Introduction to Cloud Computing

Cloud Computing Introduction

Cloud Computing Definition

•       Cloud is a parallel and distributed computing system consisting of a collection of inter-connected and virtualised computers that are dynamically provisioned and presented as one or more unified computing resources based on service-level agreements (SLA) established through negotiation between the service provider and consumers. -  Raj Kumar Buyya

•       Clouds are a large pool of easily usable and accessible virtualized resources (such as hardware, development platforms and/or services). These resources can be dynamically reconfigured to adjust to a variable load (scale), allowing also for an optimum resource utilization. This pool of resources is typically exploited by a pay-per-use model in which guarantees are offered by the Infrastructure Provider by means of customized Service Level Agreements - Vaquero

•       “... a pay-per-use model for enabling available, convenient, on-demand network access to a shared pool of configurable computing resources (e.g. networks, servers, storage, applications, services) that can be rapidly provisioned and released with minimal management effort or service provider interaction.” - National Institute of Standards and Technology (NIST)

•       Common characteristics between the most notable definitions, which a cloud should have:

•       pay-per-use (no ongoing commitment, utility prices);

•       elastic capacity and the illusion of infinite resources;

•       self-service interface; and

•       Resources that are abstracted or virtualised.

ROOTS OF CLOUD COMPUTING

•       We can track the roots of clouds computing by observing the advancement of several technologies, especially in

•       hardware (virtualization, multi-core chips),

•       Internet technologies (Web services, service-oriented architectures, Web 2.0),

•       Distributed computing (clusters, grids), and

•       Systems management (autonomic computing, data center automation).

•       From Mainframes to Clouds

•       switch in the IT world

•       From in-house generated computing power

•       into

•       Utility-supplied computing resources delivered over the Internet as Web services

•       century ago

•       when factories, which used to generate their own electric power, realized that it is was cheaper just plugging their machines into the newly formed electric power grid

•       Computing delivered as a utility can be defined as

•       “on demand delivery of infrastructure, applications, and business processes in a security-rich, shared, scalable, and based computer environment over the Internet for a fee”

•       Consumers can attain reduction on IT-related costs by choosing to obtain cheaper services from external providers as opposed to heavily investing on IT infrastructure and personnel hiring.

•       In 1970s,

•        companies who offered common data processing tasks ( payroll automation) operated time-shared mainframes as utilities,

•       It could serve dozens of applications and often operated close to 100% of their capacity.

•       In fact, mainframes had to operate at very high utilization rates simply because they were very expensive.

•       The mainframe era collapsed

•       with the advent of fast and inexpensive microprocessors

•       isolation of workload into dedicated servers

•       Incompatibilities between software stacks and operating systems

•       the unavailability of efficient computer network

•       SOA, Web Services, Web 2.0, and Mashups

•       Web services can glue together applications running on different messaging product platforms, enabling information from one application to be made available to others, and enabling internal applications to be made available over the Internet.

•       Describe, compose, and orchestrate services, package and transport messages between services, publish and discover services, represent quality of service (QoS) parameters, and ensure security in service access.

•       Created on HTTP and XML - providing a common mechanism for delivering services, making them ideal for implementing a service-oriented architecture (SOA).

•       The purpose of a SOA is to address requirements of loosely coupled, standards-based, and protocol-independent distributed computing.

•       In a SOA, software resources are packaged as “services,” which are well-defined, selfcontained modules that provide standard business functionality and are independent of the state or context of other services.

•       , information and services may be programmatically aggregated, acting as building blocks of complex compositions, called service mashups.

•       Grid Computing

•       enables aggregation of distributed resources and transparently access to them

•       share compute and storage resources distributed across different administrative domains, with their main focus being speeding up a broad range of scientific applications

•      

•       Building standard Web services-based protocols that allow distributed resources to be “discovered, accessed, allocated, monitored, accounted for, and billed for, etc., and in general managed as a single virtual system.”

