This article breaks down the core concepts of Kubernetes—including its master‑worker architecture, pod scheduling, etcd storage, service exposure, scaling mechanisms, and controller interactions—through a series of clear questions and illustrated answers to help beginners grasp the platform’s complexity.
This article traces ZuanZuan's log‑collection journey from the bare‑metal era through containerization to a cloud‑native solution, detailing the challenges, custom developments like log‑pilot, ByteCompass, and fb‑advisor, and the performance gains achieved with each transition.
The article surveys the 2023 cloud‑native landscape, highlighting the rise of cloud‑based IDEs, the mainstreaming of FinOps and GreenOps, the ubiquity of open‑source SBOMs, the maturation of GitOps and OpenTelemetry, the growing impact of WebAssembly, and several related forecasts for the industry.
Spring Cloud 2022, built on Spring Framework 6.x and Spring Boot 3.x, introduces critical bug fixes—including memory leaks and Kotlin DSL issues—and requires Java 17, with updated dependency management and new features across Gateway, Commons, and Kubernetes modules.
This article compares Prometheus and Zabbix, detailing their histories, architectures, performance, community support, and suitability for different environments, and concludes with guidance on choosing the right monitoring solution for physical servers, cloud-native deployments, and large‑scale container clusters.
This article reviews the evolution of file storage, outlines the challenges of the cloud era, and details JuiceFS's design philosophy, architecture, key capabilities, and real‑world use cases such as Kubernetes, AI, big‑data analytics, and NAS migration to the cloud.
Since JDK 9, Java has accelerated its release cadence and added a suite of cloud‑native capabilities—such as container‑aware metrics, single‑file execution, refined JVM options, fast‑fail memory controls, class‑data sharing, compact strings, active‑processor detection, and Unix‑domain sockets—to better serve modern containerized workloads.
In this comprehensive interview, ByteDance senior front‑end engineer Huang Jian shares the 2023 outlook for front‑end developers, covering cloud‑native integration, serverless adoption, micro‑frontends, AI influence, WebAssembly, career growth, and the evolving skill set required to stay competitive.
This guide provides a comprehensive overview of kubectl, covering its syntax, resource types, common flags, output formats, and dozens of practical command examples for creating, listing, describing, and deleting Kubernetes objects efficiently.
This article describes a cloud‑native CI/CD pipeline built on Kubernetes that leverages Tekton for continuous integration, Argo CD for continuous deployment, and Redmine for project management, detailing system design, task flow, usage steps, and achieved operational benefits.
Sealer 0.9 introduces a cluster‑centric model for building, sharing and running distributed applications, adds Sealer Image types, enriches Kubefile with LABEL, KUBEVERSION, APP, CNI, CSI, APPCMDS and LAUNCH directives, and outlines future roadmap and community resources.
This article provides a concise yet comprehensive overview of Kubernetes, covering its core architecture, the lifecycle of a Deployment, the role of Pods and Services, and how networking and service discovery are implemented within a cloud‑native environment.
Tencent Cloud helped the viral game 《羊了个羊》 scale from 5,000 QPS to support over 100 million daily active users in a week by using serverless Kubernetes auto‑scaling, real‑time logging, WAF/Anti‑DDoS protection, CDN, and read‑write separation with Redis, achieving high performance, availability, and scalability.
The article reproduces the AttachNotSupportedException when Java runs as PID 1 in a container, explains the JVM attach workflow and LinuxThread model, and provides practical workarounds such as creating .attach_pid files, enabling StartAttachListener, or using newer JDK images.
This article explains why Java applications running in Kubernetes can encounter Out‑Of‑Memory errors despite heap usage staying under the Xmx limit, by examining container resource limits, JVM memory models, cgroup behavior, and provides practical configuration recommendations to prevent OOM.
This article provides a comprehensive overview of Nacos, covering its origins, evolution of the configuration center, typical use cases, step‑by‑step integration methods, troubleshooting tips, high‑availability mechanisms, commercial MSE advantages, migration guides, and the upcoming features in Nacos 3.0.
This guide explains how to use Alibaba Cloud EventBridge to create and manage RocketMQ 4.0, 5.0, and self‑hosted Apache RocketMQ instances, covering architecture, API parameters, VPC and security settings, and step‑by‑step procedures for event‑driven integration and message routing.
