Performance Glossary

Browse 296 performance terms defined in plain English, from the cultural dictionary of computing.

296 Performance Terms

10xer
10xer is informal slang for someone who embodies the 10x engineer archetype, a developer believed to be roughly ten times more productive than an average peer....
Affiliate Marketing
A marketing model where partners promote a product and earn commissions for referrals, leads, or sales they generate. It is common in e-commerce, software, and...
AI Cache
A cache layer used to store and reuse AI-related outputs or intermediate results such as model responses, embeddings, retrieval results, or prompt expansions....
AI Efficiency
The ability of an AI system to deliver useful results with minimal waste in compute, latency, memory, tokens, or operational overhead. AI efficiency matters...
AI Optimization
The process of improving an AI system's quality, latency, cost, or reliability through changes to prompts, models, retrieval, infrastructure, or evaluation. AI...
AI Performance
How well an AI system performs across relevant dimensions such as accuracy, latency, throughput, cost, and user acceptance. AI performance must be judged in...
AI Speed
The responsiveness or throughput of an AI system, usually measured in latency, tokens per second, jobs per minute, or end-to-end completion time. AI speed...
Allocator
A component or algorithm responsible for managing dynamic memory allocation and deallocation — deciding where in the heap to place new objects and how to...
Annual Review
A formal yearly evaluation of an employee's work, growth, and impact, often tied to compensation, promotion, or development planning. In tech culture annual...
Anycast
A network addressing method where the same IP address is advertised from multiple locations worldwide. Traffic is automatically routed to the nearest (by BGP...
AOT Compilation
Ahead-of-Time compilation — translating high-level code to native machine code before deployment, producing a standalone binary. Eliminates cold-start latency,...
Apache Arrow
A cross-language columnar memory format designed for efficient analytical processing. Unlike Parquet (on-disk), Arrow defines how data lives in RAM, enabling...
API Pagination
The technique an API uses to split large result sets into smaller pages, commonly through page numbers, offsets, cursors, or tokens. Pagination is important...
Application Cache
A cache used by an application to store frequently accessed data, computed results, or rendered output so repeated work can be avoided. Application caches...
ARM Architecture
ARM Architecture is a runtime or systems-level concept that affects execution behavior close to the machine. It is commonly used for performance-sensitive...
Array Buffer
Array Buffer is a runtime or systems-level concept that affects execution behavior close to the machine. It is commonly used for performance-sensitive...
Bare Metal
A physical server or computer used directly without a hypervisor, virtual machine, or container abstraction layer, giving the workload exclusive access to the...
Batch Operation
An operation that handles multiple items in one request or execution unit rather than processing each item individually. Batch operations can improve...
Batch Processing Detail
Batch Processing Detail is a runtime or systems-level concept that affects execution behavior close to the machine. It is commonly used for...
Big-O
Big-O notation is a mathematical notation that describes the upper bound of an algorithm's time or space complexity as the input size grows, abstracting away...
Blazing
A casual way to say something is very fast or impressively performant, sometimes sincerely and sometimes with salesy exaggeration. In tech talk, 'blazing'...
Blocking Call
A function call that stops the current thread or execution path from doing anything else until the operation completes. Blocking calls are simple to reason...
Bloom Filter
A space-efficient probabilistic data structure that tests whether an element is a member of a set. Can return false positives ('maybe in set') but never false...
Boot Process
Boot Process is a runtime or systems-level concept that affects execution behavior close to the machine. It is commonly used for performance-sensitive...
Bounded Buffer
Bounded Buffer is a runtime or systems-level concept that affects execution behavior close to the machine. It is commonly used for performance-sensitive...
Brag Document
A running document where you record your accomplishments, impact, and contributions throughout the year. Popularized by Julia Evans. Helps you remember...
Branch Prediction
A CPU optimization that guesses which way a conditional branch (if/else) will go and speculatively executes that path before the condition is evaluated. If the...
Buffer Pool
Buffer Pool is a runtime or systems-level concept that affects execution behavior close to the machine. It is commonly used for performance-sensitive software,...
Buffer Size
The amount of data a buffer is configured to hold before data is flushed, processed, or blocked. Buffer size affects throughput, latency, memory usage, and...
Build Cache
A cache of previously compiled artifacts, dependencies, and intermediate results that speeds up subsequent builds by skipping unchanged steps. Local caches...
Build Time
The amount of time required for a build process or a specific build stage to complete. Build time affects developer feedback loops, CI throughput, and release...
Bump Allocator
Bump Allocator is a runtime or systems-level concept that affects execution behavior close to the machine. It is commonly used for performance-sensitive...
Bun
A JavaScript runtime, bundler, and package manager written in Zig, designed to be a faster drop-in replacement for Node.js. Claims dramatically faster startup...
Bus Architecture
Bus Architecture is a runtime or systems-level concept that affects execution behavior close to the machine. It is commonly used for performance-sensitive...
Busy Wait
A wait implemented by repeatedly checking a condition instead of sleeping or yielding efficiently, often used casually to describe wasteful polling. In...
Bytecode Interpreter
Bytecode Interpreter is a runtime or systems-level concept that affects execution behavior close to the machine. It is commonly used for performance-sensitive...
