Cloud Computing (4)

Editor's Desk:
The first three parts of this series discussed cloud computing in terms of basic concept, applications, architecture, and computing model, and a comparison was made with other distributed computing technologies. In the final of this series, three cases are examined to determine key factors for success in cloud computing. Several economic and technical issues associated with cloud computing—including business model, reliability, security, privacy and developing trends—are discussed.

8 Case Study
    Cloud computing is still a new phenomenon. Although many IT giants are developing their own cloud computing infrastructures, platforms, software, and services, few have really succeeded in becoming cloud computing providers. Enterprises and research institutes have mostly been engaged in building their own experimental platforms. To help identify success factors in cloud computing operation, three cases can be examined[1-3].

8.1 Google
    Google was the first to initiate cloud computing, and its success can be attributed to its applications. To meet demands associated with its search engine, Google developed a unique large-scale distributed computing platform—a cloud computing platform—which can powerfully  analyze mass data. It is this powerful platform and analysis algorithm that propelled Google to  dominance in the search engine field.

    Google’s infrastructure can be divided into three layers:

• Products: search, ads, mail, map, video, chat, and blog
• Distributed system infrastructure: Google File System (GFS), MapReduce, and BigTable
• Computing platform: a large number of computers in many data centers.

    On March 9, 2010, Google opened its Google Apps Marketplace, thus introducing the concept of an "Apps Marketplace" to cloud computing. Its well-known applications Gmail, Docs, Sites, and Calendar gave Google the advantage of a large existing user group. Google Apps Marketplace is designed to deliver a variety of products and services to users, including applications that can be installed and synchronized with Google Apps. These are generally easy to use, supporting single sign-on, Google’s general navigation, and can be integrated with a user’s own data.

    Google Apps now features some novel cloud applications for enterprise. Intuit Online Payroll for Google Apps is one such application, which integrates the online payroll application of Intuit—a financial software developer—with Google. An employee need only click an icon in Google Calendar to download the payroll. Manymoon is another eye-catching application, integrated with Google account and Google Apps engine. It includes functions such as status message display and document editing and achieves fully socialized project collaboration. The operation model of Google Apps involves leveraging its large user group to attract developers to  create applications on its platform. Profits are then shared with developers. A three-way win is realized for Google, developers, and end users.

8.2 Amazon
    Amazon Elastic Compute Cloud (EC2) allows users to lease computers to run their own applications. Through an EC2 web service interface, a user can create a virtual machine, called an "instance," to run software. Scalable deployment of applications is thus achieved. A user can create, launch and terminate server instances as needed, paying by the hour for active servers.  Such flexibility gives rise to the term "elastic."

    As well as allowing for flexible configuration, Amazon’s cloud computing platform differs from Google’s in its operation model. Users are not only provided with bare machine (a virtual machine) on which to configure static IP addresses, but also absolute control over these so they can run their own application programs. Moreover, Amazon’s EC2 can be easily integrated with other Amazon web services, such as Simple Storage Service (S3) and SimpleDB. By configuring firewalls, Amazon EC2 controls network access between instances. A user can create an Amazon Virtual Private Cloud (VPC) and connect it to an enterprise’s IT infrastructure.

    Amazon’s success in cloud computing lies in its advanced network service technologies.  Although it could never have foreseen itself as a cloud computing provider at the beginning,   Amazon can now provide users with access to its enormous idle resources (used for e-commerce)  through advanced network service technologies. Consequently, its cloud computing platform  has been developed. Amazon’s cloud computing technology is developed on the basis of network computing services.

8.3 Salesforce
    Salesforce.com is a Software as a Service (SaaS) company that distributes business software and provides services for a monthly subscription fee. Its best-known product is Customer Relationship Management (CRM), which delivers services via the Internet using the SaaS model. CRM services cover all aspects of customer relationship management—from common contact management,  catalog and order management, and opportunity management, to sales management. In 2009, Salesforce was voted by Forbes magazine as one of the fastest growing technology companies,  second only to Google. It is the first cloud computing company with annual revenue of 1 billion U.S. dollars. At present, SalesforceCRM has 7.5 million users worldwide, including Google, Cisco, Starbucks, and Deutsche Bank’s Prestitempo Division. Two important platforms of Salesforce are Sales Cloud and Force.com.