This article explains what eBPF is, why it matters for cloud‑native architectures, its key components and use‑cases in networking, observability and security, and explores current market momentum and commercialization models for leveraging eBPF in modern infrastructure.
This article explores how YouTube handles massive video uploads and billions of daily views by detailing its backend micro‑services, MySQL‑based Vitess clustering, sharding, replication, disaster recovery, cloud‑native deployment, CDN strategy, and large‑scale storage architecture.
This article details the evolution of Ctrip's log infrastructure, describing the shift from fragmented departmental logging to a unified Elasticsearch-based platform, the migration to ClickHouse for cost‑effective, high‑performance storage, and the subsequent Log 3.0 redesign that leverages Kubernetes, sharding, and a unified query governance layer to handle petabyte‑scale data.
This article compares the open‑source API gateways APISIX and KONG across community activity, plugin richness, performance under various workloads, secondary development flexibility, and their suitability for cloud‑native environments such as Kubernetes ingress and function compute integration.
vivo’s Internet Storage team details how it operates RocketMQ for low‑latency online services and Kafka for massive big‑data pipelines, outlines resource isolation, traffic balancing, intelligent throttling, and governance practices, and describes its migration from RabbitMQ and planned shift from Kafka to cloud‑native Pulsar.
This article explains the concept of transient clusters and the Stateless architecture in Volcano Engine's EMR platform, compares Stateless with traditional Stateful approaches, outlines its evolution, core components, elastic scaling features, and the business value of cost‑effective, on‑demand big‑data processing.
In a recent interview, the head of the System Operations Department at China Merchants Securities explains how their next‑generation core trading system, built on a distributed micro‑service architecture with open‑source components and cloud‑native tools, achieved Level 2 technical‑operation DevOps certification, detailing the challenges, improvements, and future plans for digital transformation.
This article provides a comprehensive comparison between Zabbix and Prometheus, covering their architecture, data collection, storage, querying, visualization, and alerting capabilities to help enterprises choose the most suitable monitoring solution for their needs.
SREWorks v1.4 introduces a Monorepo‑based frontend architecture, upgrades build tools to Webpack 5 and Rollup, adds an offline installation baseline with pre‑packaged images, and provides detailed Helm commands for upgrading and deploying the platform in cloud‑native environments.
This guide explains why MySQL monitoring is critical, defines five key metric dimensions, shows how to collect them with Prometheus and the MySQL Exporter, provides ready‑to‑use alert rules, and walks through the full setup and dashboard creation on Alibaba Cloud.
This guide explains how to route outbound container traffic through a centralized external gateway using Kube-OVN by defining a Subnet resource with specific routing and policy settings, and clarifies each required field for proper configuration.
This article details Zhaozhuan's three‑generation test environment governance evolution—starting with physical isolation, moving to automatic IP‑label traffic routing, and finally manual tag‑based routing—highlighting architectural changes, deployment processes, advantages, drawbacks, and supporting tools such as distributed tracing and debugging utilities.
This article examines YouTube’s massive backend infrastructure, detailing how the platform stores billions of videos using MySQL, Vitess for horizontal scaling, sharding, replication, disaster management, cloud‑native deployment on Kubernetes, and the supporting storage systems such as GFS and BigTable.
This article describes how Zhaozhuan’s testing environment evolved through three versions—physical isolation, automatic‑IP‑tag routing, and manual‑tag routing—detailing the architectural background, implementation principles, advantages, drawbacks, and supporting tools that dramatically reduced deployment time and resource consumption while introducing new operational challenges.
The article details the design, core features, and real‑world performance of the DragonF MPP DB, a cloud‑native, compute‑storage‑separated database that overcomes traditional MPP limitations, supports millions of daily jobs, and outlines its future roadmap for ultra‑large‑scale data platforms.
This article explains how Alibaba Cloud's Serverless Application Engine (SAE) leverages the open‑source KubeVela Workflow to overcome operational, scaling, and integration challenges, detailing the workflow design, step definitions, and three real‑world use cases that illustrate automated ops, release optimization, and rapid feature rollout.
This article explains how to design and implement a Prometheus‑based monitoring solution for big‑data components running on Kubernetes, covering metric exposure methods, scrape configurations, alerting architecture, exporter development, and practical code examples for a production‑ready setup.