C++
An extension of C that adds object-oriented programming, templates, and a vast standard library while maintaining low-level hardware control. C++ dominates...
C ABI
C ABI is the binary calling convention and data layout contract used by C-compatible code. It is commonly used for linking libraries across languages and...
Cache
Cache is a high-speed storage layer that keeps copies of frequently accessed data so future requests can be served faster than fetching from the original,...
Cache Aside
Cache Aside is a runtime or systems-level concept that affects execution behavior close to the machine. It is commonly used for performance-sensitive software,...
Cache Hit
Cache Hit is a runtime or systems-level concept that affects execution behavior close to the machine. It is commonly used for performance-sensitive software,...
Cache Key
The identifier used to look up a value in a cache, often derived from request parameters, user context, resource IDs, locale, or other inputs that affect the...
Cache Layer
A distinct caching tier inserted between a data source and its consumers to reduce latency, offload repeated work, or improve throughput. Cache layers may sit...
Cache Line
Cache Line is a runtime or systems-level concept that affects execution behavior close to the machine. It is commonly used for performance-sensitive software,...
Cache Management
The operational handling of cache population, invalidation, sizing, warm-up, eviction, observability, and failure behavior. Good cache management matters...
Cache Miss
Cache Miss is a runtime or systems-level concept that affects execution behavior close to the machine. It is commonly used for performance-sensitive software,...
Cache Policy
Cache Policy is a runtime or systems-level concept that affects execution behavior close to the machine. It is commonly used for performance-sensitive...
Cache Size
The amount of memory or storage allocated to a cache for holding entries. Cache size is a key tuning parameter because too little space causes churn and low...
Cache Stampede
A failure mode where a popular cache entry expires, causing many concurrent requests to simultaneously hit the database to regenerate it. The sudden load spike...
Cache Strategy
The overall approach used to decide what should be cached, where it should be cached, how long it should live, and how it should be invalidated. Cache strategy...
Cache Warming
Cache Warming is a runtime or systems-level concept that affects execution behavior close to the machine. It is commonly used for performance-sensitive...
Caching
Caching is the practice of storing frequently accessed data in a faster storage layer so subsequent requests skip expensive computations, database queries, or...
Caching Layer
A layer in the architecture dedicated to storing and serving cached data between producers and consumers. A caching layer often sits between the application...
Call Stack
Call Stack is a runtime or systems-level concept that affects execution behavior close to the machine. It is commonly used for performance-sensitive software,...
Capacity Planning
The process of estimating future resource needs such as CPU, memory, storage, throughput, and concurrency so systems can handle expected growth and peak...
CDN
A CDN, or Content Delivery Network, is a geographically distributed system of proxy servers and data centers that cache and deliver content from locations...
Channel Buffer
Channel Buffer is a runtime or systems-level concept that affects execution behavior close to the machine. It is commonly used for performance-sensitive...
Child Process
Child Process is a runtime or systems-level concept that affects execution behavior close to the machine. It is commonly used for performance-sensitive...
Chip Architecture
Chip Architecture is a runtime or systems-level concept that affects execution behavior close to the machine. It is commonly used for performance-sensitive...
Clock Speed
The frequency at which a processor's internal clock oscillates, measured in hertz (Hz), which determines the rate at which it can execute instructions — though...
Cloud CDN
A content delivery network service provided as part of a cloud platform to cache and serve content from edge locations closer to users. Cloud CDNs are commonly...
Code Cache
A cache used to store compiled, parsed, or otherwise optimized code representations so repeated execution can skip expensive preparation work. Code caches...
Code Performance
How efficiently code runs in terms of speed, memory use, throughput, latency, and resource consumption under realistic conditions. Code performance matters not...
Code Profile
A performance profile of code showing where time, memory, or other resources are being spent during execution. Code profiles help developers find hotspots,...
code splitting
A technique that breaks an application's JavaScript bundle into smaller chunks that are loaded on demand rather than all upfront. Code splitting ensures users...
Code Splitting
Breaking a JavaScript bundle into smaller chunks that are loaded on demand rather than all upfront. Route-based splitting loads code per page; component-based...
Cold Start
The latency penalty when a serverless function or container must be initialized from scratch — loading the runtime, dependencies, and application code before...
Compact
Describing code, data, or output that takes relatively little space while still preserving the needed meaning or functionality. In engineering discussions,...
Comptime
A Zig language keyword that forces an expression to be evaluated at compile time rather than runtime, enabling powerful metaprogramming and generic programming...
Concurrency
The ability of a system to handle multiple tasks that are in progress at the same time, though not necessarily executing simultaneously. Concurrency is about...
Concurrency Control
A Concurrency Control is a development concept used to shape how software is built, executed, or maintained in real systems. Developers rely on it to make code...
Concurrency Limit
A configured cap on how many tasks, requests, workers, or operations may run in parallel at once. Concurrency limits are used to protect downstream...
Concurrency Model
A Concurrency Model is a development concept used to shape how software is built, executed, or maintained in real systems. Developers rely on it to make code...
