Cloud Computing PDF

Introduc tion Definitions Characteristics Service Models Deployment Virtualization Cloud Services Edge Computing Cloud Computing Cloud Computing Introduc tion Definitions Characteristi...

Introduc tion Definitions Characteristics Service Models Deployment Virtualization Cloud Services Edge Computing Cloud Computing Cloud Computing Introduc tion Definitions Characteristics Service Models Deployment Virtualization Cloud Services Edge Computing Definitions 1 Introduction 2 Definitions 3 Characteristics 4 Service Models 5 Deployment Models 6 Virtualization and Elasticity 7 Typical Cloud Services Data Storage in the Cloud Communications: Publish/Subscribe Batch Processing: Map/Reduce Serveless Computing / Function as a Service 8 Edge Computing Introduc tion Definitions Characteristics Service Models Deployment Virtualization Cloud Services Edge Computing Define Cloud Computing Introduc tion Definitions Characteristics Service Models Deployment Virtualization Cloud Services Edge Computing Distributed Computing 6 Distributed System: Definition (Wikipedia) A system whose components are located on different networked computers, which then communicate and coordinate their actions by passing messages to one another. Very broad! Different models are possible: Centralized Peer-to-peer/”mesh” Hybrid Introduc tion Definitions Characteristics Service Models Deployment Virtualization Cloud Services Edge Computing Grid Computing 7 Grid Computing: Definition (Wikipedia) A combination of computer resources from multiple administrative domains applied to a common task. effectively functioning as a “virtual supercomputer” Introduc tion Definitions Characteristics Service Models Deployment Virtualization Cloud Services Edge Computing Utility Computing 8 Utility Computing: Definition (Wikipedia) The packaging of computing resources (computa- tion, storage etc.) as a metered service similar to a traditional public utility. Users only pay for what they use Introduc tion Definitions Characteristics Service Models Deployment Virtualization Cloud Services Edge Computing Cloud Computing? 9 Introduc tion Definitions Characteristics Service Models Deployment Virtualization Cloud Services Edge Computing Cloud Computing? 9 Grid Computing + Utility Computing? Very hard to define – can mean so many different things to different parties! Many definitions Cloud Computing: Definition (NIST) Cloud computing is a model for enabling ubiquitous, convenient, on-demand network access to a shared pool of configurable computing resources (e.g., net- works, servers, storage, applications, and services) that can be rapidly provisioned and released with minimal management effort or service provider interaction. This cloud model is composed of 5 essential characteristics, 3 service models, and 4 deployment models. Introduc tion Definitions Characteristics Service Models Deployment Virtualization Cloud Services Edge Computing Cloud Characteristics 10 1 Introduction 2 Definitions 3 Characteristics 4 Service Models 5 Deployment Models 6 Virtualization and Elasticity 7 Typical Cloud Services Data Storage in the Cloud Communications: Publish/Subscribe Batch Processing: Map/Reduce Serveless Computing / Function as a Service 8 Edge Computing Introduc tion Definitions Characteristics Service Models Deployment Virtualization Cloud Services Edge Computing Cloud Characteristics (1) 11 1. On-demand Self Service Ability to provision computing capabilities without intervention Computation (“aka machine”) time Storage Introduc tion Definitions Characteristics Service Models Deployment Virtualization Cloud Services Edge Computing Cloud Characteristics (1) 11 1. On-demand Self Service Ability to provision computing capabilities without intervention Computation (“aka machine”) time Storage 2. Broad network access Capabilities available over the network Accessible by thin and thick clients (e.g., web, desktop/laptops, mobile devices, etc.) Introduc tion Definitions Characteristics Service Models Deployment Virtualization Cloud Services Edge Computing Cloud Characteristics (2) 12 3. Resource pooling Multi-tenancy: the same cloud infrastructure can serve multiple customers, host multiple VMs, applications Computing resources are pooled (grouped as a single service) Storage Processing Memory Network Physical and logical resources are dynamically assigned and reassigned according to consumer demand Location independence Precise location of the resources irrelevant Only a general idea (e.g., Amazon EC2 US-east) Introduc tion Definitions