The Two Eras Of Computing

Published Date: 02 Nov 2017

It is clear that silicon based processor chips are coming to their physical limits in processing speed. A viable answer to overwhelm this limitation is to attach multiple processors employed in coordination with each other to explain large dispute problems. Hence, high performance computing needs the use of Massively Parallel Processing (MPP) systems encompassing thousands of mighty CPUs. A superior agent computing system constructed utilising MPP set is C-DAC’s PARAM supercomputer.

By the end of this century, all high performance systems will be parallel computer systems. High-end super computers will be the Massively Parallel Processing (MPP) systems having thousands of processors interconnected. To present well, these parallel systems need an functioning system fundamentally distinct from present ones. Most investigators in the area of functioning systems have discovered that these new functioning systems will have to be much lesser than customary ones to accomplish the effectiveness and flexibility needed.

Eras of Computing

The most famous two eras of computing are, i) sequential and ii) parallel era. In the past ten years, parallel machines have become important competitors to vector machines in the quest for high performance computing. A 100 years broad outlook of development of computing eras is shown in Figure 1. The computing era begins with a development in hardware architectures, pursued by system programs, applications and coming to its saturation point with its development due to difficulty in environments. Every component of computing undergoes three phases: R & D, commercialization and commodity.

fig1.TIF

Fig 1: Two Eras of Computing

Cloud Computing refers to both the applications consigned as services over the Internet and the hardware and systems programs in the data hubs that supply those services. When a Cloud is made accessible in a pay-as-you-go kind to the general public, it can be called as a Public Cloud, the service being traded is Utility Computing. Cloud Computing is the addition of SaaS and Utility Computing.

People can be users or providers of SaaS, or users or providers of Utility Computing. From a hardware issue of outlook, three facets are new in Cloud Computing.

1.  The illusion of infinite computing assets accessible on demand.

2.  The elimination of an up-front firm promise by Cloud users.

3.  The proficiency to pay for use of computing assets on a short-term basis as required and release them as needed.

Infrastructure services (Infrastructure-as-a-service), supplied by cloud vendors permit any client to use large number of compute examples effortlessly by utilizing the virtual assets to present data/compute intensive works.

1.2 High Performance Parallel Computing with Clouds and Cloud Technologies

The introduction of financial / commercial cloud infrastructure services for example Amazon EC2/S3, GoGrid and ElasticHosts permit users to use clusters effortlessly and rapidly by giving a monetary worth only for the duration used. The provisioning of assets occurs in minutes.

The accessibility of open source cloud infrastructure softwares for example Nimbus and Eucalyptus and the open source virtualization software’s for example Xen Hypervisor permits organizations to construct private clouds to advance the asset utilization of the accessible computation facilities. The likelihood of dynamically provisioning added assets by leasing from financial / commercial cloud infrastructures makes the use of private clouds more promising.

With all the overhead characteristics of cloud, the accessibility to computation power in cloud is no longer a barricade for the users who require to present large scale data/compute intensive applications. However, to present such computations, two foremost preconditions require to be satisfied: (i) the application should be parallelizable to utilize the accessible assets and (ii) there should be an befitting parallel runtime to apply it.

We have some cloud technologies for HPC for example Hadoop, Dryad and CGL-MapReduce, to diverse technical applications for example, (i) Cap3 data investigation (ii) High Energy Physics(HEP) data investigation (iii) Word Histogramming (iv) Distributed Grep and (v) Kmeans clustering.

Cloud technologies for example Google MapReduce, Google File System (GFS) , Hadoop and Hadoop Distributed File System (HDFS) , Microsoft Dryad and CGL-MapReduce take up a more data-centered set about to parallel runtimes. In these structures, the data is arranged in data/compute nodes of clusters or large-scale data centers. The computations proceed to the data in alignment to present data processing. Distributed document systems for example GFS and HDFS permit Google MapReduce and Hadoop to get access to data by distributed storage systems constructed on heterogeneous compute nodes, while Dryad and CGL-MapReduce support reading data from localized disks. The ease in the programming form endows better support for value of services for example obvious error tolerance and monitoring.

1.3 Cloud Computing Application Platforms

Cloud computing is a foremost change in IT industry. One of the most significant components of that change is the advent of cloud platforms. As its title proposes, this kind of platform permits developers to compose applications that run in the cloud, or use services supplied from the cloud or both. This new way of applications has large potential.

Similarly, if every development group that desires to conceive a cloud application should first construct its own cloud platform. Fortunately, vendors are increasing to this dispute and several cloud platform technologies are accessible today.

fig2.TIF

Fig: 2. Cloud services assembly

To understand about cloud platforms, it’s helpful to start by looking at cloud services in general. Figure 2 displays that services in the cloud can be grouped into three very broad categories. Those classes are , i) Software as a service (SaaS): A SaaS application sprints solely in the cloud. The on-premises purchaser is normally a browser or some other straightforward client, ii) Attached services: Every on-premises application presents helpful purposes on its own. An application can occasionally enhance these by accessing application-specific services supplied in the cloud, and iii) Cloud platforms: A cloud platform presents cloud-based services for creating applications. Rather than construction their own made-to-order base, for example, the creators of a new SaaS application could rather than construct on a cloud platform.