Concurrency Pattern
A Concurrency Pattern is a development concept used to shape how software is built, executed, or maintained in real systems. Developers rely on it to make code...
connection pool
A cache of reusable database connections maintained by an application or middleware (like PgBouncer) to avoid the overhead of establishing a new connection for...
Connection Pooling
Maintaining a cache of reusable database connections instead of opening a new connection for each request. Dramatically reduces latency (TCP handshakes, TLS...
Content Delivery
The process of distributing application assets, media, documents, or pages efficiently to users across different locations and devices. Content delivery often...
Context Switch
The process of saving one thread/process's state (registers, program counter, stack pointer) and loading another's so the CPU can switch between tasks. Costs...
Continuous Profiling
The ongoing collection of profiling data from running applications in production or near-production environments to understand where CPU time, memory, or other...
Core Web Vitals
A set of three user-centric performance metrics defined by Google that measure loading speed (Largest Contentful Paint), interactivity (Interaction to Next...
Core Web Vitals
Google's three key metrics for web page user experience: LCP (Largest Contentful Paint — loading speed), INP (Interaction to Next Paint — responsiveness), and...
Cost Per Acquisition
The average cost to acquire a customer or user through a specific campaign or channel, often used in performance marketing. It can be narrower and more...
Cost Per Click
The price paid for each click in a digital advertising campaign. CPC is one of the most common metrics for evaluating paid traffic efficiency.
Cost Per Lead
The average cost of generating one lead through a campaign or channel, usually before that lead becomes a qualified opportunity or customer. Teams use it to...
CPA
Usually shorthand for cost per acquisition in growth contexts, though it can also mean certified public accountant in finance contexts. In startup metrics...
CPC
Short for cost per click, the amount paid for each click in an advertising campaign. It is often analyzed alongside conversion rate to judge paid-channel...
CPL
Short for cost per lead, measuring the average acquisition cost of generating a lead. Teams track CPL to compare events, content, paid ads, and outbound...
CPU Cache
Small, extremely fast memory built into the CPU that stores copies of frequently accessed data. Organized in levels: L1 (fastest, smallest, ~32-64KB per core),...
Critical Rendering Path
The sequence of steps the browser takes to convert HTML, CSS, and JavaScript into pixels on screen: parse HTML → build DOM → parse CSS → build CSSOM → combine...
Cumulative Layout Shift
A Core Web Vital metric that measures visual stability — how much the page layout unexpectedly shifts during loading. Caused by images without dimensions,...
Custom Allocator
A specialized memory allocation strategy or implementation provided by an application instead of relying solely on the default allocator. Custom allocators are...
Debounce
A technique that delays execution of a function until a specified period of inactivity has passed. Each new trigger resets the timer. Commonly used for...
denormalization
The deliberate introduction of redundancy into a database schema to improve read performance, trading storage space and write complexity for faster queries....
Disk Cache
A Disk Cache is a development concept used to shape how software is built, executed, or maintained in real systems. Developers rely on it to make code easier...
DMA
Direct Memory Access — a hardware mechanism that allows peripherals to transfer data directly to or from main memory without involving the CPU for each byte,...
Dog and Pony Show
A presentation, demo, or managed display designed to impress rather than to represent everyday reality fully. In engineering slang, a dog and pony show is...
Dynamic Memory Allocation
A Dynamic Memory Allocation is a development concept used to shape how software is built, executed, or maintained in real systems. Developers rely on it to...
Edge Function
A serverless function that runs at CDN edge locations close to the user, rather than in a centralized cloud region. Executes in lightweight runtimes (V8...
Elysia
A TypeScript web framework optimized for Bun, focused on developer experience through end-to-end type safety and exceptional performance. Elysia uses Bun's...
ETag
An HTTP response header containing a unique identifier (hash or version) for a specific version of a resource. Clients can send If-None-Match with the ETag on...
Evil Regex
A regular expression so complex or poorly designed that it is unreadable, fragile, or prone to catastrophic backtracking and performance issues. The term often...
First Contentful Paint
A performance metric measuring the time from navigation to when the browser renders the first piece of DOM content (text, image, canvas). Signals to users that...
flame graph
A visualization of profiled software that displays the call stack hierarchy as stacked, color-coded rectangles where width represents time spent. Created by...
Flame Graph
A visualization of profiling data where the x-axis represents the proportion of time spent and the y-axis shows the call stack depth. Wider bars mean more time...
Flame Graph Culture
A culture that values evidence-driven performance debugging through profiling and visualization rather than guessing. The phrase is partly humorous, but it...
Flash Attention
An optimized attention implementation that reduces memory traffic and accelerates transformer training or inference. It influences how models are trained,...
FLOPS
Floating-point operations per second, a standard measure of numerical compute throughput often used in AI hardware discussions.
Frame Rate
The frequency at which consecutive images (frames) are rendered or displayed, measured in frames per second (FPS). 60 FPS is the standard target for smooth...
Garbage Collection Pause
A stop-the-world event where a garbage collector pauses all application threads to reclaim memory. Can range from microseconds (ZGC, Shenandoah) to seconds...
Goal Setting
The practice of defining goals clearly enough that people know what outcomes matter and can align work accordingly. In healthy organizations, goal setting...