    (1) Sales Cloud: The World’s Leading Sales Application
    Sales Cloud gives sales reps everything they need to complete their tasks, saving time spent on administration and leaving more time open for selling. It also provides sales managers with real-time visibility into a team’s sales activities. Service Cloud is a modern customer service platform, providing efficient and responsive services through channels as diverse as call centers and social networks.

    (2) Force.com: Cloud Platform for CRM
    Each enterprise has its own CRM solution, but force.com is a perfect CRM cloud platform. Because it resides within the cloud, CRM applications are never limited by underlying technology and can be customized in real-time depending on business conditions. Force.com AppExchange is an application directory created for Salesforce by third-party developers. Users can buy applications and add them to their Salesforce environment. In 2009, a total of 100,000 applications had been developed by more than 124,000 developers. Applications on Force.com can also be integrated with services delivered by providers such as Google.
The key to Salesforce’s success is its top sales application, which is highly regarded by many enterprises, and a cloud computing platform that helps Salesforce pack its applications into services and release them on the Internet. The cloud computing platform has contributed to the widespread success of Salesforce.

9 Some Issues Concerning Cloud Computing
    More than three years have passed since cloud computing was first introduced, and many enterprises, including IT giants, have begun deploying cloud computing environments, platforms, and services. On the one hand, cloud computing is developing rapidly, like a scudding cloud. On the other, debates about cloud computing are raging. Some questions need to be clarified. For instance, what exactly is the cloud? Compared with existing distribution technologies, what new features does cloud computing bring? What does the future of cloud computing look like?

    In this chapter, we will discuss issues concerning cloud computing as well as some perspectives on cloud computing development trends.

9.1 The Name "Cloud"
    Academia and enterprise view cloud computing from slightly different angles. It is interesting to note that China’s Cloud Computing Experts Association[4] is established under the Chinese Institute of Electronics (CIE) rather than the China Computer Federation (CCF). According to Ian Foster, cloud computing and grid computing are identical in terms of vision, architecture, and technology, but differ in terms of security, programming model, business model, computing model, data model, application, and abstraction[5-6]. Some Chinese scholars have coined the term "gloud" for cloud grid,  in an attempt to integrate cloud and grid computing. In the cloud-grid, cloud computing services are provided in a grid environment[7-8].

    As the technology develops, more names are likely to be coined. However, attention should not be paid to the definition of cloud computing or its features. Instead, the focus should be on whether a cloud computing environment can solve real computing problems, no matter what technologies or operational model is adopted. We should grasp the essence and forget the dross, as the Chinese story "Jiufang Gao Judges Horses" goes. In the warring states of China, Duke Mu of Qin sent Jiufang Gao in search of a swift horse. Three months later, Jiufang Gao reported that a swift horse had been found, a yellow mare in the dunes. However, a black stallion was sent back. The Duke was not satisfied. But Bo Le, a horse connoisseur, sighed and said: "What Jiufang Gao observes! He sees the essence but ignores the appearance." The horse was proven  unparalleled.

9.2 Business Model
    Advocators of cloud computing claim the pay-per-use business model is its biggest highlight. A user pays for a service based on how much of the resource is used. But this business model does not define what cloud is, and does not directly relate to the cloud computing system. Two important concepts in telecommunications network operation are Operations Support System (OSS) and Business Support System (BSS). Only BSS is related to the business model. Therefore, a pay-per-use business model does not involve cloud computing. Any system can adopt this business model as long as it can measure the resources consumed by each user. If the grid system is put into commercial operation, a pay-per-use business model can also be used.

    Popular opinion is favorable about cloud computing, and many research institutes and enterprises are in the process of building their own cloud computing platforms. However, some critics do not recognize private cloud and believe it does not exist. The reasoning is simple: If a user still has to buy, build, and manage a private cloud, it does not lower front-end investment and reduce management, which conflicts with claims made about the pay-per-use model[9].

    The pay-per-use business model is not the only success story. With its innovative user interface, Apple’s iPhone mobile has become popular. Users can purchase applications in Apple’s application store, called App Store. This model has also proven to be a successful business model.  Google Apps Marketplace now adopts the same model as Gmail's extensive user group.

