gengwg / redisconf19

Day 1

microservices architecture workshop

two pizza teams
samll autonojous teams

deliver software rapidly reliably.

• mtaintanabilty
• test
• deploy

monolithic deliver software rapidly reliably deliery imposible.

monolithic hell

app -- > a set of services:

• maintainable and testable
• loosely coupled
• independently deployable
• organized around business capabilities

database per service  
schema change withoiut cordination with other teams
looser coupling
eliminate runtime coupling, e.g. locks.
shared db but private tables
separate db

expriment safely and evolve the tech stack.

complexity

• dev: ipc, partial failure, distributed data
• testing: integration, end2end
• deployment: k8s

devops/continuous delivery
product-centered teams

pattern: reusable solution to a problem described objectively.

https://microservices.io

trasactions

transactions and queries in microservice using asynchronous messaging.

loosely coupled services

online store api
customer management
order management

sagas

series of local trasactions

syncrhonous rest api call initiates asynchronous saga.

event handlers
indicrect way to make sth happen

saga orchestrator

event sourcing
event centric approcach

k8s and redis

'making reliable scalable, agile distributed system a cs101 exercise.'

self-healing keep workloads at desired state.

run containers in multiple nodes

networking.

• communicate directly via overlay network.
• service lookup by name via internal dns servervice.
• cloud lb to route traffic.
• dispatch incoming traffic to relevant backend (ingress)

management

customre resource definition
+
customer controller 
= operator

stateful set controller

Day 2

Gerneral Session

io-thread enabled

acs. security. operational safety. bitmap at runtime.

redis 6

Google cloud ceo

oracle 22 years.

• unified platform
• different access mothold

learn 18 databases?

data powered. managed db. tune perfo. serverless platform

higher level, higher contribution. upper level of abstraction.

games

game industry use redis or die?

Redis replication

microservices at credit karma

php monolithic --> 900 micro services.

client side vs service mesh

consol for service discovery

• automate first
• think and plan ahead.
• maintainance overhead will grow exponentially.

orchestration frameworks

• k8s
• swarm
• nomad

helm

observability

env parity

unsolved

• shared staging env
• mock services
• load dependencies automatically
  • dependeny tree

tool integration is hard

independent development.

one db per service.

large logical mysql cluster on bm.

85% scala node picking up

testing
team owns
how much tests?
central pipeline compliance
dev kube

config
consol k/v store
deployed with services

secrets in vault

Turing complete

integer optimization problem

factor graph