GPU Utilization
A measure of how fully a GPU is being used for useful computation over time. Low GPU utilization can indicate bottlenecks in data loading, batching, model...
Hash Table
A data structure that maps keys to values using a hash function to compute an index into an array of buckets. Provides average O(1) lookup, insertion, and...
Heap Memory
A Heap Memory is a development concept used to shape how software is built, executed, or maintained in real systems. Developers rely on it to make code easier...
Heavyweight
Describing a tool, framework, process, or system that is large, complex, or resource-intensive relative to the problem at hand. It is often used critically...
Hot Loop
A tight frequently executed loop that matters a lot for performance. In engineering slang, hot loops are where tiny inefficiencies can become big costs because...
Hot Path
The most frequently executed code path in a system — the sequence of operations that handles the majority of traffic or computation. Optimizing the hot path...
HTTP/2
The second major version of HTTP, introducing multiplexed streams (multiple requests over one TCP connection), header compression (HPACK), server push, and...
HTTP/3
The third major version of HTTP, built on QUIC (UDP) instead of TCP. Eliminates TCP's head-of-line blocking — a lost packet in one stream doesn't stall other...
Hydration
The process of attaching JavaScript event handlers and state to server-rendered HTML on the client side. After SSR delivers static markup for fast initial...
index
A data structure (typically a B-tree or hash table) that speeds up data retrieval operations on a database table at the cost of additional storage and slower...
Index Scan vs Seq Scan
Two fundamental data access methods in databases. An index scan uses a B+ tree index to jump directly to matching rows — fast for selective queries. A...
Inference
The process of running a trained machine learning model on new data to generate predictions or outputs. Unlike training (which learns patterns), inference...
Inference
The process of using a trained model to make predictions or generate outputs from new inputs. As opposed to training (which teaches the model), inference is...
Inference Speed
How quickly a model can process inputs and generate outputs during inference, often measured in latency or tokens per second. Inference speed influences user...
Inline Assembly
An Inline Assembly is a development concept used to shape how software is built, executed, or maintained in real systems. Developers rely on it to make code...
Inline Cache
An Inline Cache is a development concept used to shape how software is built, executed, or maintained in real systems. Developers rely on it to make code...
Intersection Observer
A browser API that asynchronously observes when an element enters or exits the viewport (or another element). Replaces expensive scroll event listeners for...
io_uring
A Linux asynchronous I/O interface (since kernel 5.1) that uses shared ring buffers between user space and kernel to submit and complete I/O operations without...
Islands Architecture
A web architecture pattern where most of the page is static HTML, with small 'islands' of interactive components that are independently hydrated. Reduces...
JAX
A numerical computing library for high-performance machine learning with automatic differentiation and XLA compilation. It influences how models are trained,...
Jemalloc
A general-purpose memory allocator designed to reduce fragmentation and scale well under multithreaded workloads. It is widely used in high-performance systems...
JIT
Just-In-Time compilation — a technique where bytecode is compiled to native machine code at runtime rather than ahead of time, allowing optimizations based on...
JIT Compilation
Just-In-Time compilation — translating bytecode or interpreted code to native machine code at runtime, right before execution. Combines the portability of...
KV Cache
A transformer inference cache that stores previously computed key and value tensors for faster decoding. It influences how models are trained, evaluated, or...
L1 Cache
The smallest and fastest cache memory embedded directly on each CPU core, typically split into separate instruction (L1i) and data (L1d) caches of 32–64 KB...
L2 Cache
A mid-level CPU cache, typically 256 KB to 1 MB per core, that sits between L1 and L3 in the memory hierarchy with access latencies of roughly 10–20 cycles —...
L3 Cache
The largest on-chip CPU cache, typically 8–64 MB shared across all cores, with access latencies of 30–50 cycles. It acts as the last line of defense before the...
Largest Contentful Paint
A Core Web Vital metric measuring when the largest content element (hero image, heading, video poster) becomes visible in the viewport. Best proxy for...
Latency
The time delay between a request and its response. In networking, it's the round-trip time for a packet. In user experience, it's the time between click and...
Latency Budget
The maximum allowable end-to-end response time for a request, broken down and allocated across each component in the processing chain — such as DNS, TLS...
Latency vs Throughput
Two often-confused performance metrics: latency is the time for a single operation to complete (measured in ms), while throughput is the number of operations...
Lazy Evaluation
An evaluation strategy that delays the computation of an expression until its value is actually needed, avoiding unnecessary work and enabling infinite data...
Lazy Initialization
A creational pattern in which an object or resource is not allocated until the first time it is actually needed, deferring the cost of construction and...
lazy loading
A technique that defers the loading of non-critical resources (images, scripts, components) until they're needed, typically when they enter or approach the...
Lighthouse Score Obsession
An excessive focus on improving web audit scores, sometimes at the expense of real user outcomes or broader product priorities.
Line Profiler
A profiling tool that measures execution time or hit count for each individual line of source code, rather than at the function level. Python's `line_profiler`...
Live Coding
The practice of writing or modifying code in real time in front of an audience — during a livestream, conference talk, interview, or music performance. In the...
