SERVICE SCHEDULER
DOCKER INTEGRATION
WITH COMPOSE
ECS CLI
44. DynamoDB
Create User
Update profile
Send message
Latest messages
Back end
API Gateway
Javascript
& HTML5 app
Mobile app
Front end
Server-less Backend Applications
45. AWS Lambda in Production
Threat intelligence and analytics
Ad data analytics and routing Mobile app analytics Image content filtering
Real-time video and bidding
News content processing
News media processing Game metrics analytics
57. 3 Steps to App Nirvana With AWS Mobile Hub
1. CHOOSE & CONFIGURE FUNCTIONALITY
58. 3 Steps to App Nirvana With AWS Mobile Hub
1. CHOOSE & CONFIGURE FUNCTIONALITY
2. DOWNLOAD SOURCE
59. 3 Steps to App Nirvana With AWS Mobile Hub
1. CHOOSE & CONFIGURE FUNCTIONALITY
2. DOWNLOAD SOURCE
3. TEST & MONITOR
60. Lots of devices can be solved by trying to test on all of them, but
usually more likely to just use a subset; made worse by the fact
that new devices are appearing on the market all the time.
61. Automatically test on a large
selection of device types
View detailed reports,
logs and screenshots
Configure location, language
and application data
Integrate with existing
development workflow
AWS Device Farm
Test your app on real
smartphones and tablets
in the AWS Cloud
85. Sales lead ranking
Customer adoption models
Counterfeit goods detection
Item classification
Display ads
Customer support
Demand estimation
Search intent
The Spark For Hundreds of New Machine
Learning Applications
86. Easily create machine learning models
Visualize and optimize models
Put models into production in seconds
Battle-hardened technology
Use data from S3, RDS, and Redshift
Amazon Machine Learning
89. Build model
Validate & optimize
Make predictions
1
2
3
Batch predictions
Asynchronous predictions
with trained model
Real-time predictions
Synchronous, low latency,
high throughput
Mount API end-point with a
single click
Redmart – eCommerce, Retail, Online Grocery Store
GrabTaxi – Taxi booking app, ASEAN Unicorn
Wego – Travel Search engine, searches from over 700 travel sites
Eyeota – Data marketplace, audience targeting company
Ninjavan – flexible logistics for eCommerce, (the self-proclaimed AWS for logistics)
Gumi – Mobile gaming, Global audience, Run Brave Frontier (30mn users) on AWS
Razer – Gaming hardware and software, Wearables
PropertyGuru – Property search portal
Matchmove – Mobile payments, gaming platform
Mobilewalla - mobile ad platform
Temasys – Web RTC platform
ViSenze – Image search for ecommerce
Pie – Enterprise communications platform
Trax – Image Recognition
PocketMath – Real-time Bidding platform for Mobile Ads. Signed an EDP this year.
Zalora – Rocket Internet eCommerce, Retail, Dev/Test environment
Galaxy is a commercial meteor.js application platform
Galaxy needed to be built to be accessible to developers without sophisticated devops backgrounds
Galaxy needed to be scalable multi-tenant: 100k users, 1MM processes, 100M sessions
Wanted to use containers for isolated user processes and fast spin-up
Chose ECS because it was integrated with other parts of the AWS stack.
Argument of no software to install and operate was compelling
It supported multiple availability zones
ECS got Galaxy faster to market than Kubernetes or Mesos
ECS also provided the management API to write customer schedulers for AZ spread scheduling, rate limiting, and application based health checks.
Galaxy provides the application state management and runs as a management overlay atop ECS through the ECS APIs. Through its custom scheduler, Galaxy schedules user applications atop containers running across ECS clusters
A micro-services based architecture
Backend batch processing systems are encapsulated in containers and Amazon ECS
No extra time is spent installing software and maintaining a cluster
INTRODUCTION: Welcome Jason Fischl (FISH-EL), VP of Engineering, Remind.
