A Case for Cloud-Based Mobile Search

Release Date:2011-03-18 Author:Yan Gao, Li Fu, Zhenwei Zhang, Shengmei Luo, and Ping Lu Click:

 

1 Introduction
    The advantages of cloud computing as an IT infrastructure are becoming more apparent as Internet services develop. Cloud computing services follow a scalable delivery and usage model, which means services can be requested on demand and can be customized. Networks that provide resources are called clouds. Clouds form a kind of virtual data storage and resource pool on a web-based platform and these resources can be accessed by users.


    There are three key cloud service models [1]: Software as a Service (SaaS), Platform as a Service (PaaS), and Infrastructure as a Service (IaaS). The core concept of cloud computing is to provide on-demand services and to make terminals into mere input/output devices by improving cloud processing capacity and reducing terminal processing load.


    As an extension of Internet search technology, mobile search is a generic term for mobile-based search technology. It refers to the acquisition of on-demand information and services (Wireless Application Protocol (WAP) sites, Internet information, mobile value-added service, or local information) on a mobile terminal via Short Message Service (SMS), WAP, Interactive Voice Response (IVR) and other access modes [2].


2 Analysis of Mobile Search and Cloud Computing

 

2.1 Mobile Search Services Overview 
    Mobile search involves providing services via mobile search engines. These acquire information from user input and integrate information from different providers to build relationships between the two. Processing information on an engine means that information it can be made suitable for a mobile terminal. Compared with traditional Internet search, mobile search has the following advantages:


    (1) More Flexibility: Users are no longer restricted by a fixed terminal and can search anywhere and anytime [3].


    (2) Accurate Results: Mobile search is focused on simplicity and effectiveness. These features require the search engine to have stronger capacity for natural language analysis and to provide accurate results.


    (3) Diverse Applications


    Mobile search should be more like a basic capacity than a service. It has been widely used for entertainment data filing, travel and managing personal information [3]. Combining diverse features boosts the development of popular applications. Since the multiservice development model emerged, isolation of the mobile search services, and flexibility of search traffic have become clear requirements.


Many technical problems are encountered in the development of mobile search. These include: 
    (1) Data Expansion 
    For personalized search, more data from user access logs and other user-associated information needs to be recorded. These demand greater storage and processing capability in mobile phones.


    (2) Limited Processing Power
    Providing accurate results means increasing the processing workload of search technology.  Enhancing user experience, processing search data, and correlation analysis all require quick and powerful computing.


    (3) Information Security 
    For mobile search, the basic requirement of multiservice application is security. This is essentially different from Internet search technology.


    (4) Service Flexibility 
    Service scale differs considerably during different stages of development. Fluctuations require powerful service processing capacity for a mobile search engine. As well as cost and energy saving, service flexibility must be taken into account.

 

2.2 Cloud Computing in Mobile Search
    Key areas of mobile search development are mobile network, search technology, and user end.  The emergence of cloud computing infuses a powerful driving force for the development of mobile search.


    (1) Massive Data Storage 
    Cloud computing provides a secure and reliable data storage center for both storage and management. A cloud-based distributed network with scalable architecture not only provides mass data storage capacity but also takes advantage of storage capacity in the server itself. So computing and storage size are upgraded together, and this can significantly improve system reliability, availability, efficiency, and scalability.


    (2) Parallel Computing 
    Parallel computing provides powerful computing for mobile search applications. MapReduce is a data processing solution based on mass data storage. Computing tasks can be decomposed by the MapReduce programming mode. Real-world computing tasks can be abstracted and then implemented by a partition statute using this programming model. Parallel computing is the solution to mass data storage in mobile search.


    (3) New Solutions for Mobile Security 
    Cloud computing is combined with parallel computing, grid computing, and other emerging technologies and concepts. In applying cloud computing, security must be guaranteed for multiservice applications in the basic cloud structure. Distributed storage and distributed database are key technologies for which service security is a key requirement. Virtualization can be used to build a hardware and software wall for securing services [4]. 