LLM Throughput
The rate at which a large language model system can process requests or generate tokens over time. Throughput matters for capacity planning, queueing behavior,...
Load Shedding
A resilience strategy where a system intentionally drops or rejects excess requests during overload to protect core functionality. Unlike rate limiting...
Load Testing
Simulating expected or peak traffic patterns against a system to verify performance under load. Measures response times, throughput, error rates, and resource...
Local Cache
An in-process cache that stores frequently accessed data in the application's own memory space, avoiding network round-trips to an external cache like Redis....
Lock Contention
A performance bottleneck that occurs when multiple threads frequently compete for the same lock, causing them to block and wait instead of doing useful work....
Lock-Free Data Structure
A concurrent data structure that guarantees system-wide progress without using locks -- at least one thread makes progress in a finite number of steps...
Lookup Table
A precomputed array or hash map that replaces expensive runtime calculations with direct indexed access, trading memory for speed. Common uses include CRC...
Loop Unrolling
A compiler or manual optimization that replicates the body of a loop multiple times per iteration, reducing the overhead of branch instructions, loop counter...
LRU Cache
Least Recently Used cache — a fixed-size cache that evicts the least recently accessed item when full. Implemented with a hash map (O(1) lookup) and a...
materialized view
A database object that stores the precomputed results of a query physically on disk, unlike a regular view which re-executes the query each time. Materialized...
Memcached
A high-performance, distributed, in-memory key-value store designed for caching database query results, session data, and API responses. It uses a simple...
Memo
Short for memoization — caching the result of a function call so that subsequent calls with the same arguments return the cached value instead of recomputing....
Memory Alignment
The requirement or practice of placing data in memory at addresses that are multiples of the data type's size (e.g., a 4-byte int at an address divisible by...
Memory Allocation
Memory Allocation is the process of reserving a portion of computer memory for use by a program. Static allocation happens at compile time for global variables...
Memory Arena
A region-based memory allocation strategy where all allocations come from a single contiguous block and are freed together in bulk, eliminating per-object free...
Memory Budget
A hard or soft limit on how much RAM a component, subsystem, or entire application is allowed to consume. Memory budgets are enforced through monitoring and...
Memory Leak
A bug where a program allocates memory but never releases it, causing memory usage to grow continuously until the process crashes or the OOM killer intervenes....
Memory-Mapped File
A technique that maps a file directly into a process's virtual address space, allowing the program to read/write the file as if it were memory. The OS handles...
Memory Pool
A pre-allocated block of memory divided into fixed-size chunks that can be quickly allocated and deallocated without calling the system allocator. Eliminates...
Memory Pressure
A condition in which available physical memory is nearly exhausted, forcing the OS to aggressively page, compress, or kill processes to reclaim RAM....
Memory Profile
A snapshot or time-series recording of a program's memory usage, showing allocation sizes, frequencies, call sites, and object lifetimes. Tools like Valgrind's...
MessagePack
A binary serialization format that's like JSON but faster and more compact. Supports the same data types (maps, arrays, strings, numbers, booleans, null) but...
Mex
MATLAB Executable — a compiled C, C++, or Fortran function callable from within MATLAB, used to speed up performance-critical code or interface with external...
Micro-Optimization
A very small, localized performance improvement that saves tiny amounts of time, memory, or compute, often at the cost of readability or maintainability....
Micro Optimization Slang
Informal language around tiny performance tweaks, often used when debating whether the gains are real or worth the complexity. In engineering slang,...
Minification
The process of removing whitespace, comments, and renaming variables to shorter identifiers in JavaScript, CSS, or HTML source code to reduce file size and...
Model Performance
How well a model performs on the dimensions that matter for a task, such as accuracy, speed, cost, robustness, or user satisfaction. Model performance should...
Model Speed
The speed at which a model can return results, often measured by latency or tokens per second. Model speed matters especially in interactive products where...
Model Throughput
The rate at which a model-serving system can process requests or generate tokens over time.
MTU
Maximum Transmission Unit — the largest packet size (in bytes) that a network interface can transmit without fragmentation. Ethernet's standard MTU is 1500...
Multi-Threading
A concurrency model in which a single process spawns multiple threads of execution that share the same memory space, enabling parallel work on multi-core CPUs...
N+1 Query Problem
A performance anti-pattern where code executes 1 query to fetch a list of N items, then N additional queries to fetch related data for each item. Common with...
Native
Code compiled directly to the machine instructions of a specific processor or platform, running without an interpreter or virtual machine layer.
Non-Blocking
Describes an operation or API call that returns immediately rather than waiting for completion, allowing the calling thread to continue executing other work....
N Plus One Problem
A performance issue where one initial query triggers many additional dependent queries, often one per item, causing avoidable load and latency. In engineering...
NUMA
Non-Uniform Memory Access — a memory architecture in multi-socket servers where each CPU has fast access to its local memory and slower access to remote memory...
NVMe
Non-Volatile Memory Express — a storage protocol designed specifically for SSDs, replacing the SATA/AHCI protocols that were designed for spinning hard drives....
Object Pool
A creational design pattern that maintains a set of pre-initialized, reusable objects — such as database connections or threads — rather than allocating and...