Characteristics Service Models Deployment Virtualization Cloud Services Edge Computing Cloud Characteristics (3) 13 4. Rapid elasticity Elastic provisioning – scaling up and down Can be done automatically To consumers: pool of resources might appear to be infinite Introduc tion Definitions Characteristics Service Models Deployment Virtualization Cloud Services Edge Computing Cloud Characteristics (3) 13 4. Rapid elasticity 5. Measured service Metering of the different resources CPU (e.g., $/CPU time in ms) Network bandwidth (e.g., $/gb) Processing (e.g., $/X requests) Storage (e.g., $/gb) Monitoring, controlling, reporting Full transparency for cloud operator and consumer Introduc tion Definitions Characteristics Service Models Deployment Virtualization Cloud Services Edge Computing Service Models 14 1 Introduction 2 Definitions 3 Characteristics 4 Service Models 5 Deployment Models 6 Virtualization and Elasticity 7 Typical Cloud Services Data Storage in the Cloud Communications: Publish/Subscribe Batch Processing: Map/Reduce Serveless Computing / Function as a Service 8 Edge Computing Introduc tion Definitions Characteristics Service Models Deployment Virtualization Cloud Services Edge Computing Hierarchy of Service Models (Source) 15 Introduc tion Definitions Characteristics Service Models Deployment Virtualization Cloud Services Edge Computing Infrastructure as a Service (IaaS) (1) 16 Consumer can provision virtualized computing resources (aka VMs) Processing, storage, network, GPU Can include OS and applications, or be bare metal Example: Amazon EC2, Azure Consumer doesn’t manage the hardware (physical or virtualized) But has control over the OS, storage, applications, and limited network settings e.g., firewall, port redirection, VLans, etc Introduc tion Definitions Characteristics Service Models Deployment Virtualization Cloud Services Edge Computing Infrastructure as a Service (IaaS) (2) 17 Virtualizing a machine implies that all of its components must be virtualized as well! Processor (CPU): virtualized from “real”, physical CPUs Hardware acceleration is available on most recent processors Support for multiple cores “Standardized” metrics for modelling the performance of CPUs (e.g., Amazon vCPU) … Memory Storage Networking: SDNs Network configuration is defined by software and not purely by the hardware (routers, switches) Bandwidth, firewalls, subnets, etc. GPUs and other devices The components of a VM are not all necessarily located on the same physical machine! Introduc tion Definitions Characteristics Service Models Deployment Virtualization Cloud Services Edge Computing Data storage in the cloud 18 File storage: Common/typical storage abstractions (file systems, folders, files, etc.) Emulation of a “local” hard disk, but provided over the network Protocols: NFS, Google file system, etc. Introduc tion Definitions Characteristics Service Models Deployment Virtualization Cloud Services Edge Computing Data storage in the cloud 18 File storage: Common/typical storage abstractions (file systems, folders, files, etc.) Emulation of a “local” hard disk, but provided over the network Protocols: NFS, Google file system, etc. Object storage: Storage of objects and metadata (BLOB) ID for each “object” Typically accessed thru standard access protocols (e.g., HT TP) Version control systems (VCS) (e.g., Git, SVN) make use object storage Different storage systems are available based on customer needs (costs, frequency of data reads/writes, throughput, latency, etc.) Replication, versionning, encryption, availability in several “zones”, etc. e.g., Amazon S3, Google Cloud Storage Introduc tion Definitions Characteristics Service Models Deployment Virtualization Cloud Services Edge Computing Data storage in the cloud 18 File storage: Common/typical storage abstractions (file systems, folders, files, etc.) Emulation of a “local” hard disk, but provided over the network Protocols: NFS, Google file system, etc. Object storage: Storage of objects and metadata (BLOB) ID for each “object” Typically accessed thru standard access protocols (e.g., HT TP) Version control systems (VCS) (e.g., Git, SVN) make use object storage Different storage systems are available based on customer needs (costs, frequency of data reads/writes, throughput, latency, etc.) Replication, versionning, encryption, availability in several “zones”, etc. e.g., Amazon S3, Google Cloud Storage, Amazon Glacier Block storage (volumes): Very low-level: emulates a fixed storage block Can be mount as a networked hard disk, either over a