To talk more about the benefits of containers and using Amazon ECS, please
FireEye (Threat intelligence and analytics)
FireEye designed a search architecture for their threat intelligence and analytics products that uses AWS Lambda functions to scan data stored in S3 and stream the results back to the query GUI.
Benefits: Dramatically decreased costs from previous implementation which required an Elasticsearch cluster. There was minimal impact on performance and customer experience.
AdRoll (Ad retargeting platform)
AdRoll stores over 300TB of new compressed data in S3 every month for different teams to access and analyze. Lambda allows AdRoll to notify and route data within minutes to each team every time new data is added to S3.
Benefits: Lambda allows AdRoll to have a fast and automated way to deliver new data to different teams (e.g., machine learning, analytics, etc) .
Localytics (App analytics and marketing platform)
Localytics processes in real-time billions of data points monthly. Lambda is used to process historical and live data stored in S3 and streamed from Kinesis
Benefits: Eliminates the need to provision and manage infrastructure to run each microservice. Lambda automatically scales with the workload, processing billions of data points monthly. Lastly, it speeds the time to market for new features
Periscope (Real-time video broadcasting)
When users broadcast on Periscope, the video is broken into 3 second chunks for Lambda to process and identify if there is pornographic content.
Benefits: <Checking>
VidRoll (Video advertising platform)
VidRoll initially used AWS Elastic Beanstalk to power the business logic and dynamic configuration for real-time video ad bidding across multiple exchanges. Amazon EC2 was also used for real-time video ad transcoding. VidRoll switched to Lambda to power the business logic and video transcoding.
Benefits: Eliminates the need for developers to understand or worry about infrastructure. VidRoll has grown revenue by 10x without hiring additional technical resources to manage the volume because of the serverless architecture enabled by Lambda – the savings are passed on to their customers
Thomson Reuters (Media corporation)
Thomson Reuters cleanses text and images from various news sources and dynamically generates content for distribution to publishers using Lambda. Lambda takes master object files (e.g., photo or text) and generates different pictures sizes and document formats (XML, JSON, binary)
Benefits: Significant simplification of workflow and reduced time-to-market for new content types and formats
Associated Press (Nonprofit news agency)
AP initially used Amazon EC2 instances to scrub and merge hundreds of smaller documents into a single document. AP switched to Lambda to power the document merge process.
Benefits: Serverless and pay as you go nature allowed for simplicity and lowered costs from not having to keep around idle servers.
MLBAM (Interactive sports media platform)
Uses Lambda to combine their baseball statistics feed with Twitter, videos, and images to create a live multimedia feed for their customers.
Benefits: Increases the speed and reduces the cost of real-time stream processing.
Global Expansion: Asia-Pacific/Tokyo
ASK = Alexa Skills Kit
Device and platform proliferation is Quickly overwhelming
One approach? Focus on a smaller number of devices and platforms, but this limits audience and isn’t really a good approach?
Or go out and buy multiple devices and run every platform version, and try to test your app across them? Not ideal either?
Especially as new models with new platform versions arrive all the time
Counterfeit goods: identify risk of fake products based on counterfeit serial numbers.
Customer adoption models: targeting promotional campaigns more specifically based on likelihood of adoption for a customer: Amazon Student, Dash, Mom.
Item classification: eliminate need for human classification of new grocery items, into categories and subcategories
Sales lead ranking: Prioritize businesses who are likely to be successful deals on Amazon Local.
Search intent: detect intent of query and route to the appropriate category
Global demand estimation: estimate ASIN demand in new regions
Customer support: improve quality and efficiency of customer support experience through social media processing, routing of new information/discovery calls to correct department, detect ‘where’s my stuff’ calls and route appropriately.
Display ads: increase the click through rate of displayed ads by selecting the creative with the best performance for a customer segment: better experience for the customer as more relevant ads.
Any Startup = Self Starter Package
All others = Portfolio Package
VC = higher than 1.5M raised = NO ACTIVATE