OCSP Stapling
OCSP Stapling is a TLS optimization where the server includes a recent signed OCSP response during the handshake. Security teams use it to enforce trust,...
On Fire
Doing extremely well or going extremely badly, depending on context and tone. In tech it can praise a high-performing team or describe a system that is...
On the Metal
Running directly on physical hardware rather than inside a virtual machine, container abstraction, or managed platform layer. It often implies tighter control...
Optimization
The process of modifying code, queries, algorithms, or system configuration to improve a measurable performance metric — such as execution time, memory usage,...
Order of Magnitude
A factor-of-ten difference used to express rough scale comparisons. In engineering estimation, saying something is 'an order of magnitude faster' means roughly...
Overclocking
The practice of running a processor, GPU, or memory at a clock frequency higher than its rated specification to achieve greater performance, at the cost of...
Overhead
The additional cost, time, or complexity required to support an operation beyond the core work itself. Overhead can be technical, organizational, or...
Over-Optimize
To spend too much effort improving one aspect of a system beyond what the actual constraints justify. In engineering slang, over-optimizing usually means...
Page Cache
A region of main memory managed by the OS kernel that caches recently read file pages (typically 4 KB blocks) from disk, so that subsequent reads of the same...
Page Fault
An interrupt triggered when a program accesses a virtual memory page that isn't currently mapped to physical RAM. Minor page faults load the page from the page...
Pagination
Breaking large result sets into smaller pages returned one at a time. Offset-based (page=3&per_page=20) is simple but slow for deep pages. Cursor-based...
Parallel Decoding
Methods that generate or verify multiple output tokens in parallel to reduce latency compared with strictly sequential decoding.
Parallel Generation
Generation strategies or system designs that produce multiple outputs, branches, or partial computations in parallel rather than strictly one sequence at a...
Parallelism
The simultaneous execution of multiple computations, typically on separate CPU cores or machines. Unlike concurrency (which is about structure), parallelism is...
Parallel Processing
The simultaneous execution of multiple computations across separate CPU cores or processors to reduce wall-clock time. Unlike concurrency (which interleaves...
Partial Encryption
A technique in which only part of a file, dataset, or storage volume is encrypted instead of the whole thing, often to reduce processing time while still...
Partitioning
Splitting a large table into smaller, more manageable pieces (partitions) based on a key — typically date (range partitioning), category (list partitioning),...
Perf Tool
The Linux kernel's built-in performance analysis tool (perf) that uses hardware performance counters and software tracepoints to profile CPU cycles, cache...
Phantom Load
System load that is real and costly but not obvious from the most visible usage patterns, often coming from retries, background work, duplicated events, or...
Pipeline Parallelism
Pipeline Parallelism is a systems strategy for splitting model computation or parameters across devices. It is commonly used for training and serving workloads...
Pool
A cache of pre-initialized, reusable resources — such as database connections, threads, or objects — that can be checked out when needed and returned after...
Prefetch
To speculatively load data, assets, or DNS resolutions before they are explicitly needed, anticipating that the user or program will request them soon. In...
Preload
To force the browser or runtime to fetch a critical resource at high priority early in the page lifecycle, before the parser would naturally discover it. In...
Premature Optimization
The act of optimizing code for performance before you know where the actual bottleneck is. Knuth called it 'the root of all evil,' and he wasn't wrong —...
Profile
To measure where a program spends its time and memory, identifying bottlenecks. Profilers instrument code to collect metrics — CPU time, memory allocations,...
Profile Guided Optimization
A two-pass compiler optimization technique where the program is first compiled with instrumentation, then run against representative workloads to collect...
Profiler
A tool that measures where a program spends its time and memory, typically by sampling the call stack at intervals or instrumenting function entry/exit....
Profiling
Measuring the resource consumption (CPU time, memory, I/O, function call counts) of a program to identify performance bottlenecks. CPU profilers show which...
Progressive Loading
A technique for rendering content incrementally as it becomes available — showing skeleton screens, low-resolution placeholders, or partial data — rather than...
Protocol Buffers
A language-neutral, platform-neutral binary serialization format developed by Google. Protobuf schemas (.proto files) define message structures and generate...
Pure Component
A React component that renders the same output given the same props and state, and implements a shallow comparison in shouldComponentUpdate (via...
Query Cache
A mechanism that stores the result set of a previously executed query keyed by the query text (and sometimes its parameters), returning cached results on...
Query Optimizer
The database subsystem that transforms a parsed SQL query into an efficient execution plan by evaluating strategies like join ordering, index selection, and...
Query Plan
The execution strategy chosen by the database's query optimizer for a SQL statement. Shows which indexes are used, join order and method (nested loop, hash,...
QUIC
A transport protocol built on UDP that provides the reliability of TCP with faster connection setup, built-in encryption (TLS 1.3), and multiplexed streams...
Qwik
A JavaScript framework built around the concept of 'resumability' rather than hydration. Created by Misko Hevery (creator of AngularJS), Qwik serializes the...
Random Access
The ability to read or write any element in a data structure in O(1) time by its index or address, without traversing preceding elements. Arrays and RAM...