    (4) Flexible Platform 
    In cloud computing, mass resources are converged and integrated through virtualization in the resource layer, platform layer, and application layer, and through physical distributed integration. The cloud is not just a simple collection of resources. There is also a management mechanism so that the whole system can serve as a virtual pool for providing services and provide a flexible platform for multilevel applications of mobile search.
The cloud server generally processes complex calculations while mobile terminals are responsible for interaction with users. This reduces the processing workload in mobile devices and SaaS is implemented as well. Cloud computing also greatly reduces network requirement of mobile search [5].


3 Solution of  Cloud-Based Mobile Search
    In the following, a solution for cloud-based mobile search technology is introduced.

 

3.1 Architecture
    In the architecture for the solution (Fig. 1), cloud computing provides storage computing capability and virtualization environment support. The core features—collection, index, search, and managing—of mobile search are designed and applied using cloud technologies. The system can be deployed as a basic search structure on a mobile Internet platform or as an independent mobile search service.

 

 

3.2 Key Technologies of Cloud Computing in the Solution
    The key techologies for the solution include:


    (1) Distributed Parallel Processing 
    Distributed computing and distributed storage are used to support mass data processing. Large-scale computing tasks can be divided and ruled by MapReduce (Fig. 2) to ensure system efficiency.

 


    (2) Virtualization Platform 
    For multilevel applications of mobile search, a virtualization platform (Fig. 3) is introduced to provide a virtual service environment and to eliminate the incompatibility caused by different terminals. The virtualization platform supports the management and allocation of resources of heterogeneous virtual machines and provides a flexible virtual desktop with seamless interaction between servers and clients, seamless combination of applications, two-way audio and video transmission, and application deployment.

 


    (3) An Organized Structure for Semantic Network and System of Text to Understand Chinese Webpages.


    Semantic understanding is the basis of understanding raw data and search purpose and is a key technology that makes mobile search results simple. We create a language model according to spoken Chinese and integrate sentence level text features based on spoken Chinese into the language model. At the same time, an unsupervised machine learning method recognizes named entities of page texts with unlimited fields to improve user experience. Semantic and syntactic decoding technology is also used to support natural language understanding.

 

3.3 Key Technologies of Mobile Search Based Cloud Computing


    (1) Multiservice Index Sharing 
    Distributed file system, which distributes caching technology and virtualization technology, can isolate data in a multiservice environment. But search applications need to be provided with an index data sharing mechanism to improve search accuracy and data utilization. The core idea of this technology is as follows: Index data is placed into multiple sub-libraries to ensure data isolation; a raw service database is built on the distributed file system to ensure global shared access; index intermediate libraries are placed in the distribution database to ensure strict data isolation; and cross-service data access is controlled through a data dispatch server and data access layer to ensure only a service that has permission can read other service index intermediate libraries (Fig. 4). In specific implementations, this technology also further ensures data segregation and running effectiveness of services by accessing the image library when accessing other index libraries. 

 


    (2) Personal Real-Time Search
    By combining a message-driven mechanism and distributed parallel computing architecture, a cloud-based search engine is realized. Such an engine has the capacity to update data in real time, which provides users with real-time search services. The platform informs the search engine to modify indexes automatically by message system when logging or modifying information or services. New search results are then returned immediately to ensure users can search the latest information the first time. A web-based search engine spends between one and three days updating search results. The key idea behind this capacity is to build an efficient and accurate distributed message-driven system. By classifying information and services, an efficient data distribution mechanism is built that balances data on computing nodes, and this ensures the message is processed efficiently and accurately. The messaging system is the second source for the search engine so that crawlers are not the only way to obtain information.


    (3) Role Index
    For the cloud computing mobile search platform, the user is introduced into the basic index and search process. Using this methodology, the search engine can predict the purpose of searches and provide a specific search “scene” that improves accuracy. An automatic association analysis mechanism analyzes the relationship between user role, user purpose, and category of actual search result. It then stores those relationships into a user’s behavioral view. The search engine creates role indexes for information and services, and these bind the user roles. When a search call comes in, the search engine determines the most correlative role index according to the relationship stored in a user’s behavioral view.  It then starts an additional search on the role index in conjunction with the traditional search to improve accuracy.