•       Open Grid Services Architecture (OGSA)

•       defining a set of core capabilities and behaviors that address key concerns in grid systems

•       Globus Toolkit is a middleware that implements several standard Grid services

•       Grid brokers, which facilitate user interaction with multiple middleware and implement policies to meet QoS needs.

•       Problems in Grid Computing

•       ensuring QoS in grids is difficult

•       Activities associated with one user or virtual organization (VO) can influence, in an uncontrollable way, the performance perceived by other users using the same platform.

•       availability of resources with diverse software configurations, including disparate operating systems, libraries, compilers, runtime environments but user applications would often run only on specially customized environments

•       Utility Computing

•       grid resource management techniques failed to provide fair and equitable access to resources

•       Traditional metrics (throughput, waiting time, and slowdown) failed to capture the more subtle requirements of users.

•       No flexibility for Users about resource requirements or job deadlines, nor provisions to accommodate users with urgent work.

•       In Utility Computing

•       users assign a “utility” value to their jobs,

•       Where utility is a fixed or time-varying valuation based on QoS constraints (deadline, importance, satisfaction).

•       The valuation is the amount they are willing to pay a service provider to satisfy their demands

•       service providers then attempt to maximize their own utility to get more profit

•       Providers can choose to prioritize high yield user jobs.

•       Hardware Virtualization

•       Hardware virtualization allows running multiple operating systems and software stacks on a single physical platform

•       a software layer, the virtual machine monitor (VMM), also called a hypervisor, mediates access to the physical hardware presenting to each guest operating system a virtual machine (VM), which is a set of virtual platform interfaces

•      

•       Following technologies increased adoption of virtualization        

•       multi-core chips,

•       paravirtualization,

•       hardware-assisted virtualization, and

•       live migration of VMs

•       Traditionally, perceived benefits were

•       improvements on sharing and utilization,

•       better manageability, and

•       Higher reliability.

•       3 basic capabilities regarding management of workload in a virtualized system

•       isolation, consolidation, and migration

•       Work load Isolation

•       execution of one VM should not affect the performance of another VM

•       The consolidation of several individual and heterogeneous workloads onto a single physical platform leads to better system utilization.

•       Workload migration

•       It is done by encapsulating a guest OS state within a VM and allowing it to be suspended, fully serialized, migrated to a different platform, and resumed immediately or preserved to be restored at a later date

VMM platforms

•       VMWare ESXi

•       Provides from server and desktop virtualization to high-level management tools

•       It is a bare-metal hypervisor,

•       Means -  it installs directly on the physical server

•       others may require a host operating system

•       provides advanced virtualization techniques of processor, memory, and I/O

•       Xen

•       open-source project and has served as a base to other virtualization products

•       pioneered the para-virtualization concept

•       Guest operating system, by means of a specialized kernel, can interact with the hypervisor

•       forms the base of commercial hypervisors like

•       Citrix XenServer and

•       Oracle VM

•       KVM

•       The kernel-based virtual machine (KVM) is a Linux virtualization subsystem

•       part of the mainline Linux kernel

•       memory management and scheduling are carried out by existing kernel thus making KVM simpler and smaller than hypervisors

Virtual Appliances and the Open Virtualization Format

•       virtual appliance

•       An application combined with the environment needed to run it

•       Environment - operating system, libraries, compilers, databases, application containers, and so forth.

•       It eases software customization, configuration, and patching and improves portability.

•       An appliance is shaped as a VM disk image associated with hardware requirements, and it can be readily deployed in a hypervisor.

•       Each Hypervisor supports a different VM image format and the formats are incompatible with one another

•       Example – AMI(Amazon Machine Image) format for Amazon EC2 public cloud

•       OVF - Open Virtualization Format (OVF).

•       To facilitate packing and distribution of software to be run on VMs several vendors (VMware, IBM, Citrix, Cisco, Microsoft, Dell, and HP).

•       Aim is

•        “Open, secure, portable, efficient and extensible”.