Rate Limiter
A mechanism that controls the rate of requests a client can make to a service within a time window. Common algorithms include token bucket, sliding window, and...
Redis Pipeline
A technique for sending multiple Redis commands to the server in a single batch without waiting for each individual reply, dramatically reducing round-trip...
Rehydration
The client-side process of attaching JavaScript event handlers and component state to server-rendered HTML, making a static page interactive without...
Render Cycle
One complete pass through a UI framework's rendering process — typically triggered by a state change, encompassing the evaluation of component functions or...
Resource Pool
A cache of pre-initialized, reusable resources — typically database connections, threads, or network sockets — that are borrowed by consumers and returned...
Responsive Images
HTML techniques for serving appropriately sized images based on the viewer's screen size, pixel density, and supported formats. Uses srcset (multiple...
Ring Buffer
A fixed-size buffer that wraps around — when the write pointer reaches the end, it wraps to the beginning, overwriting the oldest data. Provides O(1)...
Rspack
A Rust-based JavaScript bundler designed as a drop-in replacement for Webpack, offering 5-10x faster build times while maintaining compatibility with the...
Runaway Process
A process that consumes excessive CPU, memory, disk, or other resources because it is stuck, looping, leaking, or otherwise behaving uncontrollably. In...
Running Hot
Operating close to system, workload, or human limits, often with little spare capacity left. It can describe servers near resource saturation or teams under...
Scale Up
Vertical scaling — increasing system capacity by upgrading the hardware resources (CPU, RAM, disk, network) of an existing machine or instance. Simpler than...
Scroll Event
A browser DOM event fired when the user scrolls an element or the page, notorious for triggering at extremely high frequency and requiring throttling or...
Server Push
A mechanism in which the server proactively sends resources to the client before the client requests them — originally an HTTP/2 feature (PUSH_PROMISE) for...
Server-Side Rendering
The technique of generating a page's full HTML on the server for each request (or at build time in the static variant), sending a complete document to the...
Short Circuit
An evaluation strategy in which logical operators (`&&`, `||`) skip evaluating the right-hand operand when the result is already determined by the left —...
SIMD
Single Instruction, Multiple Data — a CPU capability that applies one operation to multiple data elements simultaneously. A single SIMD instruction can add 8...
Skeleton Screen
A loading state that shows the page's layout structure with placeholder shapes (gray boxes, lines) instead of a spinner or blank page. Creates the perception...
Socket Pool
A cache of pre-established, reusable network socket connections maintained by an HTTP client or application to avoid the overhead of repeated TCP handshakes...
SSR
SSR, or Server-Side Rendering, is a web development technique where HTML pages are generated on the server for each request rather than being constructed in...
Stack Allocation
The practice of allocating memory on the call stack by simply adjusting the stack pointer, making allocation and deallocation essentially free (a single...
Stack Memory
A region of memory allocated per thread that grows and shrinks in LIFO (last-in, first-out) order as functions are called and return. It is managed...
Stale-While-Revalidate
A caching strategy (HTTP header and React Query/SWR pattern) that immediately returns stale cached data while simultaneously fetching a fresh version in the...
Static Dispatch
A method call resolution strategy where the compiler determines the exact function to invoke at compile time, typically through monomorphization of generics,...
StormForge
A tooling brand associated with optimizing Kubernetes resource settings and performance behavior.
Streaming Inference
Streaming Inference is the phase where a trained model processes new inputs to produce predictions or generations. It is commonly used for production APIs,...
String Builder
A mutable buffer class (e.g., Java's `StringBuilder`, C#'s `StringBuilder`, Go's `strings.Builder`) that efficiently constructs strings by appending segments...
Superscalar
A CPU architecture that can issue and execute multiple instructions per clock cycle by employing multiple execution units (ALUs, FPUs, load/store units) and an...
Swap
Disk space used as virtual memory when physical RAM is full. The OS 'swaps' memory pages to disk — functional but orders of magnitude slower. If your server is...
Tail Call Optimization
A compiler optimization that reuses the current stack frame for a recursive call that is the last operation in a function, preventing stack overflow on deep...
Tail Call Optimization
A compiler optimization that reuses the current function's stack frame for a recursive call in tail position, converting recursion into iteration. Prevents...
tail latency
The response time experienced by the slowest requests — typically measured at the 99th or 99.9th percentile (p99/p999). In distributed systems, tail latency is...
Tail Latency
The response time experienced by the slowest requests — typically measured at p99 (99th percentile) or p99.9. In distributed systems, tail latency compounds:...
Tank
To crash, fail badly, or drop sharply in quality or performance. In engineering slang, things tank when they fall off a cliff instead of merely drifting...
Task Parallelism
A form of parallelism where different tasks or functions are distributed across multiple processors, as opposed to data parallelism where the same operation is...
Tensor
Tensor is an AI or ML concept used to represent, train, evaluate, or deploy learned systems. It is commonly used for building production models and research...
Tensor Core
Tensor Core is an AI or ML concept used to represent, train, evaluate, or deploy learned systems. It is commonly used for building production models and...