filesystem or raw Different storage technologies are available: SSD,, etc. Introduc tion Definitions Characteristics Service Models Deployment Virtualization Cloud Services Edge Computing Platform as a Service (PaaS) 19 Enables the deployment, management and execution of consumer or acquired applications onto cloud infrastructure (IaaS) For customers: alleviates the need for managing the applications on their own infrastructure Can be written into a variety of languages Using a variety of libraries, services, tools supported by the provider e.g., Web apps (Heroku, Google App Engine), APIs, microservices No control over underlying cloud infrastructure (IaaS)! Control over deployed applications Might have limited control over configuration settings of the hosting environment (e.g., config files) Introduc tion Definitions Characteristics Service Models Deployment Virtualization Cloud Services Edge Computing Software as a Service (SaaS) 20 Use the provider’s specific applications Over the cloud provider’s infrastructure (hardware + software/PaaS) Accessible from various clients Thin & thick clients, mobile, web (e.g., web-based email) Consumer does not manage the underlying cloud infrastructure (network, servers, OS, storage, applications) Exception: limited user-specific application configuration settings (e.g., GMail settings) Introduc tion Definitions Characteristics Service Models Deployment Virtualization Cloud Services Edge Computing Provisioning in Service Models 21 Exercise To which service model (SAAS, PAAS, IAAS) does each of the following correspond? Editing a document online on Google Docs Testing a new Linux Kernel on an Amazon VM Accessing a MySQL database service Deploying a Python application Provisioning a virtual machine  Exercise To which service model (SAAS, PAAS, IAAS) does each of the following correspond? Editing a document online on Google Docs SaaS Testing a new Linux Kernel on an Amazon VM IaaS Accessing a MySQL database service PaaS Deploying a Python application PaaS Provisioning a virtual machine IaaS Introduc tion Definitions Characteristics Service Models Deployment Virtualization Cloud Services Edge Computing Deployment Models 22 1 Introduction 2 Definitions 3 Characteristics 4 Service Models 5 Deployment Models 6 Virtualization and Elasticity 7 Typical Cloud Services Data Storage in the Cloud Communications: Publish/Subscribe Batch Processing: Map/Reduce Serveless Computing / Function as a Service 8 Edge Computing Introduc tion Definitions Characteristics Service Models Deployment Virtualization Cloud Services Edge Computing Public cloud 23 Open use by general public Owned by business, academic, government organization, or a combination Exists on the premises the of cloud provider Example: Amazon, Google, MS Azure Introduc tion Definitions Characteristics Service Models Deployment Virtualization Cloud Services Edge Computing Private cloud 24 Exclusive use of a single organization with multiple “internal” consumers e.g. different business units within a given organization Owned, managed, operated by organization, or a third-party, or a combination May exist on or off premises Example: a large company (e.g., Google Internal Cloud) Introduc tion Definitions Characteristics Service Models Deployment Virtualization Cloud Services Edge Computing Community cloud 25 Exclusive use of a specific community of consumers from organizations with shared concerns Mission, security requirements, policy, compliance considerations Owned, managed, operated by one or more organizations in the community, a third party, or a combination of them May exist on or off premises Examples: Amazon Government Cloud, clouds that comply with BC data policies (e.g., UBC Workspace) Introduc tion Definitions Characteristics Service Models Deployment Virtualization Cloud Services Edge Computing Hybrid cloud 26 A composition of two or more distinct cloud infrastructure Can include on-premise computing/storage/network infrastructure  Exercise Choose the most accurate deployment model for each of the following (public cloud, private cloud, community cloud, hybrid cloud): 1.Due to strict privacy laws, UBC offers a storage service operated by the university and hosted on the university premises. 2.Intel-ligent operates a cloud service for its own internal needs – whenever the demand goes above the available capacity, the extra load is sent to the MS Azure cloud. 3.Your Android phone automatically stores the photos that you have taken onto the Google Photos service. 