•       An OVF package consists of a file, or set of files, describing

•       the VM hardware characteristics (e.g., memory, network cards, and disks),

•       operating system details,

•       startup, and

•       shutdown actions,

•        the virtual disks themselves, and

•       Other metadata containing product and licensing information.

•       Extensions to OVF is allowed for management of datacenters and cloud.

•       VMC – Virtual Machine Contract

•       An extension to OVF.

•       A simple example of a VMC is when a cloud consumer wants to specify minimum and maximum amounts of a resource that a VM needs to function; similarly the cloud provider could express resource limits as a way to bound resource consumption and costs.

•       Autonomic Computing

•       Systems should manage themselves, with high-level guidance from humans

•       Autonomic (self-managing) systems rely on

•       monitoring probes and gauges (sensors),

•       on an adaptation engine (autonomic manager) for computing optimizations based on monitoring data, and

•       On effectors to carry out changes on the system.

•       4 properties of autonomic systems(by IBM):

•       self-configuration,

•       self-optimization,

•       self-healing, and

•       Self-protection.

•       IBM has also suggested a reference model for autonomic control loops of autonomic managers, called MAPE-K (Monitor Analyze Plan Execute—Knowledge)

•      the concepts of autonomic computing inspire software technologies for data center automation

•      Its Tasks are

•       management of service levels of running applications;

•       management of data center capacity;

•       proactive disaster recovery; and

•       automation of VM provisioning

LAYERS AND TYPES OF CLOUDS

•      Cloud computing services are divided into three classes ( based on abstraction level of the capability provided and the service model of providers)

•       (1) Infrastructure as a Service,

•       (2) Platform as a Service, and

•       (3) Software as a Service

•      Note:-

•       These abstraction levels can also be viewed as a layered architecture where services of a higher layer can be composed from services of the underlying layer

•      

•       Public cloud

•       “cloud made available in a pay-as-you-go manner to the general public”

•       Private cloud

•       “Internal data center of a business or other organization, not made available to the general public.”

•       Community cloud

•       “shared by several organizations and supports a specific community that has shared concerns (e.g., mission, security requirements, policy, and compliance considerations)

•       Hybrid cloud

•       when a private cloud is supplemented with computing capacity from public clouds

•       Note: - “cloud-bursting” - temporarily renting capacity to handle spikes in load.

•        

DESIRED FEATURES OF A CLOUD

•      These features are essential to

•      enable services that truly represent the cloud computing model and

•      Satisfy expectations of consumers.

•      Self-Service

•      Consumers expect on-demand - nearly instant access to resources.

•      So clouds must allow self-service access

•      customers can request, customize, pay, and use services without intervention of human operators

•      Per-Usage Metering and Billing

•      Services must be priced on a short term basis (e.g., by the hour), allowing users to release (and not pay for) resources as soon as they are not needed.

•      Implement features to allow efficient trading

•       pricing,

•       accounting, and

•       Billing

•      Elasticity

•      illusion of infinite computing resources available on demand

•      it is expected that the additional resources can be

•       (a) provisioned, possibly automatically, when an application load increases and

•       (b) released when load decreases

•     scale up and down

•      Customization

•       Multi-tenant cloud a great disparity between user needs

•       Resources rented from the cloud must be highly customizable.

CLOUD INFRASTRUCTURE MANAGEMENT

•      A key challenge IaaS providers face when building a cloud infrastructure is managing physical and virtual resources

•      The orchestration of resources must be performed in a way to rapidly and dynamically provision resources to applications

•      virtual infrastructure manager (VIM)

•      The software toolkit responsible for this orchestration.