Figure 2 displays that the direct users of a cloud platform are developers, not end users. Understanding cloud platforms needs some affirmation on what the phrase "platform" entails in this context. One very broad way to believe about it is to outlook a platform as any softwares that presents developer-accessible services for creating applications.

1.3.1 General Model for Application Platforms

Now-a-days application platforms arrive mostly from on-premises platforms. A helpful way to believe about cloud platforms is to glimpse how an application developer relies his services in the on-premises natural environment convert to the cloud. Figure 3 displays a general model that can be directed to both worlds. Whether it’s on-premises or in the cloud, an application platform can be considered of as comprising three parts: i) Foundation, ii) Infrastructure services and iii) Application Services.

fig3.TIF

Fig: 3 A up to date application platform

Even though these applications exist mainly to supply services to end users, this furthermore makes them part of the application platform.

Development tools are another significant part in platforms. Modern tools can assist developers construct applications utilising all three components of an application platform. To make this abstract form more solid, on-premises platforms are found. The on-premises base examines like this: i)Operating system, ii)Local support.

For on-premises infrastructure services, usual examples are: i)Storage, ii)Integration, iii) Identity.

On-premises application services, the third class shown in Figure 3, vary broadly over distinct organizations. The cause for this is straightforward, distinct organizations use distinct applications, which in turn reveal varied services. One way to believe about these applications in the on-premises platform is to split up them into two very broad categories: i) Packaged applications and ii) Custom applications.

1.3.2 On-Premises Platforms To Cloud Platforms

Along with recounting on-premises platforms, the general model just recounted can furthermore be utilised for cloud platforms. And since on-premises and cloud platforms can be utilised simultaneously, it’s significant to realise how the two work in concert. Figure 4 shows this new world.

fig4.TIF

Fig: 4. On-premises platforms and cloud platforms

Figure 4 displays that a cloud application can be construct on a cloud base, just as an on-premises application is construct on an on-premises foundation. Both types of application can get access to infrastructure and application services supplied on-premises and in the cloud. Just as on-premises platforms support today’s applications, cloud platforms supply services for the applications were probable to construct tomorrow.

  

1.4 Cloud Computing Platforms

1.4.1 Abicloud Cloud Computing Platform

Abicloud is a cloud computing platform evolved by Abiquo, a business established in Barcelona Spain, focusing on the development of cloud platform. It can be utilised to construct, incorporate and organise public as well as personal cloud in the homogeneous environments. Using Abicloud, client can effortlessly and mechanically establish and organise the server, storage system, mesh, virtual apparatus and applications and so on. The major distinction between Abicloud and other cloud computing platforms is its mighty web-based administration function and its centre encapsulation manner. Using the Abicloud, client can complete establishing a new service by just pulling a virtual appliance with mouse. This is much simpler and flexible than other cloud computing platforms that establish new services through order lines.

1.4.2 Eucalyptus Cloud Platform

Eucalyptus is an elastic computing structure that can be utilised to attach the users' programs to the helpful systems, it is an open-source infrastructure utilising clusters or workstations implementation of elastic, utility, cloud computing and a well liked computing benchmark founded on service grade protocol that allow users lease mesh for computing capability. Currently, Eucalyptus is matching with EC2 from Amazon and support more other types of purchasers with smallest modification and extension.

1.4.3 Nimbus Cloud Computing Platform

Nimbus is an open tool set and furthermore a cloud computing solution supplying IaaS. Based on technical study in the early platform, Nimbus have sustained numerous non-scientific study domain applications. It allows users lease isolated assets and construct the needed computing natural environment through the deployment of virtual machines.

Nimbus cloud computing platform encompasses numerous distinct constituents, state purchaser, agency and resource supervisor and so on. Generally, all these functional components can be classified as three kinds. One kind is client-supported modules which are utilised to support all types of cloud clients. Context client module, cloud client module, quotation client module and EC2 client module are all pertains to this kind of component. The second kind of component is mostly service-supported modules of cloud platform, supplying all types of cloud services. It encompasses context agency module, world wide web service resource structure module, EC2 WSDL module and isolated interface module. The third kind of compoenent is the backdrop resource administration modules which are mostly utilised to organise all types of personal assets on the cloud computing platform, encompassing work service administration module, IaaS module, EC2 and other cloud platform support module, workspace navigate module, workspace asset administration module and workspace controller.

1.4.4. OpenNebula Cloud Computing Platform

OpenNebula is one of the key technologies of reservoir design and the flagship study task in virtualization infrastructure and cloud computing of European Union. Like nimbus, OpenNebula is furthermore an open source cloud service framework. It permits client establish and organise virtual machines on personal assets and it can set user’s data hubs or clusters to flexible virtual infrastructure that can mechanically acclimatize to the change of the service load. The major distinction of OpenNebula and nimbus is that nimbus applies isolated interface founded on EC2 or WSRF through which client can method all security associated matters, while OpenNebula does not.

