Cloud storage Study Guide

📖 Core Concepts Cloud Storage – Remote data storage in logical pools accessed over the Internet; the provider owns and operates the hardware, networking, and facilities. Virtualization & Elasticity – Resources are abstracted and can expand or shrink “near‑instantaneously” to match demand. Multi‑tenancy – Multiple customers share the same physical infrastructure while perceiving isolated storage. Object, File, Block Storage – Three primary service models: Object – Stores data as self‑describing objects (metadata + unique ID). File – Presents a hierarchical directory tree accessed via NAS protocols. Block – Supplies raw storage volumes that behave like a local disk. Eventual Consistency – Updates propagate to all replicas over time; reads may temporarily see stale data. Hybrid Cloud Storage – Combines on‑premises storage (or a gateway) with public cloud storage, caching “hot” data locally and keeping “cold” data in the cloud. --- 📌 Must Remember Pay‑as‑you‑go – Billing is based on actual capacity used (usually averaged monthly/quarterly). Redundancy = Fault‑tolerance – Data is replicated across geographic sites, giving high durability. Encryption – Protect data at rest (stored) and in transit (moving). Keep decryption keys with the user, not the provider. SLA Exclusions – Planned maintenance, network outages, human error, natural disasters, force‑majeure, and security breaches are typically not covered. Hybrid Cache Rule – Hot data → local cache; Cold data → cloud. --- 🔄 Key Processes Storing an Object Client sends data + metadata via API → Provider assigns a unique object ID → Data is written to multiple physical nodes → Replicas created (eventual consistency). Hybrid Data Flow User writes to local gateway → Gateway caches data locally (if “hot”) → Asynchronously pushes copy to cloud storage → Retrieval checks cache first, falls back to cloud if miss. Encryption Workflow Data encrypted on‑premises with user‑managed key → Encrypted blob sent over TLS to provider → Provider stores encrypted blob; decryption key never leaves user environment. --- 🔍 Key Comparisons Object vs. File Storage Object: Flat namespace, metadata‑rich, ideal for unstructured data (images, backups). File: Hierarchical directories, POSIX‑like semantics, suited for shared files and applications needing a traditional file system. Block vs. Object Storage Block: Raw volumes, low‑latency, needed for databases & VM disks. Object: Higher latency, but provides built‑in durability and metadata. Public Cloud vs. Hybrid Cloud Public: All data lives off‑site; simple scaling, full provider responsibility. Hybrid: Mix of on‑premises and cloud; control over latency, compliance, and cost‑optimisation via caching. --- ⚠️ Common Misunderstandings “Cloud storage is always cheaper.” It converts CAPEX to OPEX; actual cost depends on usage patterns and data transfer fees. “Data is instantly consistent everywhere.” Most services are eventually consistent; a recent write may not be visible on all replicas right away. “Provider handles all security.” Providers secure the infrastructure, but customers must manage key ownership and encryption of sensitive data. --- 🧠 Mental Models / Intuition “Data as a replicated library” – Imagine each object as a book duplicated in several branches; you can read any branch, but a new edition (update) takes time to appear everywhere. “Cache as a front‑door” – Hybrid storage is like keeping frequently used tools on a workbench (local cache) while the heavy toolbox stays in the garage (cloud). --- 🚩 Exceptions & Edge Cases Compliance‑driven restrictions – Some regulations forbid storing personally identifiable information (PII) outside a specific jurisdiction; hybrid or private clouds may be required. SLA “no‑SLA” services – Certain low‑cost storage tiers provide no formal availability guarantee; rely on the provider’s best effort. Performance variability – Bandwidth‑limited WAN links can cause latency spikes; not all workloads tolerate this. --- 📍 When to Use Which | Situation | Recommended Storage Type | |-----------|--------------------------| | Large unstructured data (media, backups) | Object storage | | Shared files, developer environments, home directories | File storage | | High‑performance DBs, VM disks, low‑latency apps | Block storage | | Need to meet strict latency or compliance locally | Hybrid cloud with on‑premises cache | | Want to avoid managing hardware & get automatic durability | Public object storage | --- 👀 Patterns to Recognize “Hot vs. Cold” data pattern – Exam questions that mention frequent access, low latency, or caching hint at hybrid storage with local cache. “Eventual consistency” clues – Any scenario describing delayed propagation or read‑after‑write anomalies points to object storage semantics. “Cost‑only‑as‑used” wording – Indicates a pay‑per‑gigabyte model, often contrasted with traditional on‑premises capital expense. --- 🗂️ Exam Traps Distractor: “Encryption at rest is optional because the provider secures the hardware.” Why wrong: Without user‑controlled keys, provider staff could access plaintext data. Trap: “SLA guarantees 100 % uptime.” Why wrong: SLAs exclude planned maintenance, outages, and force‑majeure; they usually specify a target (e.g., 99.9 %). Confusion: “Object storage is the same as file storage.” Why wrong: Objects lack a hierarchical namespace and are accessed via APIs, not file‑system calls. Mislead: “Hybrid cloud eliminates all security concerns.” Why wrong: It still inherits WAN risks and may introduce new attack surfaces (gateway devices). ---