DSST Computing and IT cheat sheet

CPU cycle: fetch → decode → execute → store (write-back); clock speed in GHz sets cycles per second, not work per cycle (IPC matters).

Memory hierarchy by speed and cost per bit, fastest first: registers → L1/L2/L3 cache → RAM → SSD → HDD → tape; each level trades capacity for latency.

Volatile memory loses contents on power loss (RAM, cache); non-volatile retains it (ROM, flash, SSD, HDD, optical).

Von Neumann architecture shares one bus/path for instructions and data (von Neumann bottleneck); Harvard architecture uses separate instruction and data paths.

Storage units: 1 KB = 1,000 bytes (SI) or 1,024 bytes (binary); 1 MiB = 1,024 KiB; 1 GiB = 1,024 MiB; 1 TiB = 1,024 GiB.

I/O methods: programmed I/O (CPU polls), interrupt-driven I/O (device signals CPU), DMA (controller moves blocks without per-byte CPU work).

RAID: 0 = striping, no redundancy; 1 = mirroring; 5 = striping with distributed parity, survives one disk failure; 6 = two parity blocks, survives two failures.

Ports and buses: USB is hot-swappable serial; SATA connects drives; PCIe connects expansion cards; HDMI/DisplayPort carry digital video and audio.

Cloud service models: IaaS supplies virtual machines and storage; PaaS supplies a managed runtime; SaaS supplies finished applications.

Proprietary/EULA license: source closed, copying and modification restricted, per-seat or per-device terms set by the vendor.

Free and open-source (FOSS) licenses grant the four freedoms (run, study, redistribute, improve); copyleft (GPL) requires derivative works to stay under the same license.

Permissive licenses (MIT, BSD, Apache 2.0) allow relicensing into closed products; Apache 2.0 adds an express patent grant.

Public domain = no copyright retained; Creative Commons licenses apply to creative works and range from attribution-only to non-commercial and no-derivatives variants.

License models: per-seat (named user), concurrent/floating (limited simultaneous users), site license, subscription (term), perpetual (one-time), volume/enterprise agreement.

OEM licenses bind to the machine they ship with and are generally non-transferable; retail licenses may be transferable under vendor terms.

Toolchain roles: source editor → compiler/interpreter → linker → loader → debugger; version control (Git) tracks changes; CI pipelines build and test automatically.

Interpreted languages execute line by line at runtime; compiled languages translate to machine code before execution, trading portability for speed.

Software audit risk: unlicensed installs, expired subscriptions, and license overuse (more installs than seats) are the common findings.

Waterfall: sequential phases (requirements → design → implementation → testing → deployment → maintenance); changes are costly after a phase is signed off.

Agile: short iterations (sprints, commonly 1–4 weeks) producing working increments; values individuals, working software, customer collaboration, and response to change.

Scrum roles: Product Owner (backlog priority), Scrum Master (process facilitator), Development Team; artifacts are product backlog, sprint backlog, increment.

Kanban: continuous flow with a visible board and work-in-progress limits; no fixed-length iterations required.

Spiral model: risk-driven, repeats planning → risk analysis → engineering → evaluation per loop; suited to large, high-risk projects.

V-model: each development stage pairs with a matching test level (requirements ↔ acceptance, design ↔ system/integration, code ↔ unit).

DevOps: continuous integration, continuous delivery, infrastructure as code, and monitoring; shortens feedback loops between build and operations.

Prototyping and RAD favor early user feedback over exhaustive upfront specification; throwaway prototypes must not ship as production code.

Maintenance categories: corrective (fix defects), adaptive (new environment), perfective (improve), preventive (reduce future risk).

Relational model: tables (relations) of rows (tuples) and columns (attributes); a primary key uniquely identifies a row; a foreign key references another table's primary key.

Normalization: 1NF removes repeating groups; 2NF removes partial dependency on part of a composite key; 3NF removes transitive dependency on non-key attributes.

ACID: Atomicity (all or nothing), Consistency (valid state transitions), Isolation (concurrent transactions do not interfere), Durability (committed data survives failure).