•      VIM or cloud operating system or infrastructure sharing software or virtual infrastructure engine

•     Features of VIM

1.      Virtualization Support.

2.      Self-Service, On-Demand Resource Provisioning.

3.      Multiple Backend Hypervisors.

1.      Some VI managers provide a uniform management layer regardless of the virtualization technology used.

2.      open-source VI managers provide pluggable drivers to interact with multiple hypervisors

4.      Storage Virtualization.

1.      Abstracting logical storage from physical storage.

2.      consolidates all available storage devices in a data center

3.      Allows to create virtual disks independent from device and location

4.      Organized in a storage area network (SAN) and attached to servers via protocols such as Fibre Channel, iSCSI, and NFS

5.   a storage controller provides the layer of abstraction between virtual and physical storage

5.    Interface to Public Clouds

1.      Hybrid cloud - institutions can make good use of their available resources and, in case of spikes in demand, extra load can be offloaded to rented resources

2.      A VI manager can be used in a hybrid cloud setup if it offers a driver to manage the life cycle of virtualized resources obtained from external cloud providers.

6.    Virtual Networking

1.      Virtual networks - isolated network on top of a physical infrastructure independently from physical topology and locations

2.      VLAN – Virtual LAN –

•       isolates traffic that shares a switched network, allowing VMs to be grouped into the same broadcast domain

•       VLAN can be configured to block traffic originated from VMs from other networks

3.      Similarly, the VPN - virtual private network

•       private overlay network on top of a public network

4.     Support for creating and configuring virtual networks to group VMs placed throughout a data center is provided by most VI managers.

7.      Dynamic Resource Allocation

1.      For Less Energy consumption in data centers there should be dynamic consolidation of VMs in a fewer number of servers.

2.      Machines that are not assigned any VM can be turned off or put on a low power state.

3.      Overheating can be avoided by moving load away from hotspots

8.      Virtual Clusters.

1.       group of interconnected VMs for multi-tier Internet applications

9.     Reservation and Negotiation Mechanism.

1.      AR - Advance Reservations

•      users request computational resources to available at a specific time

2.      Best-Effort Requests

•      when users request resources whenever available

3.      In clouds on which resources are scarce, VM placement strategies are used which support queues, priorities, and advance reservations

4.      Leases may be negotiated and renegotiated, allowing provider and consumer to modify a lease or present counter proposals until an agreement is reached.

5.      OpenPEX – uses bilateral negotiation protocol

•      allows users and providers to come to an alternative agreement by exchanging offers and counter offers

10.  High Availability and Data Recovery

1.      Aims at minimizing application downtime and preventing business disruption.

2.      Failover mechanism

•      Detects failure of both physical and virtual servers and restarts VMs on healthy physical servers.

3.      Mission critical applications

•      restarting VMs does not suffice

•      redundant and synchronized VMs are kept in a secondary physical server

•      ensures that a duplicate VM serves the application in case of failures

4.      some VI managers offer data protection mechanisms that perform incremental backups of VM images

5.     The backup workload is often assigned to proxies, thus offloading production server and reducing network overhead

Infrastructure as a Service Provider

•      Iaas Provides

•      virtual servers containing one or more CPUs,

•      running several choices of operating systems and

•      A customized software stack.

•      In addition, storage space and communication facilities are often provided.

•      Specialized features that influence the cost-benefit ratio to be experienced by user applications when moved to the cloud.

•      geographic distribution of data centers;

•      variety of user interfaces and APIs to access the system;

•      specialized components and services that aid particular applications (e.g., loadbalancers, firewalls);

•      choice of virtualization platform and operating systems; and

•      Different billing methods and period (e.g., prepaid vs. post-paid, hourly vs. monthly).

•      Geographic Presence

•      To improve availability and responsiveness, a provider of worldwide services would typically build several data centers distributed around the world

•      For example,

•      Amazon Web Services uses

•      “Availability zones” and “regions” for its EC2 service.

•      Availability zones are distinct locations that are engineered to be insulated from failures in other availability zones

•      provide inexpensive, low-latency network connectivity to other availability zones in the same region

•      Regions are geographically dispersed and will be in separate geographic areas or countries

•      User Interfaces and Access to Servers

•      Public IaaS provider must provide multiple access means to its cloud

•      Different types of UI

•      graphical user interfaces (GUI),

•      command-line tools (CLI), and

•      Web service (WS) APIs.

