Google Cloud Professional Cloud Architect Exam Guide
Software Architect’s library | blackthorn.ai
- Solutions to Sample Case Studies
- Online Proctored Exam — Hints
- Web Resources
- Mobile Resources
Note: This guide discusses v1.0 of the exam
There are many good resources (see appropriate sections at the end) to prepare for the exam — this article does not pretend to be such a resource — it shares personal experience and recommendations.
I took A Cloud Guru’s online course (awful), then Cloud Academy’s (acceptable), then Linux Academy’s (the best, but it is dead now), then Coursera’s (official high-level overview). I passed a corporate preparation class (didn’t anticipate an alarming pace). I installed all available Android apps and found two of them very useful (see Mobile Resources section). I’ve screened 4 certification books and found Dan Sulivan’s book to be the best (See Books section). It took me ~2 months to prepare learning from 2 to 4 hours every day plus full-time — on weekends and holidays. I had 3+ years of hands-on experience using GCP and passed Google Cloud Professional Data Engineer certification by the exam time.
I passed the online proctored exam on the first attempt.
Solutions to Sample Case Studies
The case studies are available on the exam website at any time. You will see all 3 on the exam. Case studies-related questions make up ~30% of all questions. You will see case studies during the exam in a side-by-side format.
Hint: do not (even) try to memorize them — learn the solutions
- Serve multi-regional instance group backends by a global HTTP Load Balancer
- Multi-regional ingest/storage — Pub/Sub + Datastore/BigQuery/Cloud Storage
- Monitoring — Stackdriver + its metrics can drive GCE group scaling
- Slow/late data — Pub/Sub (scales and buffers slow/late data)+ Dataflow (data processing, windowing, the arrangement of late data)
- Reduce latency to all customers — multi-regional GCE managed instance groups+ HTTP load balancer, multi-region Datastore
- Scaling — autoscaling managed instance groups (hardened Linux distro can be run), Stackdriver can drive scaling up/down, HTTP Load Balancer scales automatically
- User profiles and game state — store in a NoSQL transactional database (Datastore)
- Game activity — store in a time-series database for future analysis (Bigtable — if ms response time / low latency is a requirement, BigQuery otherwise).
- SQL queries on historical data > 10TB — BigQuery
- Process files uploaded by users’ mobile devices — upload to Google Storage, process via Dataflow
- Migration strategy — move data first, then build a prototype, then move the applications.
- Deployment automation — gcloud for automated management scripts: Cloud Deployment Manager, Terraform, other IaaC products
- CI/CD for on-premises datacenter or cloud — Jenkins, Spinnaker, Cloud Build, etc.
- Failover support — replicating environment on Google Cloud, replicate MySQL to Cloud SQL, on-premises/cloud application servers — DNS cutover
- Encrypt data on the wire and at rest — customer-supplied (custom) encryption keys. Customer uploads these Base64-encoded encryption keys to a private Google Storage bucket and then uses them as gsutil parameters every time he/she uses the data.
- Support multiple private connections between the production data center and cloud environment:
A) Cloud VPN (max 8 tunnels each can support up to 3 Gbps ingress/egress — in general this option works for connections ≤ 3 Gbps; has IPsec and IKE; no SLA; data travels over public internet). Not highly available by default. HA Cloud VPN is 2 paired VPN tunnels.
B) Cloud Interconnect (physical connections ~10 Gbps, same RFC 1918 IP space) Dedicated — direct cable to Google facilities (high-end option, up to 80 Gbps, ≥99.9% SLA), or Partner — cable to one of Google’s partners (up to 10 Gbps, ≥99.9% SLA)
- Lift and shift MySQL to Cloud SQL. Cloud SQL supports MySQL natively. Single region if no global footprint required. Migration strategy: create a read replica managed by Cloud SQL, sync it with on-premises MySQL, update applications to point to the replica, promote the replica to a stand-alone instance
- Move Redis 3 server to one of the following:
A) Deploy Redis server on Compute Engine
B) Export to a new Memorystore managed Redis database
- Move web application servers (microservices) to:
A) autoscaled GCE managed instance groups with custom machine types
B) Re-architect for Google Kubernetes Engine or Google AppEngine
- Move Apache Hadoop/Spark servers to Cloud Dataproc, connect Dataproc to Cloud Storage
- Replace RabbitMQ with Pub/Sub or deploy RabbitMQ on GCE managed instance group
- Jenkins, monitoring, bastion hosts, security scanners — no service equivalents, move as-is to GCE instances (custom machine types)
- iSCSI for VM hosts / Fiber channel SAN (block storage) — move to GCE persistent disks working in SAN cluster.
- NAS (image storage, logs, backups) — move to Cloud Storage or Filestore
- Convert to 100% cellular data if possible. This is the preferred option to decrease unplanned vehicle downtime to less than 1 week.
- Manage cellular devices data ingestion with Cloud IoT Core. + built-in managed security.
- Use Cloud Endpoints to manage and protect APIs
- Pub/Sub — to ingest data from all cellular devices, globally
- Dataflow processes data (including the streaming!) from Pub/Sub and inserts it into BigQuery for storage/analytics. An alternative is to stage the data Cloud Storage, clean with Dataprep, and run a job backed by Dataflow into BigQuery
- Cloud ML Engine (AI Platform) uses data from BigQuery and other sources to predict customer needs, breakdowns, optimize parameters and deploy updated parameters back to machines to update config
- Data Studio — to display analytics data/insights that can be shared with dealers via dashboards. + the same to create compelling joint offers in partnership with different companies.