9.3 Cloud Service Provider
    Another issue to be considered by cloud service providers is the profit-making model. Cloud computing has always been said to save user investment; but to date, no one has really figured out how cloud service providers should make profit. Again, a success story may prove useful in  determining how to make a profit.

 
    Google developed GFS and BigTable for storing mass data generated by its search engine. The MapReduce framework was developed for fast computing during the search process. Google’s profits are derived from huge advertising revenue generated by its search engine business.  Google has also located its data center next to a hydropower station to save energy costs. Amazon purchased many computers during the construction of its e-commerce platform, but later found that it did not need so many computers. To make full use of those idle devices and avoid waste, it packaged these devices into services by means of virtualization technology and provided them to users. Tencent is rumored to pay almost one hundred million US dollars in electricity per year.

    The success of Salesforce depends entirely on its sales application, which leads the world and is highly valued by customers. The story of Salesforce tells us that an enterprise should have a competitive product before it becomes a cloud service provider. In March 2010, Alibaba announced the close of its Internet platform, alisoft.com, and the termination of related services. Some critics question how long free SaaS can continue. In response to this question, it may be asked: Is CRM software good enough? Providing free services is obviously a fishing strategy. A service provider first gives certain benefits, and when users rely on these benefits, the service is charged. Of course, two premises are required: the software must be good enough, and the price must be reasonable. The former implies the software should improve business capability and profit enterprise. Otherwise, even if the software is free, it will not be widely adopted. The  latter requires consideration of a user’s profit margin to determine whether it is being shrunk, and whether the service fee is bearable.

    IBM established the first Chinese cloud computing center in Wuxi, Tai Hu New Town Science and Education Industrial Park. So far, no enterprise is known to be seeking storage services from the cloud computing center. This demonstrates that caution should be exercised if an enterprise or institute wants to become a cloud computing operator. As well as having an advanced distributed computing platform and excellent service software, it must limit operating cost in order to provide users with high-quality low-priced cloud services. Seeking quick success or windfall profits is detrimental to the development of Internet applications.

9.4 Reliability, Security and Privacy
    According to a news report in 2008, Amazon’s S3 became unavailable due to authentication service overload—a failure that lasted for two hours. Later, a Gossip protocol blowup led to an outage of S3 for almost eight hours. Gmail and Google Apps once suffered downtime of several hours due to contact system failure and program errors.  When users find the cloud has disappeared, their businesses are inevitably affected. Such  failures also occur in common computing  from time to time. Reliability and availability problems are not confined to cloud computing. On the contrary, in a cloud computing environment, redundancy is an important feature that greatly improves reliability and availability. 

 
    All security information technologies can be used in the cloud computing environment. The cloud computing environment is no less safe than common computing systems. Like money-keeping,  it is safer to deposit money in the bank than to keep it under the pillow. When people talk about network security, they are especially concerned about privacy. Americans, among others, are  concerned their private data may be seized by the government by means of a court order if that data is stored in a cloud. One solution to this problem involves building an InterCloud containing clouds of all operators as well as private clouds of other enterprises. When data is stored in the large cloud, others do not know which storage devices are used or the clouds in which the data  is stored. Chinese companies may be more concerned with data confidentiality than privacy and are perhaps reluctant to save their data in clouds of service providers. This is actually a trust problem. If you have no trust in the bank, you can do nothing but put your money under the pillow. Security technologies are designed to prevent attacks from hackers, and solving the trust problem requires everyone to work together to build a healthy business environment.

9.5 Future Cloud Computing
    Computing as a utility has been a long-held dream. Cloud computing aims to make this dream come true. Before discussing the development trends of cloud computing, a look at Google’s future objectives might be insightful:

• To support geographically distributed clusters.
• To create a globally unique name space for data. Currently, data is isolated by clusters.
• To provide more and better automated migration of data and computation.
• To solve the consistency problem that occurs in coupling data copies split by wide area networking, and to ensure services are delivered normally even when a cluster is down due to maintenance or other reasons.

    From these objectives, it can be seen that building large-scale geographically distributed infrastructure is still important in the field of cloud computing. However, Google’s MapReduce framework still has some limitations. Although the map and reduce functions are very suitable for large-scale parallel computing, not all computing can be easily converted into a MapReduce framework. As a result, parallel computing algorithm for general problems remains a major challenge[10].

