Cloud Pro: AWS Compute: Choose, Compare and Launch

What's inside

1. Level 1: The AWS Compute Map on One PageFree
   Learners lay out every AWS compute service by hosting model (virtual machines, containers, serverless functions, managed application platforms, batch and HPC, hybrid and edge, specialised hosting). They also learn AWS's own decision guidance and the current and former names that still turn up in docs, consoles and job adverts.
2. Level 2: The Standard Workload ProfileFree
   Learners build the reusable workload profile that every later lesson scores against. It covers CPU/GPU/accelerator and memory needs, tightly coupled versus embarrassingly parallel work, duration and state, latency, scaling, OS and licensing, data location, compliance, team skills and cost model.
3. Level 3: Amazon EC2: The Foundation Service
   Learners work through the nine-part lesson on EC2 itself: AMIs, instance types, EBS and instance store, quotas measured in vCPUs, pricing levers, the shared responsibility split, a smallest working launch with the CLI and CloudFormation or Terraform, and a clean clean-up.
4. Level 4: General, Compute, Memory and Storage Families
   Learners decode EC2 instance naming (family, generation, processor and attribute letters, size) and match the general purpose, compute, memory and storage optimised families to workload profiles. The lesson covers burstable credits and how networking and EBS bandwidth scale with size.
5. Level 5: Accelerated Instances: GPUs, Trainium, Inferentia
   Learners compare NVIDIA GPU instances with AWS Trainium and Inferentia instances for training, inference and graphics. The lesson covers the software stacks each needs (CUDA versus the Neuron SDK), regional availability, the separate accelerator quotas, and cost levers.
6. Level 6: Graviton Versus x86 on EC2
   Learners weigh AWS Graviton (Arm) against Intel and AMD x86 instances on price-performance, software compatibility, licensing and build pipelines. The quick start is a multi-architecture deployment.
7. Level 7: EC2 Spot Instances
   Learners take Spot through the nine-part lesson: how spare capacity is priced and reclaimed, the two-minute interruption notice, allocation strategies, and the workloads that tolerate interruption and those that do not.
8. Level 8: EC2 Capacity Blocks for ML
   Learners cover reserving GPU and Trainium capacity for a fixed future window with Capacity Blocks for ML. The lesson covers which instance types and durations are on offer, upfront pricing, what happens when the block ends, and how this compares with On-Demand Capacity Reservations and Spot.
9. Level 9: EC2 Dedicated Hosts and Dedicated Instances
   Learners cover single-tenant hardware: when bring-your-own-licence software or compliance rules require Dedicated Hosts, how they differ from Dedicated Instances, host-level pricing, AWS License Manager, and the operational overhead of managing host capacity.
10. Level 10: EC2 Auto Scaling
    Learners build Auto Scaling groups with launch templates, mixed instance policies, health checks and target tracking, step, scheduled and predictive scaling. They then cover warm pools, lifecycle hooks and instance refresh for safe rollouts.
11. Level 11: AWS Batch
    Learners take AWS Batch through the nine-part lesson: compute environments on EC2, Spot, Fargate or EKS; job queues; job definitions; array and multi-node parallel jobs; fair-share scheduling; and the fact that you pay only for the underlying compute.
12. Level 12: AWS ParallelCluster
    Learners cover the open-source ParallelCluster tool for building self-managed Slurm HPC clusters on EC2. The lesson covers its configuration file, head and compute nodes, Elastic Fabric Adapter (EFA), placement groups, shared FSx for Lustre storage, and the upgrade and operations burden that stays with you.
13. Level 13: AWS Parallel Computing Service (PCS)
    Learners cover AWS PCS, the managed Slurm service: clusters, compute node groups and queues; what AWS operates compared with ParallelCluster; the supported Slurm versions and limits; and how pricing adds a cluster management fee to the EC2 cost.
14. Level 14: Amazon ECS
    Learners take Amazon ECS through the nine-part lesson: task definitions, services, cluster capacity on EC2 or Fargate, capacity providers, service discovery and load balancing. ECS Express Mode is introduced as the fast path from a container image to a running web service.
15. Level 15: Amazon EKS and EKS Auto Mode
    Learners cover managed Kubernetes on AWS: control plane pricing, managed node groups, Karpenter, add-ons, version support windows and extended support costs. They then cover EKS Auto Mode, where AWS also manages nodes, scaling and core add-ons.
16. Level 16: AWS Fargate
    Learners cover Fargate as serverless capacity for ECS and EKS: task sizes, ephemeral storage, the lack of GPU support and host access, per-second vCPU and memory pricing, Fargate Spot, and Graviton support.
17. Level 17: AWS Lambda
    Learners take Lambda through the nine-part lesson: event sources, memory sizing that also sets CPU, the 15-minute timeout, payload and package limits, concurrency quotas, cold starts, SnapStart and provisioned concurrency, and pricing per request and per GB-second.