•      GUIs –

•      to launch, customize, and monitor a few virtual servers and

•      Do not necessary need to repeat the process several times.

•      CLIs

•      For automating repetitive tasks via scripts

•      e.g:- start and shutdown a number of virtual servers at regular intervals

•      WS APIs offer programmatic access to a cloud using standard HTTP requests, thus allowing complex services to be built on top of IaaS clouds.

•      Advance Reservation of Capacity

•      Advance reservations allow users to request for an IaaS provider to reserve resources for a specific time frame in the future.

•      Most clouds only support best-effort requests (user’s requests server whenever resources are available).

•      Example:-

•      Amazon Reserved Instances - advance reservation of capacity

•      It allows users to pay a fixed amount of money in advance to guarantee resource availability at any time during an agreed period and then paying a discounted hourly rate when resources are in use.

•      Only long periods of 1 to 3 years are offered.

•      Automatic Scaling and Load Balancing

•      Applications often need to scale up and down (elasticity) to meet varying load conditions.

•      It allow users to set conditions for when they want their applications to scale up and down

•      based on application-specific metrics such as

•      transactions per second,

•      number of simultaneous users,

•      Request latency, and so forth.

•      Incoming traffic must be automatically distributed among the available servers - after Scale up through automatic scaling

•      This activity enables applications

•      to promptly respond to traffic increase

•      Achieves greater fault tolerance.

•      Service-Level Agreement

•      Offered by IaaS providers to express their commitment to delivery of a certain QoS.

•      To customers it serves as a warranty.

•      An SLA usually include

•      Availability and performance guarantees.

•      Most IaaS providers focus their SLA terms on availability guarantees, specifying the minimum percentage of time the system will be available during a certain period.

•      For instance, Amazon EC2 states that “if the annual uptime Percentage for a customer drops below 99.95% for the service year, that customer is eligible to receive a service credit equal to 10% of their bill.3 ”

•      Hypervisor and Operating System Choice.

•      Traditionally, IaaS offerings have been based on heavily customized open-source Xen deployments.

•      IaaS providers needed expertise in

•      Linux, networking, virtualization, metering, resource management, and many other low-level aspects to successfully deploy and maintain their cloud offerings.

•      More recently,

•      VMWare vCloud and Citrix Cloud Center (C3) which have lowered the barrier of entry for IaaS competitors, leading to a rapid expansion in the IaaS marketplace.

PLATFORM AS A SERVICE PROVIDERS

•      Offer a development and deployment environment that allow users to create and run their applications with little or no concern to low-level details of the platform.

•      specific programming languages and frameworks are made available

•      other services such as

•      persistent data storage and

•      In-memory caches.

•      Features

•      Programming Models, Languages, and Frameworks.

•      Each model aims at efficiently solving a particular problem.

•      In the cloud computing domain, the most common activities that require specialized models are:

•      processing of large dataset in clusters of computers (MapReduce model),

•      development of request-based Web services and applications;

•      definition and orchestration of business processes in the form of work- flows (Workflow model); and

•      High-performance distributed execution of various computational tasks.

•      For user convenience, PaaS providers usually support multiple programming languages.

•      Most commonly used languages in platforms include

•      Python and Java (e.g., Google AppEngine),

•      .NET languages (e.g., Microsoft Azure), and

•      Ruby (e.g., Heroku).

•      Force.com has devised its own programming language (Apex) and an Excel-like query language, which provide higher levels of abstraction to key platform functionalities.

•      Providers that focus on Web and enterprise application hosting offer popular frameworks such as Ruby on Rails, Spring, Java EE, and .NET.

•      Persistence Options.

•      A persistence layer is essential to allow applications to record their state and recover it in case of crashes, as well as to store user data.

•      Traditionally - relational databases as the preferred persistence method.

•      These databases offer fast and reliable structured data storage and transaction processing, but may lack scalability to handle several petabytes of data stored in commodity computers.

