Cloud Computing Architecture: Part 1

Cloud Computing Architecture:  Part I

Updated in September 2010 by Dexter Duncan.

(Article submitted to magazine targeting CIOs and originally published  in September 2009.)

The latest craze in the IT (Information Technology) sector is Cloud Computing.   Some call it the new frontier.  The Wild West is here, bringing with it opportunities for new companies.  Conferences on this topic are already in their third or fourth year. Companies in this space are attracting venture capital funds even during the Global Financial Crisis.  IDC estimates the market size for Public Cloud Computing to be at USD $56 Billion by 2014, with an annual growth rate of 25%, or five times the regular IT growth rate. Extending this to 2020, the market size for Cloud Computing stands at an astronomical $213 Billion. With the growing shift in application software development from PC to Data Centers and Clouds, we expect that this opportunity will increase even higher.  With these projections, so many vendors and service providers have got into the game and so much press activity has taken place. Gartner’s 2009 IT Hype Cycle for Emerging Technologies, released this July, places “Cloud Computing” at the peak of the “hype cycle”. The lead adopter phase is over and a period of uncertainty, innovation and consolidation has begun as companies find practical uses for Cloud Computing.

Cloud computing is often called the 5th Utility [2].   Utilities such as water, gas, and electricity are fundamental in our daily life and are exploited on a pay per use basis. The existing infrastructures deliver these services almost anywhere and anytime.  The usage of these utilities is charged, according to different policies, to the end user.  The same idea of utility is applied to computing and a consistent shift towards this approach is seen with the proliferation of Cloud Computing.

Cloud Computing promises to deliver on- demand IT resources on a pay-per-use basis by dynamically growing or shrinking the virtualized resources and providing them as a service to the users over the internet. This is aimed at a global market with massive demands in real time, ranging from the end users that host their personal documents on the Internet, to enterprises outsourcing their entire IT infrastructure to external data centers. The approach of Cloud Computing makes the notion of IT as a fifth utility a reality: not only computing and storage resources are delivered on demand but the entire stack of computing is offloaded to the Cloud.

CIOs should ask: how does Cloud Computing help my organization?  Can this new commoditized infrastructure give me a competitive advantage in delivering services or does it provide new flexibility?   Many of the advantages of Cloud Computing are a culmination of advantages derived from the multiple technologies that drive it – such as Grids, Clusters, Virtualization, Utility (e.g. pay-per-use), Service Oriented Architecture and Quality of Service needs.

Some of the most famous names in Cloud Computing are giants such as IBM, Microsoft, Google and Sun Microsystems (now part of Oracle).  In addition, a number of IT vendors, including HP, Dell, EMC (VMWare), Cisco, Citrix, RedHat, AT&T, Unisys, Oracle, and small to mid-size start-ups have made product and service announcements addressing the Cloud Computing space.  Even Amazon, the bookseller, opened up its datacenters for business by selling compute and storage “on-demand” and includes this new service as one of its best revenue growth areas.

What is Cloud Computing and who are the teams playing in this space?

The CEO of Microsoft, Steve Balmer, reportedly told a reporter in New York in March 2009, “Anything that has been a server needs to be a service.”     His statement is a good summary of Microsoft’s Azure strategy, but only partly covers the definition of Cloud Computing.  A new framework race is on that allows enterprises to manage based on the demand based commodity infrastructure.  The framework needs to make it easy to manage, provision, monitor, secure and change resources when needed or based on pre-determined service levels that match the applications.

Cloud Computing Reference Model

Cloud Computing is a type of parallel and distributed system consisting of a collection of interconnected and virtualized computers that are dynamically provisioned and presented as one or more unified computing resources based on a service-level agreement [1]. According to this definition, Cloud Computing refers to the practice of delivering software and infrastructure as a service on a “pay as you go” basis.

Figure 1 illustrates the layered architecture of the Cloud Computing. Generally, the Cloud architecture defines 4 distinct layers from physical hardware to end user applications. Physical resources including spare desktop machines, clusters and datacenters form the lowest level of the stack, on top of which the virtual infrastructure is deployed. Infrastructure supporting large scale Cloud deployments are more likely datacenters hosting hundreds or thousands of machines, while small scale, in house Clouds provide a more heterogeneous scenario where the idle CPU cycles of spare desktop machines are used to leverage the compute workload. This level provides the “horse power” of the Cloud.

Figure 1: Cloud Computing layered architecture.

For full article, click here.