This article reviews the evolution of file storage, outlines challenges of cloud‑native data management, and details JuiceFS’s cloud‑native design philosophy, architecture, and key use cases such as Kubernetes, AI, and big‑data workloads.
China Agricultural Bank’s distributed core customer information project passed the Level‑3 DevOps Application Design assessment, showcasing a cloud‑native micro‑service architecture, comprehensive DevOps practices, and measurable improvements in scalability, observability, and security that set a new industry benchmark.
This article explains how to define Kubernetes applications with YAML, compares it to JSON and INI, and outlines three validation levels—structural, semantic, and security—while recommending tools and shift‑left practices to ensure reliable and secure deployments.
This article details Taobao’s serverless architecture upgrade for its homepage, covering the background of the Cloud Native 2.0 initiative, the three‑pronged system transformation, risk analysis, and a comprehensive quality‑assurance plan that includes pre‑release verification, traffic switching, and stress testing.
This article reviews the evolution of message queues, explains why Pulsar was adopted at Tencent, details the challenges of open‑source solutions, and presents the architectural design, performance and observability optimizations, plus a real‑world case study and future roadmap.
The author recounts a sudden overload caused by copying hundreds of gigabytes of small files to an Alibaba Cloud NAS, which crashed the master node of a Kubernetes cluster, leading to Docker failures, and describes step‑by‑step troubleshooting, configuration changes, and lessons learned about backups, cautious operations, and calm analysis.
This article explains how NetEase Cloud Music tackled the tight coupling between frontend UI and backend services by introducing a Backend‑For‑Frontend layer built on GraphQL, detailing the architectural choices, low‑code editor, custom directives, deployment pipeline, and the resulting improvements in scalability and developer productivity.
Apache RocketMQ 5.0 introduces a stateless proxy, gRPC‑based multi‑language SDK with immutable APIs, enhanced error handling, SimpleConsumer, and pop consumption, and the article outlines the technical improvements, roadmap, and step‑by‑step guide for developers to set up the environment, understand the new protocols, and contribute language client implementations.
This article explains why Java Agent injected via environment variables loads in every JVM inside a container, how that impacts JDK tools like java‑version and jstack, and presents a Dragonwell‑based solution using custom environment variables to limit agent loading only to business Java processes.
This article explains how Bilibili’s large‑scale Kubernetes cloud platform reduces costs and improves machine utilization by applying co‑location (mixed‑tenant) techniques, including resource‑aware scheduling, dynamic isolation, and a dedicated management console across online, offline, and idle‑machine scenarios.
This article walks through the complete lifecycle of a Kubernetes deployment, explaining how a manual upgrade request triggers API calls, creates Deployments, ReplicaSets, Pods, and how the scheduler, kubelet, and Docker work together, while also covering concepts like containers, labels, replication controllers, deployments, and autoscaling mechanisms.
This guide explains the underlying snapshot and transaction‑log storage of ZooKeeper, outlines three key scenarios for using MSE’s import‑export feature, and provides step‑by‑step instructions with code snippets to back up, migrate, or analyze ZooKeeper clusters efficiently.
The article explains Kubernetes’ built‑in Horizontal Pod Autoscaler, then details the custom Water Pod Autoscaler (WPA) that extends HPA with dual‑signal (load and SOA registration) detection, dual‑threshold scaling, noise filtering, configurable cooldown, frequency limits, tolerance buffers, and integrated alerting for reliable elastic scaling.
This article walks through the entire lifecycle of deploying an application on Kubernetes, explaining how Docker containers differ from virtual machines, the role of Pods, ReplicationControllers, Deployments, and how automatic scaling with HPA and VPA keeps services reliable and efficient.
The article outlines Ctrip’s microservice journey from its 2013 inception, detailing the evolution of its frameworks, the complexities of operating multiple stacks, the challenges faced, and the design of a progressive cloud‑native service‑mesh architecture built on Istio, Envoy, and custom operators.
This article details the evolution of 58.com’s deep learning inference platform, describing the transition from the original SCF‑based architecture to a cloud‑native Istio gateway (architecture 2.0), and explains design choices, traffic‑management, adaptive rate‑limiting, observability, model pre‑warming, and performance improvements.
