2. What to Expect from this Session
Old-Days Challenges
Solutions with ECS
What’s the Benefits
ECS v.s. Serverless
Demo
3. A Bit about Me
• IT Pro and developer for the past 15
years
• Chief Architect of Astra
Cloud(miiicasa.com) from Taiwan
• Experienced in IoT cloud platform
across multiple AWS regions
• AWS All-5 Certificates holder
7. What We Expect
resource consolidation and isolation
fully-managed immutable infrastructure
self-healing and auto scaling(both infra and service)
designed for micro-services
spot instances with maximal cost optimization
8. EC2 Contaner Services
a highly scalable, fast, container management service that makes it easy to run, stop, and manage
Docker containers on a cluster of Amazon EC2 instances
14. Benefits
fully-managed instances by ASG
fully-managed containers and services by ECS
self-healing and auto-scaling out-of-the-box
EC2 Instance Auto Scaling and ECS Service Auto Scaling
focus on your micro-services, instead of infrastructure
cost optimization with spot, <$10 per m3.medium per month
16. ECS over Serverless
cpu-bound computing
require higher memory(>1.5GB)
long running processes(>5min or even more)
invocation-intensive functions or libraries
microservice endpoints for internal only
- OpenResty or Express may be preferable
preferable languages like golang, Erlang, Ruby, php.etc.
use cases like MQTT, websockets,etc.
higher concurrent execution(>1000)
- lambda soft limit = 100
- with even more restrictions inside VPC
17. Serverless over ECS
short-living with limited CPU/memory usage
not invocation-intensive
microservice endpoints mostly for external
- with API Gateway and Custom Authorizer
preferable in java8, nodejs4.3 or python2.7
lower concurrent execution(<100)
you want to focus on the code, nothing else
“glue” AWS services
you need quick implementation, PoC or Demo
18. ECS with Serverless
ECS for queue/workers scenario with ASG and CloudWatch
- serverless enqueue, ECS dequeue
Async ecs run-task from Lambda or service proxied by API
Gateway
CloudWatch Events + Lambda + ECS(cron-like batch job)
20. ChatOps with IM
auto-scaling and self-healing - design for failure
Ops on-the-go (anywhere, anytime)
create “sns2im” SNS topic and subscribe it with Lambda
alarm notifications with pre-defined action buttons
100% serverless architecture(API Gateway and Lambda)