系统管理及进化自主算法研究 SaaS
Chapter 1 Introduction
Traditionally software applications would be reside on the samehost machine, on which they would be accessed from. However, SaaS changes thisby hosting software applications on a remote server. These software applicationsand services are then accessed by remote users via the Inter. There are a fewbasic points that are mon to all SaaS implementations. Software applicationsor services are stored remotely. A user can then accesses these services or softwareapplications from anywhere in the world via the inter. And in most cases, usersdo not have to install anything onto their host machine. All they require is a Webbrowser to access these services (although in some cases, a Web browsers mayrequire an additional plug-in or add-on for certain services). These simple set ofpoints are mon for all SaaS implementations, whether they are for businessand enterprise use or even just for individual use. Inside SaaS we have differentmodels depict in figure 1-1: In traditional model number1 in figure 1-1 each customer accesses thecustomized version of the software. There are various instances of the sameapplication running on the servers. All users within an organization connect to asingle instance of the application which is different from any other instance accessed by any other organization. This model does not offer the full benefits of amature SaaS application, but it helps the vendors reduce costs by reusing portionsof earlier code base and consolidating hardware resources. This would also helpthem in entering the SaaS maret quicly with their existing application suite. Thesoftware administration still remains troublesome since any upgradation needs tobe done for all the customized instances.In configured model number2 in figure 1-1 customers access their owninstance of the same application which is configured to suit individual needs. Theinstance used by each customer remains isolated from others and hence requiresthe same amount of hardware resource space as that of traditional model.Architecturally there is a single code base for all customers, which reduces thesoftware service costs for the provider. Any upgradation is easily provided to allcustomers. However, shifting from traditional model requires lot moredevelopmental wor.In single instance model number3 in figure 1-1 every customer accesses asingle instance of the application, with metadata which is configured to give aunique experience to each user. Security policies and authorization proceduresensure that the information provided by each customer remains isolated from thatof others. The end-user has no indication about the multi-tenancy of the datawhich occurs at the bac end. The vendor saves hardware costs since there isonly one instance of the application running and hence not much storage space isrequired on the servers. Scalability is an issue since the server resources anddatabase space is shared among all users. As the number of users increases, up-gradation of hardware resources bees troublesome.
.........
Chapter 2 Modeling SaaS Applications
2.1 Introduction
Modeling SaaS application is very important field, building a SaaS byleveraging existing technology is a challenge issue and needs brand new softwaretechnology. It is useful for business purposes, because it should be easily adopted inseveral domains, such as healthcare, education and OA (Office Automation). Forthat concern modeling SaaS application is needed. From state of art we haveobserved the SaaS model have variation in application layers. Lie in [89] worflowis different from instance to other. In addition the data layer it has variousinformation as described in [90] for big organization. However the service layer ithas different architectures in small and medium enterprises [91] [92].In this chapter we have drown our framewor of SaaS application for allchapters. We begin by show novel model for SaaS application. At first we definedgeneral architecture for SaaS application. Depend on model driven development wederived SaaS meta-model layers for all application layers. Our research is differentfrom existence wor because it represents all application layers in one model. Thatclassifies the SaaS application management in three perspectives. To demonstratethis new opinion we tae online booing SaaS application as running example. Infourth term we have described service architecture for SaaS application. By novelalgorithm we realize self-configuration of model in section five. Finally wesummarized the research wor of this chapter.
2.2 Framewor of SaaS Applications
Our main contributions represented in new design for SaaS applicationexecution management framewor. We have followed the migration framewor ofapplication to SaaS application in [93]. But the general SaaS application frameworin this thesis we introduced it by including modeling SaaS application as the firstwor in chapter2. Chapter3 showed the second novelty is autonomic managementfor QoS SaaS application. We realized evolution adaptability SaaS model inchapter4. In chapter5 we joined our contributions in one prototype for autonomicSaaS system as depicted in figure 2-1 bellow. System modeling is very important issues in software engineering, because ithas given very big power in application development. For that we have defined ournovel architecture of the SaaS application. Then described our model bymeta-model concept to show we could easy management SaaS model. An application architecture specifies the technologies to be used toimplement one or more (and possibly all) information systems in terms of data,process, and interface, and how these ponents interact across a wor.Architecture is a transferable abstraction of a system. As we study from recentresearches architecture tae a big part to develop SaaS application [94] [95] [96][97]. Our novelty here is to architect conceptual model for SaaS application asdepict in figure 2-2. It different from other wor by integrated all application layersand adds autonomic management part will be clear in describing bellow. Thisarchitecture included main three parts: Our model for SaaS application has been classified in three managementperspectives that depict in figure 2-3. According to the ind of the service candetermine the management perspective. The reasonability of this classification isvariation of application layers from perspective to perspective.
