Academia.eduAcademia.edu

Figure 2 – uploaded by Aufeef Chauhan

See full PDFdownloadDownload figure
Table 2: Electronic Data Sources electronic databases are expected to cover most of the peer-reviewed literature on software engineering and computer sciences according to our work reported in [9, 15]. We also performed complementary manual searches on high impact cloud computing conferences and journals (to be described in Search and Study Selection Process section) that results in selection of 3 papers from Wiley, but we have not shown it in Table 2 because systematic searches were not performed on Wiley.

Table 2 Electronic Data Sources electronic databases are expected to cover most of the peer-reviewed literature on software engineering and computer sciences according to our work reported in [9, 15]. We also performed complementary manual searches on high impact cloud computing conferences and journals (to be described in Search and Study Selection Process section) that results in selection of 3 papers from Wiley, but we have not shown it in Table 2 because systematic searches were not performed on Wiley.

Related Figures (34)

Table 1: Research Questions and their respective Rationale Target Data Sources - We performed systematic searches for journal papers on electronic databases that had been accessible online. We did not look for information in books and printed sources. Table 2 presents a list of our selected electronic database. These four
Table 1: Research Questions and their respective Rationale Target Data Sources - We performed systematic searches for journal papers on electronic databases that had been accessible online. We did not look for information in books and printed sources. Table 2 presents a list of our selected electronic database. These four
Search and Study Selection Process — An SLR usually takes a significant amount of time to complete, certainly much more than the traditional literature review. Given the continuously srowing landscape of cloud-based research and practice, we decided to carry out this SLR in nultiple stages. We carried out first phase of searches in July 2011 using the protocol Jeveloped for this SLR. Those searches helped us to select the first set of papers for ‘eviewing and gaining an initial understanding of the literature. Three more phases of searches were carried out in: April 2013, December 2013 and July 2015. Our multi-staged strategy to carry out this SLR enabled us to include a large number of papers reporting mature ‘esults. Figure 1 provides some details of the searches and their outcomes at different stages. We performed systematic searches in the four digital libraries (i.e. IEEE, ACM, Springer and ScienceDirect) using the search query described earlier. We retrieved 1491 papers from all he searches. We filtered the papers according to the inclusion and exclusion criteria and selected 14 journal papers, which were rigorously reviewed for setting the foundation of this review. We repeated the whole search process again in April 2013 to the papers published between 2011 and April 2013. After performing the searches and merging them together, we got 2529 papers. After applying the inclusion and exclusion criteria, we selected another 38 papers. We performed the third and fourth rounds of searches by following the same process in December 2013 and May 2015. After combining the search results from the previously selected papers and removing the duplicates, we selected 111 journal papers. Then we performed complementary manual searches on high impact conferences and journals that were known for publishing papers on cloud computing research including IEEE International Conference on Cloud Computing, IEEE/ACM International Symposium on Cluster, Cloud and Grid Computing, IEEE International Conference on Cloud Computing Technology and Science, IEEE/ACM International Conference on Utility and Cloud Computing, and Software Practice and Experience Journal. As a result we selected 22 more papers in the review that resulted in 133 papers selected for this review. ~~ «| — ay. ere w¥+7T