Seata 1.6.0 release notes detail major enhancements such as Oracle and PostgreSQL multi‑primary‑key support, Dubbo 3 integration, JDK 17 compatibility, ARM64 Docker images, extensive bug fixes, performance optimizations, test updates, contributor acknowledgments, and links to downloads, documentation, and the GitHub repository.
The article introduces Nacos 2.2.0, explains its core functions, details major changes such as removal of legacy code, new data‑source, custom‑environment, and trace plugins, lists added features, refactorings, bug fixes, and provides reference links for further reading.
This article explains Kubernetes pod architecture, networking, external exposure, and how Services use virtual IPs and selectors to provide load balancing and dynamic discovery of pod changes, including the role of kube-proxy and the limitations of using Nginx for pod-level balancing.
The Nacos 2.2.0 GA release introduces several new plugins—including rate‑limiting, custom environment variable, and datasource plugins—removes legacy redundant code, enhances tracing and batch registration, and delivers extensive refactoring and bug fixes to improve stability and extensibility for cloud‑native service discovery and configuration management.
The article provides a detailed technical overview of Baidu’s LiangYiFu 3.0 quantum computing platform, describing its cloud‑native architecture, five client tiers, the QaaS scheduling layer, extensible modules, and the enterprise‑focused LiangXi solution, while highlighting how it abstracts hardware complexities for developers.
This article explains how Dubbo 3.0’s Triple RPC protocol now supports custom exception serialization, detailing the original exception handling flow, the limitations of header‑based error transport, and the proposed TripleWrapper‑Protobuf solution for reliable client‑side error capture.
This article explains what etcd is, compares it with Zookeeper, describes its architecture and core components such as WAL, snapshots and boltdb, outlines its key features, and shows how it powers service registration, watch mechanisms, cluster monitoring and leader election in cloud‑native systems.
This article compares the main cloud‑native ingress gateway options—Nginx Ingress, Alibaba Cloud ALB, and Alibaba Cloud MSE—covering their architecture, exposure methods, configuration nuances, health‑check settings, reload behavior, and practical selection guidelines for Kubernetes deployments.
The article examines when and how large enterprises should adopt microservice architecture, discussing cultural, technical, and operational impacts, benefits of splitting monoliths, key governance considerations, and the importance of DevOps and skilled teams.
ByteDance’s cloud‑native transformation details a layered technical system, multi‑year Kubernetes‑based evolution, unified multi‑cluster resource management, and hierarchical scheduling, illustrating how the company achieves high development speed, resource efficiency, and prepares for next‑generation serverless infrastructure.
Kubernetes 1.26, themed “Electrifying,” introduces 37 enhancements—including registry changes, storage upgrades, signed release artifacts, Windows high‑privilege containers, metric and scheduling improvements—while promoting 11 features to stable, deprecating 12 APIs, and emphasizing sustainability and carbon‑footprint awareness.
This article explains how the ParallelGCThreads JVM flag defaults differ across JDK versions, why older JDKs report excessive GC threads inside limited‑CPU containers, and how to verify and retrieve accurate container resource limits using cgroup files and updated JDK releases.
This article explains why the financial sector is moving its big‑data workloads to cloud‑native platforms, compares cloud‑native systems with traditional Hadoop, describes deployment options such as Serverless YARN and Arcee Operator, and details the high‑performance GRO scheduler, agent, and ResLake resource‑lake architecture that together improve resource utilization, reduce costs, and ensure reliable, low‑latency processing for finance workloads.
OpenCloudOS, a community‑driven Linux distribution, combines full‑stack hardware support, autonomous supply‑chain control, and extensive performance optimizations to deliver a high‑availability, cloud‑native server OS that rivals CentOS while offering 99.999% uptime and up to 50% performance gains in real‑world workloads.
The article explains how cloud‑native architectures, data governance, intelligent fusion, and privacy computing are driving the evolution of big data, recounting the history from Google’s early papers and Hadoop to modern managed services, compute‑storage separation, AI‑powered recommendation platforms, and real‑world success cases.
This article details iQIYI's multi‑year journey of mixing offline and online workloads using Dragonfly and Koordinator, covering architectural evolution, key factors for successful co‑location, resource‑allocation strategies, the role of Anolis OS, pilot results, and future directions.
GitOps leverages Git as an immutable source of truth to streamline CI/CD pipelines for Kubernetes, enhancing productivity, security, and compliance by providing observable, auditable deployments, centralized control, and easy rollbacks, while requiring dedicated tools such as Flux or Weave GitOps Core for full implementation.