18. Level 18: AWS App Runner and Migrating Off It
    Learners cover App Runner's model and limits, and its status: closed to new customers since 30 April 2026, while existing customers keep using it without new features. They then plan a blue/green migration to ECS Express Mode, which AWS recommends, using weighted DNS.
19. Level 19: AWS Elastic Beanstalk
    Learners cover Elastic Beanstalk as a managed application platform on top of EC2, Auto Scaling and load balancers: platforms and their retirement schedules, environment tiers, deployment policies, and the fact that you pay only for the underlying resources.
20. Level 20: Amazon Lightsail
    Learners cover Lightsail's bundled virtual servers, containers, databases and networking at fixed monthly prices. The lesson covers bundle limits, data transfer allowances, VPC peering, and the point at which a workload should move to EC2.
21. Level 21: AWS Outposts: Racks On Premises, Servers Closed
    Learn what AWS Outposts runs in your own data centre, why second-generation racks (including the single-rack option AWS announced in September 2026) are now the only path for new customers after AWS stopped selling the 1U and 2U Outposts servers, and how it fits all nine parts of the lesson pattern.
22. Level 22: Local Zones and Edge Placement for Latency
    Learn how AWS Local Zones put a small set of EC2, EBS and container capacity close to end users, how they differ from Wavelength and Outposts, and which instance types, services and pricing to check in each zone before you rely on one.
23. Level 23: ROSA: Managed Red Hat OpenShift on AWS
    Learn when Red Hat OpenShift Service on AWS beats EKS: what Red Hat, AWS and you each operate, how cluster and worker fees stack on top of EC2 costs, and the hosted-control-plane versus classic architecture choice.
24. Level 24: Amazon EVS: VMware Cloud Foundation in Your VPC
    Learn how Amazon Elastic VMware Service runs VMware Cloud Foundation on EC2 bare-metal hosts inside your own VPC, what you still operate yourself, how licensing works now that Broadcom owns VMware, and when to rehost on EVS versus refactor to native EC2.
25. Level 25: Sizing Compute-Intensive and HPC Workloads
    Learn to turn the workload profile into a concrete machine choice for CPU-, memory-bandwidth- and GPU-bound jobs by matching cores, memory per core, network (EFA) and storage throughput to what the code actually limits on.
26. Level 26: On-Demand vs Savings Plans vs Reserved vs Spot
    Learn how every AWS purchase option prices the same capacity, including Compute and EC2 Instance Savings Plans, Reserved Instances, Spot, Capacity Reservations and Capacity Blocks, and which mix suits steady, spiky and interruptible demand.
27. Level 27: Worked Total Cost of Ownership for a Workload
    Build a full monthly and three-year total cost for one compute-intensive workload: compute plus storage, data transfer, licensing, support, managed-service fees and the engineering time each option costs to run.
28. Level 28: Fair Benchmarking Across Instances and Clouds
    Learn to design benchmarks that give comparable, repeatable results across instance types and providers: representative workloads, a fixed software stack, enough repeat runs, noisy-neighbour and burst controls, and cost-per-result metrics.
29. Level 29: Networking Compute Depends On
    Learn the VPC building blocks every compute service relies on: subnets, routing, NAT, endpoints, security groups, placement groups, EFA and ENA bandwidth, and load balancers, and how each one limits performance and cost.
30. Level 30: Storage Compute Depends On
    Learn to match storage to compute access patterns: instance store, EBS volume types, EFS, FSx for Lustre, ONTAP and OpenZFS, and S3 as a data lake, with the throughput, IOPS and cost each one gives a scratch, shared or durable dataset.
31. Level 31: Identity and Access for Compute Workloads
    Learn how compute gets its permissions safely: instance profiles, ECS task roles, EKS Pod Identity and IRSA, Lambda execution roles, IMDSv2, plus Systems Manager access instead of SSH keys, and how least privilege shapes these choices.
32. Level 32: Migration Paths Between Compute Services
    Learn the realistic routes between hosting models: on-premises or VMware to EC2, EC2 to containers, ECS and EKS in either direction, App Runner to ECS Express Mode, Beanstalk to ECS, and containers to Lambda, with the effort, risk and tooling for each.
33. Level 33: Building a Weighted Compute Selection Scorecard
    Turn the workload profile into a weighted scorecard with provider-neutral dimensions (fit, performance, limits, cost, operations burden, skills, compliance, portability, ecosystem) and objective scoring rules for each.
34. Level 34: Scoring AWS Against Another Cloud
    Apply the scorecard to AWS and a second provider (Azure or Google Cloud) for the same workload, using equivalent services, dated price and quota evidence and the same benchmarks, so the comparison is fair rather than vendor-led.
35. Level 35: Final Challenge: Compute Decision Workshop Exam
    Bring the whole course together by running a full decision workshop on a compute-intensive workload: profile it, shortlist services, check limits, size and cost it, benchmark it, score AWS against another cloud, deploy a proof of concept, and write 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 →