4.The healthcare network provides a cloud service that allows for storing patient information from multiple establishments (hospitals, clinics, pharmacies, etc.) scattered across the province. The information is stored on a layer of servers deployed/spread over the various establishments. The service respects various compliance policies pertaining to storing critical patient information.  Exercise Choose the most accurate deployment model for each of the following (public cloud, private cloud, community cloud, hybrid cloud): 1.Due to strict privacy laws, UBC offers a storage service operated by the university and hosted on the university premises. Private cloud or community cloud 2.Intel-ligent operates a cloud service for its own internal needs – whenever the demands goes above the available capacity, the extra load is sent to the MS Azure cloud. Hybrid cloud 3.Your Android phone automatically stores the photos that you have taken onto the Google Photos service. Public cloud 4.The healthcare network provides a cloud service that allows for storing patient information from multiple establishments (hospitals, clinics, pharmacies, etc.) scattered across the province. The information is stored on a layer of servers deployed/spread over the various establishments. The service respects various compliance policies pertaining to storing critical patient information. Community cloud Introduc tion Definitions Characteristics Service Models Deployment Virtualization Cloud Services Edge Computing Virtualization 27 1 Introduction 2 Definitions 3 Characteristics 4 Service Models 5 Deployment Models 6 Virtualization and Elasticity 7 Typical Cloud Services Data Storage in the Cloud Communications: Publish/Subscribe Batch Processing: Map/Reduce Serveless Computing / Function as a Service 8 Edge Computing Introduc tion Definitions Characteristics Service Models Deployment Virtualization Cloud Services Edge Computing What is virtualization? 28 Decoupling the physical resources (physical hardware) into virtual resources Why virtualize? Cloud provider might have heterogeneous hardware Offering a consistent configuration to the different customers of the cloud CPU performance Amount of memory Storage Network bandwidth Offering additional isolation (reliability) Virtualization of resources happen at different levels based on the service model! Introduc tion Definitions Characteristics Service Models Deployment Virtualization Cloud Services Edge Computing Virtualization: IaaS (1) 29 Hardware-level virtualization (lowest level) 1 Physical machine ⇒ n virtual machines Hypervisor: VMWare, VirtualBox, MS HyperV, Xen, etc. Run over an OS or “bare-metal” Nowadays, virtualization is hardware-assisted: can run at near-native speeds Introduc tion Definitions Characteristics Service Models Deployment Virtualization Cloud Services Edge Computing Virtualization (IaaS) (2) 30 Hardware-level virtualization (lowest level) Virtualized Hardware CPU (modern CPUs support virtualization extensions - e.g., Intel VT, AMD-V) Memory: portions of the RAM of the host machine are reserved Storage: virtual hard drives and other I/O peripherals, data center storage Network: virtual network adapters, virtualized networks/subnets GPU: for specific (AI) applications Introduc tion Definitions Characteristics Service Models Deployment Virtualization Cloud Services Edge Computing Virtualization: PaaS / SaaS 31 Virtualization of the combined resources of a pool of machines (VMs) Build over IaaS virtualization layer Processing power (CPU) Pools of memory Distributed data storage Virtualized networking and adressing Introduc tion Definitions Characteristics Service Models Deployment Virtualization Cloud Services Edge Computing Elasticity 32 Allocating a pool of resources from the provider in an “elastic” manner, according to current needs Resource allocation can be made directly according to user requirements (e.g., Amazon EC2 dashboard, or thru the command-line) Can be triggered by the needs of higher-level apps & services deployed over higher-level layers (e.g., PaaS) Introduc tion Definitions Characteristics Service Models Deployment Virtualization Cloud Services Edge Computing Elasticity: IaaS 33 Two approaches to scalability: Vertical: more powerful hardware (limited) Horizontal: partitioning / sharding Elasticity: IaaS (Infrastructure as a Service) Allocating new VM instances Deallocating instances which aren’t needed anymore Allocating storage, RAM, network (or a specific network configuration), etc. (can be properties of the VMs) Introduc tion Definitions Characteristics Service Models