This article recounts Ximalaya's multi‑year containerization effort, detailing the evolution from early Docker templates and Marathon to Kubernetes, the development of internal tools like barge and k8s‑sync, health‑check strategies, deployment patterns, and the practical lessons gained from integrating containers with existing middleware.
This guide explains how to use Alibaba Cloud's Serverless Function Compute together with Message Service and Connector to create a cost‑effective, high‑throughput, low‑latency message‑consumption pipeline, covering architecture, key features, step‑by‑step configuration, code examples, and best‑practice recommendations.
This guide explains how to extend Kubernetes to edge nodes using OpenYurt and KubeVela, covering node‑pool creation, Helm‑based Nginx Ingress deployment, custom traits for differential configuration, patch strategies, and verification commands, all illustrated with code snippets and images.
This guide explains how to push a Kubernetes cluster toward its theoretical limit of 5,000 nodes by detailing official limits, master node sizing for GCE and AWS, kube‑apiserver high‑availability and connection‑count tuning, scheduler and controller‑manager leader election settings, kubelet optimizations, and DNS anti‑affinity configuration.
This article provides a comprehensive overview of microservice architecture, outlining the four main technology dimensions—microservice management, storage, business support, and DevOps—and lists the most relevant tools and platforms for each area to help architects choose optimal solutions.
This guide explains how to collect essential Windows metrics with Prometheus, set up Grafana dashboards for CPU, memory, disk, network, and process monitoring, and configure alert rules, while also comparing self‑hosted and Alibaba Cloud Prometheus solutions for seamless Windows observability.
This guide walks you through using Rainbond to rapidly deploy the Spring Cloud Blade microservice platform, covering front‑end frameworks, back‑end modules, required services such as Nacos, Redis and Sentinel, configuration files, component setup, and final topology verification.
This article explains how to implement a full-link gray (canary) release in a Spring Cloud Alibaba microservice architecture by marking traffic at the gateway, propagating the gray tag through request headers, customizing Ribbon load‑balancing, and configuring Nacos metadata for selective service routing.
This article details Qunar's migration from a Prometheus‑based monitoring stack to VictoriaMetrics, describing the limitations they faced, the architectural redesign using vmagent, vmcluster, and vmalert, and the resulting performance improvements and operational benefits for large‑scale Kubernetes environments.
This article explains how front‑end engineers can transition from traditional static‑site deployment to fully cloud‑native web applications by leveraging Alibaba Cloud’s Serverless services—OSS, CDN, and Function Compute—to achieve automatic scaling, zero‑ops maintenance, cost efficiency, and enhanced security for dynamic APIs, tasks, and media processing.
This article explains Istio’s core networking concepts, detailing how sidecar and init containers are injected into a Kubernetes cluster, the roles of the data and control planes, and provides step‑by‑step commands and YAML examples for manual and automatic sidecar injection.
This article explains how RocketMQ's message lifecycle can be fully observed using OpenTelemetry‑based metrics, covering producer, broker, and consumer stages, and shows practical monitoring, alerting, and troubleshooting practices for cloud‑native deployments.
This article explains how to publish a Kubernetes application, use Config APIs like Deployment, Service, and Ingress, leverage Kustomize and a custom kustz.yml abstraction to manage resources across versions, and deploy everything with kubectl.
Since Kubernetes 1.25 the default container image registry moved from k8s.gcr.io to the multi‑provider CDN‑backed registry.k8s.io, improving download speed and resilience, but requiring clusters to update image references or mirror images for restricted environments.
The article explores Apache Flink’s eight‑year journey to becoming a top‑level Apache project, Alibaba’s extensive contributions, the rise of stream‑batch unified computing, its impact on real‑time data integration, cloud‑native deployment, and the emerging Flink‑based data‑warehouse and serverless solutions.
ByteDance’s cloud‑native computing team, led by Li Yakun, details how they transformed a Hadoop‑centric big‑data stack into a Kubernetes‑driven platform—customizing storage, middleware, and scheduling—to support petabyte‑scale workloads, achieve over 40% resource utilization, and sustain rapid product growth.
This article analyzes the evolution of network functions from hardware to virtual machines and finally to cloud‑native containers, explains CNF deployment in 5G core and access networks, details SDEWAN architecture and gray‑release processes, and outlines future trends and design principles.