TensorFlow
TensorFlow is a machine learning framework for defining, training, and deploying numerical computation graphs. It is commonly used for deep learning research,...
Tensor Parallelism
Tensor Parallelism is a systems strategy for splitting model computation or parameters across devices. It is commonly used for training and serving workloads...
TensorRT
TensorRT is an NVIDIA inference optimization runtime for compiled neural networks. It is commonly used for serving GPU models with lower latency and higher...
Thread
The smallest unit of execution that an operating system can schedule. Threads within a process share memory space, enabling parallel execution but requiring...
Thread Pool
A collection of pre-created worker threads that process tasks from a shared queue. Eliminates the overhead of creating/destroying threads per request (which...
Thread Starvation
A condition where all threads in a pool are occupied by long-running or blocked tasks, leaving no threads available to process incoming work — causing requests...
Throttle
A technique that limits function execution to at most once per specified time interval, regardless of how many times it's triggered. Unlike debounce (which...
Throughput
The amount of work a system can handle per unit of time — requests per second, transactions per minute, data processed per hour. Not the same as latency: you...
Throughput AI
The rate at which an AI system can process requests, tasks, or tokens over time under real operating conditions. Throughput matters for capacity planning and...
Thundering Herd
A concurrency problem where a large number of processes or clients are simultaneously woken up or triggered by a single event, causing a massive spike in...
Time Complexity
A measure of how an algorithm's running time grows as the input size increases, expressed in Big-O notation — for example, O(n) means linear growth, O(n²)...
Time to First Byte Culture
A performance-minded culture where teams care deeply about responsiveness metrics such as time to first byte and treat latency as a product and infrastructure...
Token Per Second
A throughput measure showing how many tokens a model can generate each second during inference.
tokens per second
The standard throughput metric for language model inference, measuring how many tokens a model can generate per second. Higher TPS enables more responsive user...
TPU
TPU is a specialized accelerator designed for large-scale tensor operations. It is commonly used for training and inference workloads with dense linear...
Trace
A record of the execution path through a system, capturing the sequence of function calls, service hops, or operations. In distributed systems, a trace follows...
tree shaking
A build optimization that eliminates unused code (dead code) from JavaScript bundles by analyzing the dependency graph of ES module imports. Named...
Tree Shaking
A dead code elimination technique that removes unused exports from JavaScript bundles during the build step. Analyzes ES module import/export relationships to...
Triton Kernel
A custom GPU compute kernel written with Triton to optimize operations common in machine learning workloads.
Turbolinks
The predecessor to Turbo Drive, Turbolinks was a Ruby on Rails library that sped up page navigation by intercepting link clicks, fetching the new page via...
Turbopack
A Rust-based JavaScript bundler developed by Vercel as the successor to Webpack, designed to be dramatically faster through incremental computation and native...
UDP
UDP, or User Datagram Protocol, is a connectionless transport protocol that sends packets (called datagrams) without establishing a connection or guaranteeing...
Unrolling
A compiler or manual optimization technique that replaces a loop with repeated copies of its body, reducing the overhead of branch instructions and...
USE Method
A methodology for analyzing infrastructure performance: Utilization (how busy the resource is), Saturation (how much work is queued), and Errors (error...
V8 Isolate
A lightweight, sandboxed instance of the V8 JavaScript engine that provides memory and execution isolation without the overhead of a full container or VM....
Vectorization
Transforming scalar operations into SIMD (Single Instruction, Multiple Data) operations that process multiple data elements in a single CPU instruction. Can...
Virtual DOM
A lightweight in-memory representation of the real DOM used by frameworks like React to optimize UI updates. When state changes, a new virtual DOM tree is...
Wasm
Short for WebAssembly — a portable, low-level binary instruction format designed as a compilation target for languages like C, C++, Rust, and Go, executing in...
WebAssembly
A binary instruction format designed as a portable compilation target for programming languages, enabling near-native performance in web browsers. WebAssembly...
Web Vitals
A set of user-centric performance metrics defined by Google — most notably the Core Web Vitals: Largest Contentful Paint (LCP), Interaction to Next Paint...
web worker
A mechanism for running JavaScript in a background thread separate from the main page thread, preventing CPU-intensive tasks from blocking the user interface....
Web Worker
Web Worker is a JavaScript API that enables running scripts in background threads separate from the main execution thread of a web page, allowing...
Worker Pool
A collection of pre-spawned threads or processes that wait for tasks on a shared queue, amortizing the cost of thread/process creation across many jobs and...
Work Stealing
A scheduling strategy where idle threads 'steal' tasks from busy threads' queues, achieving automatic load balancing without central coordination. Each thread...
Write Back
A caching strategy where writes update only the cache initially, marking the cache line as dirty, and the modified data is flushed to the backing store (main...
Write Through
A caching strategy in which every write operation updates both the cache and the underlying backing store synchronously, guaranteeing that the cache and the...
X Factor Slang
Informal use of x factor for the hard-to-quantify quality that makes a person, tool, or plan unusually effective. In engineering slang, the x factor often...
Zap
A high-performance, structured logging library for Go developed by Uber, designed to minimize memory allocations and CPU overhead by using a strongly-typed,...

Related Topics