•      Cloud computing domain - distributed storage technologies have emerged, which seek to be robust and highly scalable, at the expense of relational structure and convenient query languages.

•      For example,

•      Amazon SimpleDB and Google AppEngine datastore offer

•      schema-less,

•      Automatically indexed database services.

•      Data queries can be performed only on individual tables; that is, join operations are unsupported for the sake of scalability.

CHALLENGES AND RISKS

•      Security, Privacy, and Trust

•      “Current cloud offerings are essentially public ... exposing the system to more attacks.”

•      Challenge - to make cloud computing environments as secure as in-house IT systems.

•      Data encryption, VLANs, and firewalls – can be used.

•      Security and privacy affect -> massive use of third-party services and infrastructures. So, the trust toward providers is fundamental.

•      Legal and regulatory issues –

•      physical location of data centers determines the set of laws that can be applied to the management of data

•      Specific cryptography techniques are not allowed in some countries.

•      Data Lock-In and Standardization

•      User data locked-in by a certain provider.

•      Users may want to move data and applications out from a provider that does not meet their requirements.

•      Currently, no standard methods of storing user data and applications.

•      Consequently, they do not interoperate and user data are not portable.

•      Answer àstandardization

•      The Cloud Computing Interoperability Forum (CCIF)

•      Unified Cloud Interface (UCI) - aims at creating a standard programmatic point of access to an entire cloud infrastructure

•      Open Virtual Format (OVF) - virtual appliances can be made portable

•      virtual appliances seamlessly run on hypervisor of different vendors

•      Availability, Fault-Tolerance, and Disaster Recovery

•      Users seek for a warranty (for Availability, Fault-Tolerance, and Disaster Recovery) before they can comfortably move their business to the cloud.

•      SLAs with QoS requirements must be set up between customers and cloud computing providers to act as warranty.

•      An SLA specifies the details of the service to be provided, including availability and performance guarantees.

•      Additionally, metrics must be agreed upon by all parties, and penalties for violating the expectations must also be approved.

•      Resource Management and Energy-Efficiency

•      Challenge faced by providers - efficient management of virtualized resource pools.

•      Challenge-1:

•      finding a good mapping of VMs onto available physical hosts

•      For mapping, multiple dimensions are to be considered. They are:-

•      number of CPUs, amount of memory, size of virtual disks, and

•      Network bandwidth.

•      Dynamic VM mapping policies pre-empts low-priority allocations in favour of higher-priority ones by using

•      Suspend, migrate, and resume VMs.

•      Additional Challenge (regarding Migration of VMs)

•      Detecting when to initiate a migration,

•      which VM to migrate, and

•      Where to migrate.

•      Note: - Policies may take advantage of live migration of virtual machines to relocate data center load without significantly disrupting running services.

•      Additional concern (regarding note)

•      trade-off between

•      the negative impact of a live migration on the performance and stability of a service

                                                And

•      the benefits to be achieved with that migration

•      Another challenge

•      Outstanding (large) amount of data to be managed in various VM management activities

•      Such amount of data is a result of particular abilities of virtual machines

•      ability of traveling through space (i.e., migration) and time (i.e., checkpointing and rewinding),

•      Operations that may be required in load balancing, backup, and recovery scenarios.

•      In addition,

•      Dynamic provisioning of new VMs and replicating existing VMs require efficient mechanisms to make VM block storage devices (e.g., image files) quickly available at selected hosts.

•      Data centers consumes large amounts of electricity

•      Example

•      USD 2.6 million per year ( given by HP)

•      1.3 MW of power for 100 server racks

•      1.3 MW are required by the cooling system

•      Risk

•      data centers significantly impact the environment in terms of CO2 emissions from the cooling systems

•      Note:-

•      Dynamic resource management improves utilization and consequently minimize energy consumption in data centers.

•      This can be done by consolidating workload onto smaller number of servers and turning off idle resources

Download links -   pdf        word