This article recounts AirQo's transition from a traditional CI/CD workflow to a GitOps model using ArgoCD, detailing the operational pain points they faced, the benefits gained such as automated sync, self‑service for developers, and easier rollbacks, and offering practical insights for similar cloud‑native teams.
HeLian Health’s journey from a fast‑iteration monolith to a stable, cloud‑native microservice architecture illustrates how modular design, Dubbo‑Nacos integration, Kubernetes, and Alibaba Cloud MSE can overcome low‑fault tolerance challenges in the medical check‑up domain while boosting release efficiency and reducing operational costs.
The interview with Alibaba researcher Wang Feng reviews Flink's eight‑year journey to a top Apache project, its massive scale at Double 11, the push toward unified stream‑batch computing, emerging storage challenges, and the roadmap for cloud‑native, real‑time data warehousing.
This interview with Alibaba researcher Wang Feng (aka Mo Wen) explores Apache Flink’s eight‑year journey to top‑level Apache status, its unified stream‑batch architecture, the rise of Flink Table Store and CDC, and how cloud‑native deployments are reshaping real‑time big data processing.
The article explains that cloud‑native is a comprehensive technology system and methodology—encompassing containers, micro‑services, DevOps, and continuous delivery—while linking lean thinking principles to operational transformation and emphasizing the need for business‑focused, platform‑centric ops teams.
This article explains how to achieve end‑to‑end gray (canary) traffic routing for microservice chains on Kubernetes by using MSE's lane features and configuring Ingress controllers, covering the required concepts, configuration steps, and verification methods.
Docker, an open‑source platform for lightweight, portable containers, offers numerous practical applications—from deploying software and running minimal Linux distributions to hosting servers, creating custom images, and streamlining development environments—making it a versatile tool that reshapes DevOps and IT workflows.
The Open Cluster Management (OCM) project releases v0.9.0 with enhanced security, service exposure, plugin configurability, API updates, and outlines upcoming features like independent control‑plane deployment, ArgoCD pull‑mode integration, placement strategies, and edge‑scenario support.
This article analyzes the challenges of traditional Kubernetes cluster installation, proposes a fully automated, zero‑touch deployment framework that abstracts and packages all required resources, and explores its architecture, implementation using Go, cloud‑edge collaboration with KubeEdge, and future extensions toward CRD‑based operators.
This tutorial explains the background of the Kubernetes Gateway API, compares it with the traditional Ingress API, and provides a hands‑on walkthrough for installing the Gateway API CRDs, configuring Istio to use a Gateway resource, exposing a sample httpbin service, updating routes, and cleaning up the environment.
This article walks through building, compiling, containerizing, and deploying a simple mock‑response Wasm plugin for Higress on a Kubernetes cluster, explaining the code structure, configuration, plugin lifecycle, and the three revolutionary features of Wasm‑based gateway extensions.
This article explains how RocketMQ integrates with Prometheus and Grafana to provide comprehensive metrics, tracing, and logging, detailing the exporter architecture, deployment choices, span topology, dashboard examples, and ARMS‑based alerting for cloud‑native message‑queue observability.
This article reviews the practical applications and experimental explorations of StarRocks at 360, describing the cloud‑native lake‑warehouse product Yunzhou, its three‑tier architecture, performance comparisons with Trino using TPCH 100 GB, challenges of Kubernetes integration, and future directions for storage‑compute separation.
Former GitHub CTO Jason Warner argues that the widespread shift to microservices is a major architectural error, advocating a monolith‑first approach and emphasizing organizational challenges, while case studies from GitHub, Uber, and Botify illustrate both the pitfalls and strategic alternatives to microservice adoption.
Guoxin Securities’ QianKun centralized operation platform showcases a cloud‑native, micro‑service architecture that achieved 99.95% availability through containerization, multi‑region deployment, AI‑driven capacity forecasting, and comprehensive DevOps practices, offering a 24/7 seamless account‑opening experience and setting industry benchmarks.
This tutorial walks you through installing Devtron on Kubernetes, creating additional Kind clusters, adding them as managed clusters, configuring environments, adding custom Helm repositories, and deploying a sample Spring Boot application across multiple clusters using Devtron's UI and chart groups.
