Cloud Pro: Google Cloud Compute: Choose, Compare and Launch

What's inside

1. Level 1: Every Google Cloud Compute Service on One PageFree
   Learn to sort every Google Cloud compute service by hosting model (VMs, fleets, batch and HPC, accelerators, containers, serverless, specialised hosting), and connect Google's own decision guidance and the current and former product names to each one.
2. Level 2: The Standard Workload ProfileFree
   Build the reusable workload profile that every later lesson scores against: compute and accelerator needs, memory, coupling, duration and state, latency, scaling, OS and licensing, data location, compliance, team skills and cost model.
3. Level 3: Compute Engine Fundamentals
   Learn what Compute Engine actually gives you (instances, machine types, disks, images, zones and the shared-responsibility line), the limits and quotas that bite first, and the smallest working VM with gcloud and Terraform, plus clean-up.
4. Level 4: General-Purpose Machine Families
   Compare the general-purpose series (E2, N-series, C-series and Tau T2D/T2A) on CPU platform, shared vs dedicated cores, custom shapes, disk types and price, so you can pick a sensible default for web, app and database tiers.
5. Level 5: Compute-Optimised: C2, H3 and H4D
   Learn when the highest per-core performance and HPC-tuned series (C2, H3 and H4D) beat general-purpose machines, what they cost, and the constraints on shapes, regions and networking that come with them.
6. Level 6: Memory-Optimised Machine Families
   Size large in-memory workloads such as SAP HANA, in-memory databases and analytics on the memory-optimised series, including their certification, availability and commitment constraints.
7. Level 7: Accelerator-Optimised: A3, A4 and G2
   Choose between accelerator-optimised GPU series (A4X, A4, A3, A2, G4, G2) for training, fine-tuning, inference and graphics, and learn how GPU capacity is actually obtained: quotas, reservations, Spot and DWS.
8. Level 8: Spot VMs
   Learn how Spot VMs trade preemption risk for deep discounts, how to design workloads that survive being preempted, and where Spot capacity is unreliable.
9. Level 9: Sole-Tenant Nodes
   Understand when dedicated physical hosts are worth paying for (bring-your-own-licence rules, compliance isolation and workload separation) and what node groups, affinity labels and overcommit mean in practice.
10. Level 10: Bare Metal Instances on Compute Engine
    Learn when Compute Engine bare metal instances, which run without a hypervisor, are the right answer for nested virtualisation, custom hypervisors, low-level performance tools or licence-bound software, and what you give up compared with VMs.
11. Level 11: Managed Instance Groups
    Run fleets of identical VMs with managed instance groups: templates, autoscaling, autohealing, regional distribution, rolling updates and stateful MIGs.
12. Level 12: Google Cloud Batch
    Run embarrassingly parallel and container or script jobs on Google Cloud Batch without managing a scheduler, including task arrays, retries, Spot provisioning and GPU jobs.
13. Level 13: Cluster Toolkit for HPC
    Deploy a Slurm-based HPC cluster with Cluster Toolkit (formerly HPC Toolkit) blueprints for tightly coupled MPI work, with compact placement, high-bandwidth networking and shared storage.
14. Level 14: Cloud TPU
    Learn when Google's TPUs beat GPUs for training and inference: TPU generations, slices and topologies, frameworks, capacity options, and getting a TPU VM running.
15. Level 15: GKE Standard
    Run containers on GKE Standard, where you manage node pools, machine types, upgrades and autoscaling, and learn which cluster limits and costs matter most.
16. Level 16: GKE Autopilot
    Learn how GKE Autopilot moves node management to Google, bills per pod resource request, and restricts what you can run, so you can choose between Autopilot and Standard for a workload.
17. Level 17: Cloud Run Services
    Deploy request-driven containers on Cloud Run services, using concurrency, instance limits, CPU allocation modes and scale to zero to manage cost and latency.
18. Level 18: Cloud Run Jobs
    Run tasks that finish and exit (batch processing, migrations and scheduled work) as Cloud Run jobs, with parallel tasks, retries and task timeouts.
19. Level 19: Cloud Run Worker Pools
    Run always-on background work that does not handle requests (queue consumers, pull-based workers and agents) on Cloud Run worker pools, and learn their launch stage, scaling model and limits.