SQL subsets: DDL (CREATE, ALTER, DROP), DML (SELECT, INSERT, UPDATE, DELETE), DCL (GRANT, REVOKE).

CRUD maps to Create (INSERT), Read (SELECT), Update (UPDATE), Delete (DELETE).

NoSQL families: key-value, document, column-family, graph; chosen for horizontal scale and schema flexibility rather than strict joins.

Data warehouse vs. data mart vs. data lake: warehouse = integrated enterprise store for analysis; mart = subset for one department; lake = raw files of varied formats.

Backup types: full copies everything; incremental copies changes since the last backup of any type; differential copies changes since the last full backup.

Data integrity controls: validation rules, referential integrity constraints, access permissions, and audit logs.

OSI seven layers, bottom up: Physical, Data Link, Network, Transport, Session, Presentation, Application.

TCP/IP four layers: Network Access (link), Internet, Transport, Application; TCP is connection-oriented and reliable; UDP is connectionless and best-effort.

IPv4 addresses are 32 bits; IPv6 addresses are 128 bits; a /24 IPv4 network has 256 addresses, 254 usable hosts.

Private IPv4 ranges: 10.0.0.0/8, 172.16.0.0/12, 192.168.0.0/16; these are not routed on the public Internet and require NAT.

DNS resolves names to IP addresses; DHCP assigns addresses automatically; ARP maps IP to MAC on a local network.

LAN vs. WAN vs. MAN by scope; Ethernet and Wi-Fi are LAN technologies; MPLS and leased lines are WAN technologies.

Topologies: star (central switch, single point of failure), mesh (redundant paths), bus (shared medium, collisions), ring (token or dual-ring).

Wireless standards: 802.11 is Wi-Fi; 802.3 is Ethernet; Bluetooth is a short-range personal-area network (WPAN).

Bandwidth is capacity (bits per second); latency is delay (milliseconds); throughput is achieved rate, always at or below bandwidth.

PII (personally identifiable information) is data that can single out an individual, alone or combined: name, address, ID number, biometrics, device identifiers.

Fair information principles: notice, choice/consent, access, correction, data minimization, purpose limitation, security, accountability.

Privacy vs. security vs. confidentiality: privacy is control over personal data; security protects data from unauthorized access; confidentiality is the secrecy property.

Anonymization removes direct identifiers; pseudonymization replaces them with keys; re-identification attacks can defeat weak anonymization.

Cookies, tracking pixels, and browser fingerprints are common web tracking mechanisms; opt-in consent is required in many jurisdictions for non-essential tracking.

Data breach notification duties typically require prompt notice to regulators and affected individuals when personal data is compromised.

Workplace and school monitoring (email scanning, keystroke logging, CCTV) is generally lawful where disclosed and proportionate.

Privacy-enhancing technologies: encryption at rest and in transit, access controls, data retention limits, and differential privacy for aggregate statistics.

Cross-border data transfer restrictions limit moving personal data to jurisdictions with weaker protection unless safeguards apply.

Copyright protects original expression fixed in a tangible medium (text, code, music, images), not ideas, facts, or methods; term is typically life of author plus 70 years in many jurisdictions.

Patent protects novel, non-obvious, useful inventions and processes for a limited term (commonly 20 years from filing) in exchange for public disclosure.

Trademark protects signs identifying source of goods or services (names, logos, slogans); renewable indefinitely while in use and distinctive.

Trade secret protects confidential business information (algorithms, formulas, customer lists) with no registration; protection lasts only while secrecy is maintained.

Software is generally protected by copyright (code) and may be patented (technical methods); APIs and interfaces may be treated differently across jurisdictions.

Fair use/fair dealing is a defense, not a right: factors include purpose, nature of work, amount used, and market effect.

Work made for hire: rights in work created by an employee within scope of employment vest in the employer by default.

Public domain works have no copyright owner; Creative Commons and open-source licenses grant permissions in advance under stated conditions.