Deployment Virtualization Cloud Services Edge Computing Elasticity: PaaS 34 Elasticity: PaaS (Platform as a Service): Automatic provisioning of VM/physical resources (IaaS layer) to execute the PaaS application The elasticity of the application itself might or might not be done automatically Introduc tion Definitions Characteristics Service Models Deployment Virtualization Cloud Services Edge Computing Elasticity: PaaS 34 Elasticity: PaaS (Platform as a Service): Automatic provisioning of VM/physical resources (IaaS layer) to execute the PaaS application The elasticity of the application itself might or might not be done automatically Example: for a request-based application, the PaaS “execution layer” could provision / allocate enough resources as necessary from the IaaS layer to satisfy the current volume of requests The unit of measure for control & billing purposes can then be different (higher-level) compared to the billing metrics for the IaaS layer For instance, the customer can be billed by the number of requests or for the execution time alloted for handling the requests (as opposed to billing for the “raw” CPU usage, memory, etc. of the VM) Introduc tion Definitions Characteristics Service Models Deployment Virtualization Cloud Services Edge Computing Elasticity: SaaS 35 Elasticity: SaaS (Software as a Service): Fully managed provisioning of the PaaS layer e.g., Gmail will provision enough combined resources at the PaaS layer, which in turn will provision enough resources at the IaaS layer (type and # of VMs, network, etc.) Introduc tion Definitions Characteristics Service Models Deployment Virtualization Cloud Services Edge Computing Data Storage in the Cloud 36 1 Introduction 2 Definitions 3 Characteristics 4 Service Models 5 Deployment Models 6 Virtualization and Elasticity 7 Typical Cloud Services Data Storage in the Cloud Communications: Publish/Subscribe Batch Processing: Map/Reduce Serveless Computing / Function as a Service 8 Edge Computing Introduc tion Definitions Characteristics Service Models Deployment Virtualization Cloud Services Edge Computing Data Storage in the Cloud 37 How can data be stored across different nodes? Introduc tion Definitions Characteristics Service Models Deployment Virtualization Cloud Services Edge Computing Data Storage in the Cloud 37 How can data be stored across different nodes? Distributed File Systems Google FS, Hadoop Provides file-system like abstractions in a distributd manner Block Storage Amazon S3 (storage of objects, can be files) Databases: SQL NoSQL (e.g., Key-value Stores, MongoDB, etc.) Introduc tion Definitions Characteristics Service Models Deployment Virtualization Cloud Services Edge Computing Data Storage in the Cloud: Properties 38 Scalability High availability Low latency Durability Fault tolerant Predictable costs Introduc tion Definitions Characteristics Service Models Deployment Virtualization Cloud Services Edge Computing Data Storage in the Cloud: Properties 38 Scalability High availability Low latency Durability Fault tolerant Predictable costs Tradeoff: the CAP Theorem Consistency Availability Partition tolerance Pick only two :-) Cloud storage systems often opt for eventual consistency Introduc tion Definitions Characteristics Service Models Deployment Virtualization Cloud Services Edge Computing Communications: Publish/Subscribe 39 1 Introduction 2 Definitions 3 Characteristics 4 Service Models 5 Deployment Models 6 Virtualization and Elasticity 7 Typical Cloud Services Data Storage in the Cloud Communications: Publish/Subscribe Batch Processing: Map/Reduce Serveless Computing / Function as a Service 8 Edge Computing Introduc tion Definitions Characteristics Service Models Deployment Virtualization Cloud Services Edge Computing Publish/Subscribe Paradigm 40 Provides an elegant way to decouple content producers (publishers) from content consumers (subscribers) Publications are matched against subscriptions Many flavours of publish/subscribe Publish/Subscribe - Example Introduc tion Definitions Characteristics Service Models Deployment Virtualization Cloud Services Edge Computing Publish/Subscribe Paradigm 40 Provides an elegant way to decouple content producers (publishers) from content consumers (subscribers) Publications are matched against subscriptions Many flavours of publish/subscribe Publish/Subscribe - Example Introduc tion Definitions Characteristics Service Models Deployment Virtualization Cloud Services Edge Computing Topic-Based Publish/Subscribe 41 Very common flavour of pub/sub Subscription