Cloud Start

By: Christian Vecchiola

Melbourne, Australia

Cloud Computing Reference Model and Technologies

In order to introduce a reference model for Cloud Computing, it is important to provide some insights on the definition of the term Cloud. There is no universally accepted definition of the term.  Folks from University of Cal-Berkeley [1]  notice that “Cloud Computing refers to both the applications delivered as services over the Internet and the hardware and system software in the datacenters that provide those services”. They  identify the Cloud with both the datacenter hardware and the software. A more structured definition is given by group at University of Melbourne [2] who define a Cloud as a “type of parallel and distributed system consisting of a collection of interconnected and virtualized computers that are dynamically provisioned and presented as one or more unified computing resources based on service-level agreement”. There is an agreement on the fact that Cloud Computing refers to the practice of delivering software and infrastructure as a service, eventually on a pay per use basis. In the following, we  illustrate how this is accomplished by defining a reference model for Cloud Computing.

The above gives a layered view of the Cloud Computing stack. It is possible to distinguish four different layers that progressively shift the point of view from the system to the end user. The lowest level of the stack is characterized by the physical resources on top of which the infrastructure is deployed. These resources have different natures: clusters, datacenters, and spare desktop machines. Infrastructure supporting commercial Cloud deployments are more likely to be constituted by datacenters hosting hundreds or thousands of machines, while private Clouds can provide a more heterogeneous scenario in which even the idle CPU cycles of spare desktop machines are used to leverage the compute workload. This level provides the “horse power” of the Cloud.

The physical infrastructure is managed by the core middleware layer whose objectives are to provide an appropriate run time environment for applications and to exploit the physical resources.  To provide advanced services, such as application isolation, quality of service, and sandboxing, the core middleware  relies on virtualization technologies. Among the different solutions for virtualization, hardware level virtualization and programming language level virtualization are the most popular. Hardware level virtualization guarantees complete isolation of applications and a fine partitioning of the physical resources, such as memory and CPU, by means of virtual machines. Programming level virtualization provides sandboxing and managed execution for applications developed with a specific technology or programming language (i.e. Java, .NET, and Python). On top of this, the core middleware provides a wide set of services that assist service providers in delivering a professional and commercial service to end users. These services include: negotiation of the quality of service, admission control, execution management and monitoring, accounting, and billing.

Together with the physical infrastructure the core middleware represents the platform where the applications are deployed in the Cloud. It is very rare to have direct access to this layer. More commonly, the services delivered by the core middleware are accessed through a user level middleware. This provides environments and tools simplifying the development and the deployment of applications in the Cloud: web 2.0 interfaces, command line tools, libraries, and programming languages. The user level middleware constitutes the access point of applications to the Cloud.

The Cloud Computing model introduces several benefits for applications and enterprises. The adaptive management of the Cloud allows applications to scale on demand according to their needs: applications  dynamically acquire more resource to host their services to handle peak workloads and release when the load decreases. Enterprises do not have to plan for the peak capacity anymore, but provision as many resources as they need, for the time they need, and when they need it.  Moreover, by moving their IT infrastructure into the Cloud, enterprise reduce their administration and maintenance costs. This opportunity becomes even more appealing for startups, which start their business with a small capital and increase their IT infrastructure as their business grows. This model is also convenient for service providers that want to maximize the revenue from their physical infrastructure. Besides the most common “pay as you go” strategy more effective pricing policies can be devised according to the specific services delivered to the end user. The use of virtualization technologies allows a fine control over the resources and the services that are made available at runtime for applications. This introduces the opportunity of adopting various pricing models that benefit either the customers or the vendors.

The model endorsed by Cloud Computing provides the capability of leveraging the execution of applications on a distributed infrastructure that, in case of public clouds, belongs to third parties. While this model is certainly convenient, it also brings additional issues from a legal and a security point of view. For example, the infrastructure constituting the Computing Cloud is made of datacenters and clusters located in different countries where different laws for digital content apply. The same application is considered legal or illegal according to the where is hosted. In addition, privacy and confidentiality of data depends on the location of its storage. For example, confidentiality of accounts in a bank located in Switzerland may not be guaranteed by the use of data center located in United States. In order to address this issue some Cloud Computing vendors have included the geographic location of the hosting as a parameter of the service level agreement made with the customer. For example, Amazon EC2 provides the concept of availability zones that identify the location of the datacenters where applications are hosted. Users have access to different availability zones and decide where to host their applications. Since Cloud Computing is still in its infancy the solutions devised to address these issues are still being explored and will definitely become fundamental when a wider adoption of this technology takes place.

(edited by: Dexter Duncan)

REFERENCES:

[1]  M. Armbrust, A. Fox, R. Griffith, A.D. Joseph, R. Katz, A. Konwinski, G. Lee, D. Patterson, A. Rabkin, I. Stoika, M. Zaharia, Above the Clouds: a Berkeley view of Cloud Computing, Technical Report, UC Berkeley Reliable Adaptive Distributed Systems Laboratory, available at http://abovetheclouds.cs.berkeley.edu

[2]  R, Buyya, C.S. Yeo, S. Venugopal, J. Broberg, I. Brandic, Cloud Computing and emerging IT platforms: vision, hype, and reality for delivering IT services as the 5^th utility, Future Generation of Computer Systems, 25 (2009), 599–616.