OpenNebula is furthermore an open and flexible virtual infrastructure administration device, which can use to synchronize the storage, mesh and virtual methods and let users dynamically establish services on the circulated infrastructure as asserted by the share systems at data center and isolated cloud resources. Through the central interfaces and OpenNebula data center natural environment, users can effortlessly establish any kinds of clouds.

OpenNebula is mostly utilised to organise the data center of private cloud and infrastructure of cluster and it furthermore support hybrid cloud to attach the localized and public infrastructure. This is very helpful to construct high scalable cloud computing environment. Besides, OpenNebula furthermore carries public cloud platform by supplying interfaces and purposes to virtual machines, storage and mesh administration and so on.

1.5 Tools for Building Clouds

Today, two development methodologies are conveying many to cloud-based development: i) distributed and ii) agile development. These notions are impelling the wrapper on existing development apps, needing a new set of devices that can accommodate new development, checking and deployment methods.

  Distributed development is the natural by-product of the Internet. Distributed development is global development, which adds its own trials with collaboration and code management. There are large applications currently out there for distributed code administration, i) git and ii) subversion are two such tools and are broadly utilised in distributed environments already.

  These online code administration tools only notify part of the article, however. Even in a distributed development environment, programmers are still dragging down code to their localized machines, modifying it, then uploading back into the repository to impel it along to anything step in the merchandise workflow is next.

  Developers can proceed in the direction of a more collaborative work procedure by integrating browser-based incorporated development environments (IDEs) into their toolset. These interfaces endow developers to code out in browser space, examples encompass Ace, Coderun Studio, Codemirror and Ymacs.

1.5.1 Open source tools for construction and organising clouds

Open source expertise is going to gravely influence the cloud computing world and there are two major causes why: Open source software is vitally free and it is not usually encumbered by the software license of proprietary software.

A number of open source tools have currently had a large influence on cloud computing: Linux and Xen, for example. But there are other significant open source offerings that can advantage cloud users. These encompass KVM, Deltacloud, Eucalyptus, Cloud.com's CloudStack.

There are eight key components to address when construction an internal or external compute cloud. They are i) Shared Infrastructure, ii)Self-Service Automated Portal, iii)Scalable, iv)Rich Application Container, v)Programmatic Control, vi)100% Virtual Hardware Abstraction, vii)Strong Multi-Tenancy, and viii)Chargeback.

1.6 Programming in Cloud

Cloud computing comprises of two aspects of meaning, to recount the rudimentary platform amenities and, on the other hand, to construct applications on this platform.

First, cloud computing-based platform for programming is built. There are numerous programs which can be utilised to construct a basis for cloud computing platform programming. For example, customary distributed systems and grid platform can be utilised as the basis part of cloud computing platform facility and then construct up an whole cloud computing platform by abstraction on such a structure.

1.6.1 MapReduce Distributed Programming

MapReduce is a mighty distributed programming procedure, which is furthermore a function of dialect form utilised to deal with huge data groups and in which, only two functions are provided: Map and Reduce.

Map function presents a dedicated procedure for each data item set and comes back a new data set after disposing of the project. In a data item, a Reduce function gets implementation of the aim procedure in the data piece sets.

1.6.2 Chubby

Chubby is a highly accessible, distributed data secure service. When any machines falls short, Chubby double-checks the consistency of the backup utilising the Paxos algorithm. Each unit in little distributed document systems of Chubby can be utilised to supply secure services. Currently, this language is mostly utilised on the basis of Google’s cloud computing platform. Hadoop and Pig language which is constructed on peak of Hadoop is a kind of open-source implementation for programs model.

1.6.3 Dryad and DryadLINQ

Dryad and DryadLINQ created by Microsoft Research Silicon Valley, is created to supply a distributed computing platform. In recent years, this platform has been broadly utilised internally at Microsoft, in specifically utilised on Microsoft’s own cloud computing platform Azure.

Dryad is created to be extending in the computing platforms of all dimensions, from single-core computers, to small cluster of the composition of a multiple computers and then to having data hubs comprised of thousands of computers. The aim of DryadLINQ is to supply a high-level language interface, in order programmers can effortlessly convey out large-scale distributed computing. Two key technologies, namely, Microsoft Dryad and LINQ are combined by DryadLINQ.

Dryad can develop distributed operations performed on Dryad engines and is responsible for self-acting parallel processing of the jobs and sequence of fuctions required when data is delivered. In supplement, it boasts a variety of easy-to-use and sophisticated characteristics, for example strongly drafted data and Visual Studio incorporated debugging.

1.6.4 Programming of Dynamics Languages

Computing assets can extend dynamically as asserted by the dimensions of the task, since computing platform has a solid high expansion flexibility and platform abstraction. As a outcome, programs can run without being influenced by the influence of infrastructure change. The fast and well inspired way for enterprise applications are to use programming procedure of dynamic language. Therefore, not only the code created by dynamic language can be established to achieve enterprise applications in the Cloud’s Client, but also cloud projects accomplished by dynamic languages can be discovered in parts of cloud infrastructure.

In the existing cloud computing platform, there are still adversities in integration and interoperability of services and applications, they still desire to take benefit of services supplied by cloud computing and incorporate the services they need.