language: a key (topic name) Publications tagged with a topic T , sent to all subscribers of T Example - Weather Reports Introduc tion Definitions Characteristics Service Models Deployment Virtualization Cloud Services Edge Computing Topic-Based Publish/Subscribe 41 Very common flavour of pub/sub Subscription language: a key (topic name) Publications tagged with a topic T , sent to all subscribers of T Example - Weather Reports Introduc tion Definitions Characteristics Service Models Deployment Virtualization Cloud Services Edge Computing Applications of Topic-Based Pub/Sub 42 Traffic alert systems Weather alert systems Desirable properties: Scalability & Mobile notif. frameworks Social networks Elasticity Low Latency Reduced & Predictable IoT Multiplayer Games Costs Introduc tion Definitions Characteristics Service Models Deployment Virtualization Cloud Services Edge Computing Batch Processing: Map/Reduce 43 1 Introduction 2 Definitions 3 Characteristics 4 Service Models 5 Deployment Models 6 Virtualization and Elasticity 7 Typical Cloud Services Data Storage in the Cloud Communications: Publish/Subscribe Batch Processing: Map/Reduce Serveless Computing / Function as a Service 8 Edge Computing Introduc tion Definitions Characteristics Service Models Deployment Virtualization Cloud Services Edge Computing Map/Reduce (Source) 44 Functional Decomposition: Breaking a large problem into a set of small problems Each small problem: can be solved by a functional transformation of input data can be executed in complete isolation (parallel computing) Examples (next slides) – what do these Linux programs do? grep wc (word count) Introduc tion Definitions Characteristics Service Models Deployment Virtualization Cloud Services Edge Computing grep with MapReduce 45 Partitioning the files to be searched onto several nodes (map) Executing “grep” on each instance Partitioning the intermediate results to send them to a “reducer” Concatenation of all intermediate results Introduc tion Definitions Characteristics Service Models Deployment Virtualization Cloud Services Edge Computing wc with MapReduce 46 Partitioning the files on which the wordcount should be performed onto several nodes (map) Executing “count” on each instance to compute the number of occurences of each word Partitioning the intermediate “counts” to send them to a “reducer” (e.g., by hashing the words) Merging of all results (adding the partial counts for each word) Introduc tion Definitions Characteristics Service Models Deployment Virtualization Cloud Services Edge Computing grep and wc with MapReduce 47 Introduc tion Definitions Characteristics Service Models Deployment Virtualization Cloud Services Edge Computing Serveless Computing / Function as a Service 48 1 Introduction 2 Definitions 3 Characteristics 4 Service Models 5 Deployment Models 6 Virtualization and Elasticity 7 Typical Cloud Services Data Storage in the Cloud Communications: Publish/Subscribe Batch Processing: Map/Reduce Serveless Computing / Function as a Service 8 Edge Computing Introduc tion Definitions Characteristics Service Models Deployment Virtualization Cloud Services Edge Computing Function as a Service (FaaS) 49 Application made of a set of functions Executed upon certain events being triggered Web request File upload Change to DB Timer Executed within containers (thin VMs) Full isolation FaaS functions are stateless! Changes in state must be persisted to durable storage Example: Amazon Lambda, Google Cloud Functions, MS Azure Functions Some “functions” can be executed “at the edge”: e.g., lambda@edge Introduc tion Definitions Characteristics Service Models Deployment Virtualization Cloud Services Edge Computing Edge Computing 50 1 Introduction 2 Definitions 3 Characteristics 4 Service Models 5 Deployment Models 6 Virtualization and Elasticity 7 Typical Cloud Services Data Storage in the Cloud Communications: Publish/Subscribe Batch Processing: Map/Reduce Serveless Computing / Function as a Service 8 Edge Computing Introduc tion Definitions Characteristics Service Models Deployment Virtualization Cloud Services Edge Computing Edge Computing 51 Edge Computing, from a systems point of view, aims at processing the data as close as possible to where the data is produced and consumed. The definition by itself is very vague, and can refer to various models that are described in the literature. For instance, for cloud providers, the “edge” can refer to smaller/micro cloud deployoments that are more “localized” In other contexts, the “edge” can refer to processing the data onto the devices themselves that are becoming more and more powerful For instance, Raspberry Pi devices run a full Linux OS Introduc tion Definitions Characteristics Service Models Deployment Virtualization Cloud Services Edge Computing Edge Computing, why is it important? 