Infra for startup

64

Transcript of Infra for startup

Simplify Value Investing "Buy a Wonderful Company at a Fair Price"

ONLY! 6 engineers

There is somethingbehind right?

Few engineerscan do more

What Jitta believe

Great Infra = development speed

+ more creative + more growth

Many companiesuse cloud as just another

data center

Many companiesuse build and maintain

everything in-house

See Jitta ’s journey…

Git Repo

Github GitlabPaid service

Support from all 3rd parties Many tools (come with more

expenses)

(Enterprise version for your cloud)

Open source and paid versionfew tools and 3rd party

Internal CI runner need someone to maintain

(or buy support) made with Ruby on rails

Issues / Review tools / Test integration

First deploy!

Arg!!

$ git push heroku master

$ heroku logs -t

$ heroku scale web=10

Continuous Integration

JenkinsPaid service

Support from all 3rd parties Many tools (come with more expense)

(Travis has Enterprise version for your cloud)

Open sourcesmany tools java based

need someone to maintain (or buy support)

Auto deploy on green

(in that time)

Poor quality of service Many issues random happened

Network / Application crash!

So we decided to move out!

6 months later they added metrics and enhanced all systems T_T

Cloud

Both are covered all regions

Few services Various services

Same technology as Google Same technology as major startups use

Cloud

Database as a service (SQL, NoSQL, Bigtable a like) Compute Instances

Container Management Load Balancing

VPN support Firewall

Same technologies as Google use on their own products

Microservices help us!

After test greenNo deploy and build tools

like we had with heroku

Too many micro servicesto take care between

we’re growing

centralize

( made it simple to maintain )

A node.js microservices framework

jitta/spinal

Start a node called “english”

Start a node called “thai”

the same pattern

for our future

real time dashboard real time stat

rest api job queues

caching command line

features

……….

many things happened

Disk crashed Database crashed Machined crashed Everything crashed Many many crashed

How to create a perfect and stable infrastructure?

Make machines random reboot

Level up!

And now ….

Google Container EngineNo more machines need to maintain Support docker

Easy scaling containers Auto scaling the machine

Internal services discovery Auto mount disk between machines

Google Container Engine

All of these things to keep our engineers happy

more time x more creative love to push code

don’t afraid to maintain

Now we’re developingOur own deploy board

to allow engineers control release and rollback anytime

Manage Infrastructure is not a thing that

we can create a how-to

(only tear can tell you what is the right thing)

Choose the right technology at the right time that always

match with your own start-upand learn from all mistakes

We’re hiring…