    The success of Salesforce demonstrates that users really want solid applications that can solve practical problems. Therefore, one development trend of cloud computing is to develop new cloud computing applications, such as Facebook, that can provide users with rich experience.  Another area of interest is the integration of diverse services with a cloud computing infrastructure platform.

    From the perspective of cloud providers, using commercial computing, storage, and network facilities to build a large-scale data center and then to sell these resources using a pay-per-use model may cost less but be more profitable than building a medium-sized data center. This is because a large data center can be shared by large user groups during different periods. From a user’s point of view, new software companies can use cloud infrastructure as a starting point to build their own data centers. Overload or insufficiency risks suffered by traditional service methods can be avoided, where internal data centers are used to provide services. Other companies or organizations also benefit from the "elasticity" of cloud computing.

    Cloud exchange and market infrastructure may be the most suitable option for the future of cloud computing. In such an infrastructure, an enterprise or organization is not just a cloud computing provider or user. It builds its own cloud computing platform (private cloud) based on its average computing services in order to drive the development of its own business. But it also  merges this private cloud into the global cloud to form a large Internet cloud. This mode of development is similar to the Internet: As network technology arose, many enterprises built their own networks. Later, with the introduction of Transmission Control Protocol/Internet Protocol (TCP/IP), those isolated networks were interconnected to form the Internet. In this development mode, an enterprise may use other cloud resources in the Internet when its own computing resources are overloaded. It can also open its idle resources to other users. The resource can be priced according to a market mechanism. An enterprise pays others to use resources, and in turn, gets paid by providing its resources. Another benefit of Internet cloud is the prevention of  monopolistic behavior due to over-dependence on a provider. Only a win-win philosophy can make cloud computing a success.

    In summary, cloud computing involves integrating all resources, including hardware, software, platforms, applications, and services. Simple interfaces must be provided for easy use of cloud resources. 

References
[1] Cloud Computing [EB/OL]. http://zh.wikipedia.org or http://en.wikipedia.org [2010-02-05].
[2] LEBK A, KLEMS M, NIMIS J, et al. What’s Inside the Cloud? An Architectural Map of the Cloud Landscape [C]//Proceedings of the 2009 ICSE Workshop on Software Engineering Challenges of Cloud Computing (CLOUD’09), May 23, 2009, Vancouver, Canada. Washington, DC, USA: IEEE Computer Society, 2009: 23-31.
[3] HAYES B. Cloud Computing [J]. Communications of the ACM, 2008, 51(7): 9-11.
[4] China Cloud Computing [EB/OL]. http://www.chinacloud.cn, [2010-04-03].
[5] FOSTER I, ZHAO Y, RAICU I, et al. Cloud Computing and Grid Computing 360-Degree Compared [C]// Proceedings of the 2008 Grid Computing Environments Workshop (GCE’08), Nov 12-16, 2008, Austin, TX, USA. Washington, DC, USA: IEEE Computer Society, 2008: 10p.
[6] GEELAN J. Twenty One Experts Define Cloud Computing [J/OL]. Cloud Computing Journal (2009-01-24). http://cloudcomputing.sys-con.com/node/612375.
[7] ZHANG Liangjie,ZHOU Qun. CCOA: Cloud Computing Open Architecture [C]// Proceedings of the IEEE 7th International Conference on Web Services (ICWS’09), Jul 6-10, 2009, Los Angeles, CA, USA. Washington, DC, USA: IEEE Computer Society, 2009: 607-616.
[8] ARMBRUST M, FOX A, GRIFFITH R, et al. Above the Clouds: A Berkeley View of Cloud Computing [R]. Technical Report UCB/EECS-2009-28. Electrical Engineering and Computer Sciences, University of California, Berkeley. 2009.
[9] BUYYA R, YEO Chee Shin, VENUGOPAL S, et al. Cloud Computing and Emerging IT Platforms: Vision, Hype, and Reality for Delivering Computing as the 5th Utility [J]. Future Generation Computer Systems,2009, 25 (6): 599-616.
[10] DEAN J, GHEMAWAT S. MapReduce: Simplified Data Processing on Large Clusters [J]. Communications of the ACM, 2008, 51(1): 107-113.