Cloud Computing Core Cloud Infrastructure Services Syllabus

Core Cloud Infrastructure Services syllabus — full chapter and topic list

Expand any chapter to see its topics and sub-topics. This is the whole examinable outline for Core Cloud Infrastructure Services in Cloud Computing, not a summary of it.

1. 1

   Compute Services

   4 topics

   • Virtual Machine Instances
   • Auto Scaling Groups
   • Dedicated Hosts and Bare Metal
   • GPU and Specialized Compute
2. 2

   Storage Services

   5 topics

   • Object Storage
   • Block Storage
   • File Storage and Shared File Systems
   • Archival and Cold Storage
   • Data Durability and Replication
3. 3

   Networking in the Cloud

   5 topics

   • Virtual Private Cloud (VPC)
   • Load Balancing
   • DNS and Content Delivery Networks (CDN)
   • VPN, Direct Connect, and Peering
   • NAT Gateways and Internet Gateways
4. 4

   Database Services

   4 topics

   • Relational Databases
   • NoSQL Databases
   • In-Memory Caching
   • Data Warehousing
5. 5

   Identity and Access Management

   4 topics

   • Users, Groups, and Roles
   • Policies and Permissions
   • Multi-Factor Authentication (MFA)
   • Federation and Single Sign-On (SSO)
6. 6

   Monitoring and Logging

   4 topics

   • Metrics and Dashboards
   • Centralized Logging
   • Alarms and Notifications
   • Distributed Tracing and Observability

Core Cloud Infrastructure Services flashcards for Cloud Computing

25 of 51 cards from the Core Cloud Infrastructure Services deck — real questions with worked answers.

1. What is a virtual machine (VM) instance in cloud computing?

   A software-emulated computer running on shared physical hardware via a hypervisor. It has its own virtual CPU, memory, storage, and network interface, and is billed on-demand while sharing the underlying host with other tenants (multi-tenancy).

2. How are cloud compute instances typically classified into families, and what does each optimize for?

   General-purpose (balanced CPU/memory), Compute-optimized (high CPU-to-memory ratio), Memory-optimized (large RAM for in-memory workloads/databases), Storage-optimized (high local disk I/O), and Accelerated/GPU (hardware acceleration for ML/graphics).

3. Compare on-demand, reserved, and spot (preemptible) instance pricing models.

   On-demand: pay per second/hour, no commitment, highest price. Reserved: 1–3 year commitment for large discounts. Spot/preemptible: bid on spare capacity for up to ~90% savings but can be reclaimed with little notice, suited to fault-tolerant/batch work.

4. What is the difference between vertical scaling and horizontal scaling?

   Vertical scaling (scale up) increases the size/power of a single instance (more CPU/RAM). Horizontal scaling (scale out) adds more instances to distribute load. Horizontal scaling improves availability and has near-limitless headroom; vertical scaling is limited by max instance size.

5. What is an Auto Scaling Group (ASG) and what three capacity parameters define it?

   An ASG automatically adjusts the number of running instances to match demand. It is defined by minimum size, maximum size, and desired capacity; it launches/terminates instances to keep the group at desired capacity within the min/max bounds.

6. Contrast dynamic (target-tracking), scheduled, and predictive auto scaling policies.

   Dynamic/target-tracking: reacts to a live metric (e.g., keep CPU at 50%). Scheduled: scales at known times (e.g., business hours). Predictive: uses forecasts/ML on historical patterns to scale ahead of anticipated demand.

7. In an Auto Scaling Group, what is a health check and how does it maintain capacity?

   A periodic probe (instance status or load-balancer check) that determines if an instance is healthy. If an instance fails, the ASG terminates it and launches a replacement to keep the group at desired capacity, providing self-healing.

8. What is a Dedicated Host and how does it differ from a standard multi-tenant instance?

   A Dedicated Host is a full physical server allocated to a single customer, giving control over instance placement, visibility into sockets/cores (useful for per-socket software licensing/BYOL), and physical isolation—unlike shared multi-tenant hardware.

9. What is a bare-metal instance and why choose one over a virtualized instance?

   A bare-metal instance gives direct access to the underlying physical server with no hypervisor. Chosen for maximum performance, workloads sensitive to virtualization overhead, running your own hypervisor, or applications needing hardware features/licensing that require non-virtualized access.

10. Why are GPUs used for machine learning and graphics workloads instead of CPUs?

    GPUs contain thousands of smaller cores optimized for massively parallel arithmetic (SIMD), giving very high throughput on matrix/tensor operations common in deep learning and rendering, whereas CPUs have few cores optimized for sequential, low-latency tasks.