 

3.4 Application Case 
    The following illustrates the benefit of using cloud-based mobile search.


    An enterprise has a set of news search systems for mobile users. The system servers provide the main search function and are combined with minicomputers with largest available 200 caps access traffic. There are still many redundant caps for traffic (average: 10 caps, max: 50 caps). A major incident happens, and people want to find out what is going on. So they use the most convenient tool, a mobile terminal, to search for information. This brings about a sharp rise in service access traffic to 200 caps. Such a spike in access traffic would challenge a mobile search system with a traditional structure. When the system is unable to complete user requests, a large number of requests accumulate in the system tray buffer, or are rejected, and this leads to delayed response times or denial of access. So user satisfaction with mobile search is greatly affected.


    To solve this problem, the enterprise reconstructs the system based on cloud technology.  Existing minicomputers are kept and are given the main computing processing capacity. When traffic rises sharply, the new search system with distributed structure and virtualization technology allows the new PC server to join into the service system (20-50 caps processing capability for one node) in one minute. With the added advantages of cloud computing (Fig. 5), the system assigns user requests to N access points, and the data of each access point is assigned M processing nodes for processing according to the computing capability of every node. In this way, large-scale service requests can be handled by multinode load sharing and multinode parallel processing. After the peak, the corresponding nodes can be released. In practice, one minicomputer and 6-8 PC servers doubles system capability to 400 caps a minute. Similarly, this can be used for support in emergency situations to ensure service availability.

 


4 Conclusion 
    As the basis of mobile Internet architecture, cloud computing will be vigorously developed. Technologies between mobile search and cloud computing will fuse together, and more services will be introduced with this converged technology that will bring about greater convenience for work and life in the twenty first century.

 

References
[1] Y. Liu, “Summary of cloud computing and mobile cloud computing application,” Information and Communications Technologies, No. 2, pp.14-20, Feb. 2010.
[2] Q. Yuan, “Mobile Search Technologies and Service Development,” Telecom Network Technology, no. 4, pp. 38-42, Apr. 2007.
[3] T. Guo, “Competition Promoting Development: Mobile Search Trends,” Globrand. [online]. Available: http://www.globrand.com/2010/344424.shtml
[4] Deyi Li, “Analysis of Hot Topics in Cloud Computing,” ZTE Communications, vol. 8, no. 4, pp. 1-5, Dec. 2010.
[5] P. Zhao, “Cloud Computing Technology and Its Applications,” ZTE Communications, vol. 8, no. 4, pp. 34-38, Dec. 2010.

 

 

Biographies

Yan Gao (gao.yan6@zte.com.cn) graduated from Chongqing University in 2010, majoring in Communication and Electronic Information. She is now a pre-research engineer with ZTE Corporation. Her research areas are Chinese word segmentation technology, video search technologies, and the relevant aspects of cloud computing technology.

Li Fu (fu.li3@zte.com.cn) is product and project manager at the Chongqing Institute of ZTE Corporation. His reasearch areas include cloud computing and the construction of product platforms.

 

Zhenwei Zhang (Zhang.Zhenwei@zte.com.cn) is head of the Chongqing Institute of ZTE Corporation. He is responsible for product operation and daily affairs of the Institute. His main research interest is relevant aspects of cloud computing technology and technologies of the Internet of Things.

 

Shengmei Luo (luo.shengmei@zte.com.cn) graduated from Harbin Institute of Technology in 1996, majoring in Communications and Electronic Information. He is a chief engineer and architect at ZTE Corporation. He is also a member of the China Cloud Computing Committee and heads pre-research into new technologies. He was awarded a second prize for scientific and technological progress with several invention patents. He has published a number of academic papers in core national communications journals.

 

Ping Lu (lu.ping@zte.com.cn) joined ZTE Corporation in 1996. He is the president of the Business Research Institute of ZTE Corporation. He has served as the minister of the Network Department, product manager of operational support, product manager of WDSS, and deputy general manager of business software products.

[Abstract] Mobile search is beset with problems because of mobile terminal constraints and also because its characteristics are different from the traditional Internet search model. This paper analyzes cloud computing technologies—especially mass data storage, parallel computing, and virtualization—in an attempt to solve technical problems in mobile search. The broad prospects of cloud computing are also discussed.

[Keywords] mobile search; cloud computing; parallel computing; virtualization