20. Level 20: GPUs on Cloud Run
    Serve inference and run GPU work on Cloud Run with fully managed NVIDIA GPUs, and learn the GPU types, regions, cold starts and cost trade-offs compared with GPU VMs and GKE.
21. Level 21: Cloud Run Functions (formerly Cloud Functions)
    Learn how Cloud Run functions, formerly Cloud Functions, run event-driven, single-purpose code on Cloud Run, and when a function fits better than a full Cloud Run service. The lesson follows the nine-part pattern, covering triggers, limits and trade-offs.
22. Level 22: App Engine Standard Environment
    Assess App Engine standard as a fully managed, sandboxed platform for web applications. Cover its supported runtimes, scaling types, instance classes and the limits that steer new projects towards Cloud Run.
23. Level 23: App Engine Flexible Environment
    Understand App Engine flexible as containers running on Compute Engine VMs managed by Google. See how it differs from standard on start-up time, minimum instances, pricing and operations, and why its nearest alternatives are usually Cloud Run or GKE.
24. Level 24: Google Cloud VMware Engine
    Learn how Google Cloud VMware Engine runs a dedicated VMware private cloud on Google infrastructure. Cover the node types, licensing and the minimum-footprint costs that decide whether lift-and-shift of vSphere estates is worth it.
25. Level 25: Bare Metal Solution
    Evaluate Bare Metal Solution for specialised workloads, such as Oracle databases with licensing or certification constraints. Compare it with Compute Engine bare metal instances and sole-tenant nodes.
26. Level 26: Sizing Machine Families for Compute-Intensive and HPC
    Turn a workload profile into a concrete machine family, shape and placement choice for compute-intensive and HPC jobs. Cover cores against memory bandwidth, network tiers, compact placement and accelerator fit.
27. Level 27: Pricing Models: On-Demand, Commitments and Spot
    Compare how on-demand pricing, sustained use discounts, resource-based and spend-based committed use discounts, and Spot pricing apply across Compute Engine, GKE, Cloud Run and accelerators. Learn the risks each one carries.
28. Level 28: Worked Total Cost of Ownership
    Build a full, dated total-cost model for one workload, including compute, storage, network egress, licences, support and operations effort. Use it to compare two Google Cloud hosting options.
29. Level 29: Fair Benchmarking of Compute Options
    Design benchmarks that compare compute options honestly. Use representative workloads, controlled variables, repeated runs and cost-normalised results rather than headline vendor numbers.
30. Level 30: Networking That Compute Depends On
    Learn the VPC, subnet, firewall, load balancing, Private Service Connect, Serverless VPC Access and Direct VPC egress choices that every compute service relies on, and how they shape latency, security and cost.
31. Level 31: Storage and Identity for Compute Workloads
    Match Persistent Disk, Hyperdisk, Local SSD, Filestore, Parallelstore-style parallel file systems and Cloud Storage to workload I/O patterns. Secure the compute with service accounts, Workload Identity and IAM least privilege.
32. Level 32: Migration Paths onto Google Cloud Compute
    Compare rehost, replatform and refactor paths onto the compute services. Use Google's migration tooling and a staged cutover plan that limits risk and downtime.
33. Level 33: Building a Weighted Selection Scorecard
    Turn the standard workload profile into a weighted scorecard whose dimensions can be scored for any service or cloud. Include clear scoring rules and evidence requirements.
34. Level 34: Comparing Google Cloud with Another Provider
    Apply the scorecard to compare Google Cloud fairly with an equivalent service on another cloud. Match like-for-like offerings and normalise for price, limits, region coverage and operational effort.
35. Level 35: Final Challenge: Compute Decision Workshop Exam
    Work through a full decision workshop on a compute-intensive workload, drawing on the whole course: profile, shortlist, sizing, cost, benchmark plan, networking, storage and identity, migration and scorecard. Finish with a one-page selection record.

Access

The first 2 levels are free with a free account. Every level, the podcast edition and the AI tutor come with All Access at £4.99/month or any Creator plan — see pricing.

Start free →