This document discusses the opportunities offered by Google Cloud Platform (GCP). It describes how infrastructure provisioning, software deployment, outages, and scaling were challenging prior to cloud computing. GCP provides capabilities like auto-scaling, load balancing, monitoring and self-healing that address these issues. It also offers serverless architectures, containers, machine learning APIs, and tools to automate infrastructure that improve agility, performance and costs. While some concerns about cloud adoption include security, lock-in and speed, the document argues GCP addresses these through encryption, open standards and massive parallelism.
3. Adding a new server:
* call (using a landline) or send a fax to
the provider
* pay via bank transfer
* wait for a few days/weeks
* set up the server on your own server
room
* hope it won't break
4. deploying software:
* code locally on your OS
* submit to CVS and manually build
* send package and SQL separately to IT
* wait until the time slot they give you
(next week, probably)
* test (by hand) everything is working
* hope it won't break
9. Provisioning infrastructure: AppEngine
“Zero ops” applications. Just deploy
and forget*
Of course you still need to worry about
monitoring, backups, security.. but
infrastructure and scaling are automatic
*you need to adjust to the sandbox
17. Cuatro cosas imposibles antes de la
merienda
* Mi servicio siempre disponible,
aunque un volc n est escupiendo datos sobre miá é
centro de datos
* Sobrevivir al black friday, la superbowl o el brexit
* Acabar el trabajo de una semana en 20 minutos
* Consultar “toda la informaci n del mundo” en 5 segundosó
18.
19.
20. A squirrel did take out half of our
Santa Clara data centre two years back
Mike Christian, Yahoo Director of Engineering
24. Some data center outages reported in 2015:
* Amazon Web Services
* Apple iCloud
* Microsoft Azure
* IBM Softlayer
* Google Cloud Platform
* And of course every hosting with scheduled
maintenance operations (rackspace, digital
ocean, ovh...)
29. Auto scaling.
* Sobrevivir al black friday, la superbowl o el brexit
Add auto-scaling capabilities depending
on
* CPU usage
* Requests
* Your own custom metrics
32. * Acabar el trabajo de una semana en 20 minutos
33.
34. * 7 horas de 16 CPUs procesando sin parar
despu s...é
35. Each second of screen time would require 5,000 processor hours to
realize. Given the nature of the deadlines, our teams needed the
ability to spike to 15,000 cores simultaneously on-demand just to stay
on schedule. And because we don’t realize profits until the end of the
project, we needed to spend as little as possible to get up and
running.
41. A nadie le gusta tener que despertarse a las cinco de la
ma ana para reiniciar un servidor porque sus clientesñ
est n sin servicio: monitorizaci n, alertas, y sistemasá ó
auto-reparables.
47. Machine Learning
* Train first
* Ask later
* or use a pre-trained model
- google Vision API
- google Speech API
- google Translate API
- google Prediction API
http://www.pivothead.com/build/
https://www.youtube.com/watch?time_continue=119&v=eve8DkkVdhI
57. Why not the cloud
It's too slow
I am limited in what I can do
I will get vendor lock-in
I cannot legally host my data in the cloud
Google will spy on my data
58. Google will spy on my data
* GCP is not Gmail. SLA
* Encryption at rest and in transit
* Bring your own keys
* ISO standards
59. Cannot host my
data on the cloud
public/private hybrid clouds
Cloud carrier interconnect,
direct peering and
CDN interconnect