Google Professional Cloud Architect Certification Case Study Analysis

Once again, I’m going to use RapGenius annotations to mark and analyze parts of the case studies.

Individual Cases

1. Dress4Win

https://genius.it/cloud.google.com/certification/guides/cloud-architect/casestudy-dress4win

This is a web development company running existing servers in a colocated data centre.  They care about IAM, ability to scale and lowering costs.

Summary:

Migrate Existing components to managed services

  • MySQL -> CloudSQL
  • Apache Beam -> Dataflow
  • Java Tomcat -> potentially App Engine
  • Nginx static content -> Cloud Storage and CDN
  • Redis has no equivalent yet
  • Hadoop/Spark servers -> Dataproc
  • MQ servers -> Cloud PubSub
  • Monitoring servers -> Stackdriver suite
  • Back up their images, logs onto Cloud Storage

Other

  • Use automated/repeatable resource creation tools such as Cloud SDK and Cloud Deployment Manager
  • Use managed services (aka avoid Compute Engine)
  • Use VPN tunnels (CloudVPN and Cloud Router)

2. TerramEarth

https://genius.it/cloud.google.com/certification/guides/cloud-architect/casestudy-terramearth

This is a industrial vehicle manufacturer that streams live data about their vehicles in the field through cell network, or batch uploaded through their servicing centres.  They care mostly about turnaround time on their data, not costs.

Summary:

  • Case study is almost completely data oriented
  • Two sources of big data: 90% from batch uploads from service partners, 10% realtime uploads through cell network
  • Pub sub, cloud storage and dataflow for live data from mobile networks
  • Cloud storage, dataflow for data from service centres
  • Store data in BigQuery warehouse
  • Alternative to cloud storage/pubsub is to VPN data from their FTP to dataflow or interconnect
  • IAM issues on how to share data with their dealers and potential partners

3. MountKirk

https://genius.it/cloud.google.com/certification/guides/cloud-architect/casestudy-mountkirkgames

This is a mobile gaming company starting project from scratch.  They want to focus on scaling and analysis of game data.

Summary:

Game server

  • Autoscaling Managed instance groups for custom Linux distros on Compute Engine for game server
  • NoSql database, probably Cloud Datastore for game server (BigTable is more for analytics in petabytes of data).

Analytics

  • Pub sub to handle/buffer data from mobile phones and game servers
  • Cloud storage to handle uploaded files from mobile phones (and maybe cloud functions to process them)
  • Dataflow to process all of the cloud storage and pub sub data
  • Bigquery data warehouse to house all the analytics
  • Maybe Datalab for analysis and visualization of the data

4. JencoMart

https://genius.it/cloud.google.com/certification/guides/cloud-architect/casestudy-jencomart

This is a retailer with an online shop.  They want to expand into Asia.

Summary:

  • Requires Global load balancing and low latency in Asia (interconnects/peering?)
  • Potentially use App Engine for their LAMP stack (Customer Loyalty Portal)
  • Use of custom Compute Engine (RAM and CPU) with persistent disks
  • Google Cloud Storage 100 TB
  • Can move one of the DBs (Postgres not Oracle) to CloudSQL
  • Want to explore data analytics in the future (BigQuery, Dataflow, Dataproc, Datalab
    Cloud Machine Learning / Tensorflow)

Summary by Product

Dress4Win TerramEarth Mountkirk JencoMart
Compute
Compute Engine (Premade or custom VMs) AVOID (not managed service) Required for game server Custom CPU/RAM
App Engine (PaaS for apps or backends) (Maybe) Java Tomcat Maybe LAMP stack
Kubernetes Engine (Run containers)
Cloud Functions (Serverless) Maybe process mobile uploads
Big Data
Big Query (Fully managed large scale data warehouse) Store data in data warehouse House data To explore
Cloud Dataflow (Real time batch and stream processing) Migrate Apache Beam Data from service partners, cell networks Process data To explore
Cloud Dataproc (Managed Hadoop & Spark) Migrate Hadoop & Spark To explore
Cloud PubSub (Ingest streams from anywhere) Migrate MQ servers Data from cell networks Buffer data from mobile To explore
Cloud Datalab (Explore, analyse, visualize datasets) Maybe To explore
Identity & Security
IAM (Fine grained identity and access management) Yes Share access to data with outsiders
Cloud Identity Aware Proxy
Storage & DBs
Cloud Storage (Equivalent to S3) Nginx static content, backup images and logs Data from service partners, cell networks Store mobile uploads Yes
Cloud SQL (Fully managed MySQL or Postgres) Migrate Mysql Postgres (but still has Oracle)
Cloud Bigtable (Fully managed NoSQL, petabytes)
Cloud Spanner (Horizontally scalable ACID relational SQL)
Cloud Datastore (NoSQL, smaller data than bigtable, has transactions) Game server data
Persistent Disk Attached to compute engine
Management Tools
Stackdriver (Monitoring, logging, tracing)
Cloud Deployment Manager Use automation tools
Cloud Shell/APIs Use automation tools
Networking
VPC VPN tunnels Maybe VPN
Load balancing Global Global
Cloud CDN Nginx static content
Cloud Interconnect VPN tunnels Maybe VPN Asia low latency