+ All Categories
Home > Internet > Event driven infrastructure

Event driven infrastructure

Date post: 07-Jan-2017
Category:
Upload: shiva-narayanaswamy
View: 426 times
Download: 0 times
Share this document with a friend
33
EVENT DRIVEN INFRASTRUCTURE DevOpsDays Melbourne, 2015 Shiva N (@shivamaan)
Transcript

EVENT  DRIVEN  INFRASTRUCTURE  

DevOpsDays  Melbourne,  2015  Shiva  N  (@shivamaan)  

A  BANANA  WAREHOUSE  AT  THE  ZOO  

RUN  A  BANANA  INVENTORY  EVERY  HOUR  

GIVE  ME  5  BANANAS  

MONKEY  TOOK  5  BANANAS  

Run  a  banana  inventory  

Plot  monkeys  banana  consumption  pattern      

Analytics  on  monkey  banana  consumption  across  all  zoos    

EVENT  DRIVEN  INFRASTRUCTURE  

Event  Bus  

Event  Producers  

Event  Transport  

Event  Consumers  

- Real  time  (or  near  real  time)  - Push  notifications  - One-­‐way  ‘fire-­‐and-­‐forget’  - Immediate  action  at  event  consumer  - Informational  (“monkey  took  bananas”)  and  not  commands  (“perform  banana  inventory”)  

5  CHARACTERISTICS  

WHY?  

- Decouple  responsibilities  - Communicate  by  triggering  events  - Scale  out  components  horizontally  - Higher  responsiveness  - High  availability  &  Resiliency  

WHAT  ABOUT  INFRASTRUCTURE?  

AUTOMATED  CONFIGURATION  MANAGEMENT  

- Puppet  - Chef  - Ansible  - Salt  - Etc.  

INFRASTRUCTURE  AS  CODE  

CLOUDFORMATION  +  CUSTOM  RESOURCES  =  NO  LIMITS  

STACKSTORM  

HOMEGROWN  TOOLS/SCRIPTS  

AWS  LAMBDA  A  whole  new  world  

-  A  zero-­‐administration  compute  platform  -  Stored  procedures  for  the  cloud  -  Function-­‐as-­‐a-­‐Service  (NodeJS/Java)    -  Event  driven  from  other  services    -  …or  triggered  externally  and  even  chained  

WHAT  IS  LAMBDA?  

WHAT  EVENTS  DOES  YOUR  INFRASTRUCTURE  REACT  TO?  

SIMPLE  EVENTS  

SPIKY  LOAD  EVENTS  

-  Pre  warming  (Time)  -  Auto  scaling  (Metrics)  -  Feed  forward  loops  

(Predictive)    

COMPONENT  FAILURE  EVENTS  

- EC2  auto  healing  - SQS  design  pattern  - Service  discovery  +  Container  recycling  – Consul  – Serf  

DISASTER  RECOVERY  EVENT  

SECURITY  EVENTS  

- DDos  Attacks  - Instance  compromise  

STREAMING  EVENTS  

Cloudtrail  logs  streaming  to  Cloudwatch  

Subscribe  Kinesis  stream  to  Cloudwatch  logs  

Lambda  func8on  subscribed  to  Kinesis  watching  for  pa:erns  

SNS  no8fica8on  for  insecure  pa:erns  in  Cloudtrail  logs  

COMPLEX  EVENTS  

- Event  Observer  Pattern  - Intelligent  build  pipelines  

SENTIENT  EVENTS  ?!  

HOW  DOES  THIS  AFFECT  “THE  DEVOPS”  

OPERATIONAL  LIFECYCLE  

– Immutable  infrastructure  – No  patching  or  in  place  deployments  

SERVERLESS  ARCHITECTURES  

MONITORING  AND  METRICS  

– Centralized  logging  – Log  Analytics  – Build  metrics,  application  metrics,  customer  experience  metrics  instead  of  CPU,  memory  and  OS  metrics  

– Event  Storms    

 

IS  THIS  A  SCARY  PATH?  

THANK  YOU  


Recommended