52 The cloud is widespread and is typically “reachable” from anywhere – however, in some cases, it might not be the best option (in the IoT landscape). Why? Introduc tion Definitions Characteristics Service Models Deployment Virtualization Cloud Services Edge Computing Edge Computing, why is it important? 52 The cloud is widespread and is typically “reachable” from anywhere – however, in some cases, it might not be the best option (in the IoT landscape). Why? 1 Limited connectivity For devices that are deployed in an isolated environment, or if the wireless communication is intermittent Introduc tion Definitions Characteristics Service Models Deployment Virtualization Cloud Services Edge Computing Edge Computing, why is it important? 52 The cloud is widespread and is typically “reachable” from anywhere – however, in some cases, it might not be the best option (in the IoT landscape). Why? 1 Limited connectivity For devices that are deployed in an isolated environment, or if the wireless communication is intermittent 2 Lower latency Some critical apps must react very quickly to specific events – in that case, sending the data to the cloud (back and forth) can push the latency above a critical level e.g., smart vehicles Introduc tion Definitions Characteristics Service Models Deployment Virtualization Cloud Services Edge Computing Edge Computing, why is it important? 52 The cloud is widespread and is typically “reachable” from anywhere – however, in some cases, it might not be the best option (in the IoT landscape). Why? 1 Limited connectivity For devices that are deployed in an isolated environment, or if the wireless communication is intermittent 2 Lower latency Some critical apps must react very quickly to specific events – in that case, sending the data to the cloud (back and forth) can push the latency above a critical level e.g., smart vehicles 3 Alleviating the dependence to a third party Some IoT devices are too strongly coupled to a specific cloud-based service, and can stop working if the cloud service ceases to exist, or if the operator changes its terms of use. Introduc tion Definitions Characteristics Service Models Deployment Virtualization Cloud Services Edge Computing Edge Computing, why is it important? 52 The cloud is widespread and is typically “reachable” from anywhere – however, in some cases, it might not be the best option (in the IoT landscape). Why? 1 Limited connectivity For devices that are deployed in an isolated environment, or if the wireless communication is intermittent 2 Lower latency Some critical apps must react very quickly to specific events – in that case, sending the data to the cloud (back and forth) can push the latency above a critical level e.g., smart vehicles 3 Alleviating the dependence to a third party Some IoT devices are too strongly coupled to a specific cloud-based service, and can stop working if the cloud service ceases to exist, or if the operator changes its terms of use. 4 Security requirements It might be preferable to process the data locally to meet the needs of security requirements and policies. Introduc tion Definitions Characteristics Service Models Deployment Virtualization Cloud Services Edge Computing Edge Computing, why is it important? 52 The cloud is widespread and is typically “reachable” from anywhere – however, in some cases, it might not be the best option (in the IoT landscape). Why? 1 Limited connectivity For devices that are deployed in an isolated environment, or if the wireless communication is intermittent 2 Lower latency Some critical apps must react very quickly to specific events – in that case, sending the data to the cloud (back and forth) can push the latency above a critical level e.g., smart vehicles 3 Alleviating the dependence to a third party Some IoT devices are too strongly coupled to a specific cloud-based service, and can stop working if the cloud service ceases to exist, or if the operator changes its terms of use. 4 Security requirements It might be preferable to process the data locally to meet the needs of security requirements and policies. 5 Reducing the costs In a context in which a lot of data is produced and consumed, processing some of the data locally can reduce the volume of data that is sent and processed in the cloud.