11. Name specialized compute accelerators beyond GPUs and their purpose.

    TPUs (tensor processing units, custom ASICs for neural-net math), FPGAs (reconfigurable hardware for custom acceleration), and inference-specific chips. They trade general flexibility for higher performance/efficiency on targeted workloads like training or low-latency inference.

12. What is object storage and what is its core data model?

    Object storage manages data as discrete objects in a flat namespace (buckets/containers), each object containing the data (blob), a unique key/identifier, and rich metadata. Accessed via HTTP REST APIs rather than a filesystem path or block device.

13. List key characteristics that distinguish object storage from block and file storage.

    Object: flat namespace, metadata-rich, HTTP/REST access, virtually unlimited scale, immutable objects (overwrite whole object), high durability, higher latency. Best for unstructured data—media, backups, static assets, data lakes.

14. What is block storage and how is it typically used?

    Block storage exposes raw fixed-size blocks as a virtual disk/volume attached to a single instance. The OS formats it with a filesystem. Used for boot volumes, databases, and any low-latency, high-IOPS random-access workload.

15. Differentiate IOPS, throughput, and latency as block-storage performance metrics.

    IOPS: input/output operations per second (random-access rate). Throughput: data transferred per second, e.g. MB/s (sequential bandwidth). Latency: time to complete a single I/O request (ms). IOPS matters for small random ops; throughput for large sequential transfers.

16. What is file storage / a shared file system in the cloud, and how is it accessed?

    A managed, network-attached filesystem presenting a hierarchical directory tree that can be mounted concurrently by many instances using protocols like NFS (Linux) or SMB/CIFS (Windows). Ideal for shared content, home directories, and lift-and-shift apps needing POSIX semantics.

17. Compare object, block, and file storage on sharing and access protocol.

    Block: single-instance attach, raw device (iSCSI-like), formatted by OS. File: multi-instance concurrent mount via NFS/SMB, hierarchical. Object: multi-client via HTTP REST, flat namespace with keys. Block=disk, File=shared folder, Object=web-accessible blobs.

18. What is archival / cold storage and what trade-off defines it?

    A low-cost storage tier for rarely accessed, long-retention data (backups, compliance archives). It trades very low per-GB cost for higher retrieval latency (minutes to hours) and retrieval fees, versus instantly accessible hot storage.

19. How do storage classes/tiers manage cost over a data object's lifecycle?

    Data moves from hot (frequent access, high cost, low latency) to cool/infrequent-access, then to cold/archive (low cost, high retrieval latency). Lifecycle policies automate transitions based on age or access patterns to minimize total cost.

20. Define data durability and how it differs from availability.

    Durability is the probability that stored data is not lost/corrupted over time (e.g., '11 nines,' 99.999999999%). Availability is the probability the data/service is accessible when requested (e.g., 99.99% uptime). Durability = data survives; availability = you can reach it now.

21. What does '11 nines' (99.999999999%) of durability imply in practical terms?

    It means an expected annual loss of about one object per 100 billion objects—i.e., if you store 10 million objects you would expect to lose one object roughly once every 10,000 years. Achieved by redundant, replicated storage.

22. Contrast synchronous and asynchronous replication.

    Synchronous: a write is acknowledged only after all replicas confirm, giving zero data loss (RPO≈0) at the cost of higher latency and tighter distance limits. Asynchronous: primary acks immediately and propagates later, giving lower latency but possible data loss on failure.

23. What are RPO and RTO in the context of replication and disaster recovery?

    RPO (Recovery Point Objective): maximum acceptable data loss measured in time (how far back the last usable copy is). RTO (Recovery Time Objective): maximum acceptable downtime to restore service. Lower RPO/RTO cost more and drive replication strategy.

24. How does erasure coding provide durability more efficiently than full replication?

    Erasure coding splits data into $k$ data fragments plus $m$ parity fragments, so any $k$ of the $k+m$ fragments reconstruct the data—tolerating $m$ failures with storage overhead $\frac{k+m}{k}$, far less than the 3x overhead of triple replication.

25. What is a Virtual Private Cloud (VPC)?

    A logically isolated, software-defined virtual network within a public cloud where you control your own IP address range (CIDR block), subnets, route tables, gateways, and security controls—giving a private, customizable network for your resources.

See more Core Cloud Infrastructure Services flashcards →

Planning Core Cloud Infrastructure Services for Cloud Computing

Core Cloud Infrastructure Services (Cloud Computing) FAQ