- Datalab — visual notebooks for working with BigQuery/Cloud ML Engine (AI Platform) data for ML/analytics
- Multi-regional / global services — for expansion-related questions
- Export BigQuery to Cloud Storage regularly — backup strategy
Online Proctored Exam — Hints
The exam is available both in a test center and online format. I took the online proctored exam and will briefly outline major points:
- Prior to the exam: create Webassessor Biometric Profile (they will take your photo), download and install the Sentinel app, and grant it required permissions (camera and microphone access, etc.)
- The app and exam are demanding on internet speed. Test your system before the exam! All checks should be green. Do not try to pass the exam if even one check is red — it will definitely cause you issues and may become the only root of failure.
- Bring 2 forms of your government-issued photo ID (driver’s license and ID card for example) where your face and full name are visible.
- A supervisor will look at you during all 2 hours of the exam through your laptop or web camera and listen to you via your microphone.
- Your table must be clear, no hand watches are allowed, only one screen allowed (the rest must be turned off). No other people are allowed.
- Hand notes are prohibited — but you can write comments and use an integrated whiteboard in the Sentinel app.
- You are not allowed to leave your desk during the exam. Make sure you’ve visited WC :)
- The exam takes place in the dedicated desktop application which blocks access to other applications and screens.
- The supervisor will ask you to record your workplace all around, including the ceiling and floor using your laptop or mobile phone.
- You will have to show your identity documents to the camera, they must be accepted. I recommend choosing documents with expiration dates (like local/foreign passports), attendees report issues with driver licenses without expiration dates.
- My supervisor was late by 36 (!) minutes once. All this time I thought I messed up the exam time or there was a bug in the system.
- Other attendees report that the Sentinel app is buggy: freezes without a reason, shows errors, etc. Remember: do not close the application window by yourself — it will be treated as (an empty) exam submission. Deep breath, calm down and call/write support.
- Do not hesitate to contact support in case of any issues before, during, or after the exam — they are pretty friendly and helpful.
- Former Linux Academy Lucidchart — this is a gem, don’t miss it
- Official GCP sample questions — another must-have, pass them the day before the exam
- Official GCP exam roadmap — planning and gap discovery
- Coursera — Preparing for the Google Cloud Professional Cloud Architect Exam — good for the high-level overview — not for the learning! — and test questions analysis in the last 4–6 days prior to the exam
- Examtopics test sets — bloody hack. For free! Many exam questions could be found on this forum. Not all answers are correct so be mindful and check comments from other contributors.
- Do not forget to set up a free 1 year 300$ credit GCP account
Mobile apps are very handy as the device is always with you — you can train in a subway or before bedtime.
- Test prep. for Google Professional Cloud Architect by Magic Bytes Soft — rating is 9/10
- Linux Academy mobile application — poor UI, but their videos are brilliant. Question/answer cards are a convenient tool for memorizing key information.
- Official Google Cloud Certified Professional Cloud Architect Study Guide by Dan Sulivan — great book, covers 100% of the exam objectives, contains a lot of relevant quizzes. Google a bit and download it for free
Do not wonder that there are no more books in this section — you will not need them once you’ve got Dan Sulivan’s book.
- What is the motivation to pass this certification?
Employee’s major motivation is to get the badge for the CV (to sell yourself more expensive later) + learn how to make the best possible decisions architecting on the Google Cloud Platform. But mostly these badges are desired by business owners, as a certain number of certified employees is required to sign a partnership with the cloud provider (i.e. GCP, AWS, Azure, etc.) and get a flow of projects from them.
- Is the exam useful for Developer/Engineer?
It does not teach you how to use GCP services, it teaches what tools are appropriate for which case.
- Is the exam useful for Architect?
An architect must know what tools are best to use in which case + tradeoffs. Sample Case Studies serve as reusable GCP patterns.
- Typical preparation time?
1–2 months, 2–3 hours a working day + full weekends
- Is it hard?
- Linux Academy was a great resource. RIP
- A Cloud Guru is low-grade. Even quizzes are messed up.
Hint: you can hack their subscription system by creating a new account every 7 days of a free trial with a new email
- Udemy test exams don’t correspond to reality, don’t buy them!
- Coursera’s preparation exam is a great high-level overview. Pay special attention to how they inspect and resolve quizzes. This is the way your thinking should work during the exam.
- Pay special attention to Google’s sample questions. After you complete the form, answers become visible. Make sure you understand why you were right or wrong in each response.
- “You have to have at least 2 years of hands-on GCP experience”
I had 3+ years of GCP development experience and a Google Cloud Professional Data Engineer certificate.
- Official expert-led Google Cloud training
I took them in the scope of the corporate program and I didn’t learn a thing from an ordinary guy reading GCP docs aloud
- Make sure you’ve memorized solutions for the sample case studies (Mountkirk Games, Dress4Win, TerramEarth). Solutions, not the cases.
- Mind buying the exam voucher ~1 month in advance. The exam is in high demand — if you decide to buy the voucher in the last few days, very possibly you will end up in a situation when there are no free time slots in the near 2–3 weeks.