Infringement remedies include injunction, damages (actual or statutory), account of profits, and criminal penalties for commercial-scale piracy.

Telecommunications regulation typically separates carriage (transmission) from content; common carriers may face non-discrimination and interconnection duties.

Net neutrality rules, where adopted, prohibit blocking, throttling, and paid prioritization of lawful Internet traffic.

Spectrum is licensed by national regulators (e.g., FCC in the US, Ofcom in the UK); unlicensed bands (2.4 GHz, 5 GHz) are open under power limits.

Universal service obligations fund access in high-cost or rural areas and may subsidize schools, libraries, and low-income users.

Lawful interception and data retention rules require carriers to retain or disclose certain records under legal process, subject to privacy law.

Interconnection and roaming agreements, number portability, and emergency-call (911/112) obligations are common regulatory requirements.

Slamming (switching a customer's carrier without consent) and cramming (unauthorized charges on a bill) are prohibited practices.

Radio interference rules: devices must accept interference and must not cause harmful interference in unlicensed bands.

Regulatory bodies also police mergers, rate regulation for dominant carriers, and consumer complaint processes.

Netiquette basics: identify yourself, quote only relevant text, stay on topic, avoid ALL CAPS (read as shouting), and proofread before sending.

Email etiquette: concise subject lines, one topic per message, reply-all only when all recipients need it, and respect others' time with clear asks.

Spam and phishing are prohibited; never forward chain letters, hoaxes, or unverified warnings; do not harvest addresses.

Flame wars escalate from personal attacks; respond to the argument, not the person, and avoid public shaming (doxxing).

Acceptable use policies (AUPs) govern organizational networks: no unauthorized access, no disruptive traffic, no commercial use on private systems.

Respect bandwidth and shared resources: large attachments and streaming consume common capacity; use appropriate channels.

Attribution matters: credit sources and authors when reposting; do not misrepresent others' words.

In synchronous chat and forums, avoid cross-posting the same question to many groups and respect each community's posted rules.

In professional settings, treat others' personal data and screenshots with consent and confidentiality.

AI categories: narrow (task-specific, deployed today), general (human-level across tasks, hypothetical), and superintelligent (beyond human capability, speculative).

Machine learning types: supervised (labeled data), unsupervised (structure in unlabeled data), reinforcement (reward from interaction).

Turing Test: a machine passes if a human interrogator cannot distinguish its responses from a human's; it measures behavior, not understanding.

Search techniques: breadth-first (complete, memory-heavy), depth-first (memory-light, may not terminate), heuristic/A* (goal-directed with estimated cost).

Expert systems separate a knowledge base (facts and rules) from an inference engine (forward chaining from data, backward chaining from goals).

Neural networks learn weights from examples; deep learning uses many layers; training requires data, compute, and validation to avoid overfitting.

Bias in AI arises from unrepresentative training data or flawed objectives; transparency, explainability, and human oversight are proposed mitigations.

Robotics combines perception, planning, and actuation; autonomy levels describe how much human intervention is required.

AI governance concerns include accountability for automated decisions, data protection, and disclosure when users interact with a bot.

Outsourcing moves work to an external provider; offshoring moves work to another country; near-shoring keeps it in a nearby country or time zone.

Follow-the-sun support hands work between time zones so issues progress 24 hours a day.

Language, currency, date/time formats, and address conventions are localization concerns; internationalization designs software to be adaptable without code changes.

Regulatory fragmentation: data protection (GDPR), export controls, and tax rules differ by country and affect where data and services may reside.

Digital divide: unequal access to connectivity and devices across and within countries; globalization can widen or narrow it.

Time zones complicate scheduling; UTC is the standard reference; local time adds offset and daylight-saving rules.

Global supply chains for hardware create concentration risk (single-region fabrication) and tariff exposure.

Cultural differences affect communication norms (directness, hierarchy, holidays) and thus virtual team management.

Currency and payment systems: exchange rates, purchasing power, and local payment methods shape pricing and contracting.