CHAPTER 3 AUTONOMIC ALGORITHM FOR SAAS QOS MANAGEMENT .. 46
3.1 INTRODUCTION ......................... 46
3.2 APPLICABILITY OF SAAS APPLICATIONS ................... 46
CHAPTER 4 AUTONOMIC ALGORITHM FOR SAAS APPLICATIONSEVOLUTION........................ 61
4.1 INTRODUCTION ..................................... 61
4.2 CASE-BASED REASONING FOR SAAS APPLICATIONS .................................. 61
CHAPTER 5 AUTONOMIC MANAGEMENT FOR SAAS SYSTEMS................. 75
5.1 INTRODUCTION ............. 75
5.2 AUTONOMIC ALGORITHM FOR SAAS SYSTEM MANAGEMENT .................... 75
Chapter 5 Autonomic Management for SaaS Systems
5.1 Introduction
In this chapter we have designed novel algorithm for autonomic managementSaaS system. It depends on all algorithms that applied in previous chapters for SaaSapplication. We verify this algorithm in SaaSEHR case study to obtain autonomicmanagement for one example of health SaaS system. According to our model inchapter2 has three management perspectives user, tenant, provider represented bypatient, hospital, administrator respectively. The experimental realize theself-optimization for patient and hospital perspective. In addition it satisfyself-healing problem in administrator perspective.
5.2 Autonomic Algorithm for SaaS System Management
By integrating the wor presented in previous chapters， we obtained the goalfor autonomic management of SaaS systems. Because the business integrationinfrastructures via Web services are being mon and wors must bereconfigured to respond to rapidly changing requirements (and changes in businessworld) by SaaS system. We have focused on nonfunctional requirement lieapplicability and performance. We proposed the autonomic algorithm for SaaSsystem management (AASS) as depicted in table 5-1. AASS detects changes ofexecution environment and user's requirements by collecting real data from worldthrough detecting agents. And the activating agents is used to replay activator e.g.,by activating suitable intelligent agents. At first we need to transfer data intoinformation in monitoring process after that model adaptation at runtime by learningfrom monitored data. The value of business in medical system it need to improve and optimizationmedical treatment and environment. First we can facilitate the medical costing byelectronic payment. Secondly sharing of resident information will reduce themedical cost. Thirdly government can monitoring and controls the effective disease.Fourthly researching and consulting services will be easy and successful. Fifthlymedical treatment can be any time everywhere.In this SaaSEHR case study we have looed at some problems. The systemneed to customize and configure all application layers to serve many hospitals and alot of patients. This configuration should be dynamically by system without userintervention. Healthcare it is service must be available any time anywhere to save human life without pain. The services need to be classifying in ordering necessity.Measure and improve the QoS requirements is very important to give better servicecare. In almost problems in healthcare it has the same manner to solve. For thatmeta-heuristic methods will be good way because it is easy to formalizationnowledge to solve problem by similarity.
.............
Conclusion and Future Wor
SaaS is growing exponentially and numerous enterprises of all sizes cangenerate major performance improvements and cost savings through its application.The research and development of autonomic algorithms and systems have beenmajor challenges to QoS adaptability and evolution of the SaaS applications. Inorder to obtain meta-heuristic evolution, we propose novel models and algorithmsto enable the SaaS systems self-optimizing, self-configuring, self-healing, andself-protecting. The enhanced SaaSEHR system is developed as a case study toverify and valid the presented autonomic algorithms. The major contributions of thethesis are summarized below:(1) The novel meta-model and models are proposed for SaaS applications. Bythe meta-model, we define four layers to posite system and describe theassociations and dependencies among layered constitutes. In the businessprocess layer, it is necessary to sharing worflow among SaaS providers,tenants, and users to improve efficiency of SaaS system and to obtaineffective control to the consumer services. The meta-model and modelsallow us to improve the quality of SaaS systems in three perspectives ofSaaS providers, tenants, and users, and support to develop autonomicalgorithms for the SaaS systems.(2) AAQS algorithm is proposed to manage the SaaS QoS for the adaption to thechange of SaaS requirements. The quality evaluation models of DSUTPs andCSUTPs are adopted in AAQS to evaluate the applicability and performancemetrics of SaaS services. The formal languages of Probabilistic ComputationTree Logic (PCTL) and Continuous Stochastic logic (CSL) are used todefine QoS requirements formally. The applicability and dependences ofSaaS services is also described from three perspectives of SaaS providers,tenants and users. By using PRISM as model checer, empirical evaluationand analysis are conducted, and the results shown, AAQS enable thedetection and prediction of violation for SaaS applications.
..............
参考文献（略）