Search and Study Selection Process — An SLR usually takes a significant amount of time to complete, certainly much more than the traditional literature review. Given the continuously srowing landscape of cloud-based research and practice, we decided to carry out this SLR in nultiple stages. We carried out first phase of searches in July 2011 using the protocol Jeveloped for this SLR. Those searches helped us to select the first set of papers for ‘eviewing and gaining an initial understanding of the literature. Three more phases of searches were carried out in: April 2013, December 2013 and July 2015. Our multi-staged strategy to carry out this SLR enabled us to include a large number of papers reporting mature ‘esults. Figure 1 provides some details of the searches and their outcomes at different stages. We performed systematic searches in the four digital libraries (i.e. IEEE, ACM, Springer and ScienceDirect) using the search query described earlier. We retrieved 1491 papers from all he searches. We filtered the papers according to the inclusion and exclusion criteria and selected 14 journal papers, which were rigorously reviewed for setting the foundation of this review. We repeated the whole search process again in April 2013 to the papers published between 2011 and April 2013. After performing the searches and merging them together, we got 2529 papers. After applying the inclusion and exclusion criteria, we selected another 38 papers. We performed the third and fourth rounds of searches by following the same process in December 2013 and May 2015. After combining the search results from the previously selected papers and removing the duplicates, we selected 111 journal papers. Then we performed complementary manual searches on high impact conferences and journals that were known for publishing papers on cloud computing research including IEEE International Conference on Cloud Computing, IEEE/ACM International Symposium on Cluster, Cloud and Grid Computing, IEEE International Conference on Cloud Computing Technology and Science, IEEE/ACM International Conference on Utility and Cloud Computing, and Software Practice and Experience Journal. As a result we selected 22 more papers in the review that resulted in 133 papers selected for this review. ~~ «| — ay. ere w¥+7T
Table 4: Primary Studies’ Distribution over Thematic Categories
Table 4: Primary Studies’ Distribution over Thematic Categories
Figure 3(a) shows the distribution of the reviewed studies over the digital libraries used in this SLR. ScienceDirect is the most prominent source with 67 (66%) studies. IEEE is the second with 41 (22%) studies published followed by Springer with 20 (12%) studies. There are 3 papers (2.3%) published by Wiley and 2 (1.5%) published by ACM.
Figure 3(a) shows the distribution of the reviewed studies over the digital libraries used in this SLR. ScienceDirect is the most prominent source with 67 (66%) studies. IEEE is the second with 41 (22%) studies published followed by Springer with 20 (12%) studies. There are 3 papers (2.3%) published by Wiley and 2 (1.5%) published by ACM.
Cloud Environments Used The reviewed papers have used different cloud environments to implement and deploy the proposed solutions as shown in Table 6. Amazon is on the top of the list with 29 reviewed papers using Amazon as the environment to implement the proposed solutions. Google App
Cloud Environments Used The reviewed papers have used different cloud environments to implement and deploy the proposed solutions as shown in Table 6. Amazon is on the top of the list with 29 reviewed papers using Amazon as the environment to implement the proposed solutions. Google App
Deployment Models Used Table 6: Cloud Environment used for research reported in the primary studies ne i Ae Sie ae ‘SLTMASUDS se SSS Tiree = | Heewsl PESRIES ohes It has been stated that cloud computing solutions can have different deployment models. The reviewed papers reported solutions for not only the three commonly known deployment models (i.e. public, private and hybrid), but some papers also used local infrastructure as a simulated cloud environment. Figure 4 shows the number of papers that used different deployment model or simulated cloud environments. Our SLR has revealed that Amazon Web Services (AWS), Windows Azure and Google App Engine are commonly used public cloud environments. Eucalyptus, OpenStack, AWS and Windows Azure are common choices for implementing hybrid solutions. Eucalyptus, Flexiscale cloud platform, UC-Cloud and
Deployment Models Used Table 6: Cloud Environment used for research reported in the primary studies ne i Ae Sie ae ‘SLTMASUDS se SSS Tiree = | Heewsl PESRIES ohes It has been stated that cloud computing solutions can have different deployment models. The reviewed papers reported solutions for not only the three commonly known deployment models (i.e. public, private and hybrid), but some papers also used local infrastructure as a simulated cloud environment. Figure 4 shows the number of papers that used different deployment model or simulated cloud environments. Our SLR has revealed that Amazon Web Services (AWS), Windows Azure and Google App Engine are commonly used public cloud environments. Eucalyptus, OpenStack, AWS and Windows Azure are common choices for implementing hybrid solutions. Eucalyptus, Flexiscale cloud platform, UC-Cloud and
Figure 4: Studies Distribution with respect to Cloud Deployment Models RESERVOIR have been used for private clouds. Grids, cloud simulations toolkits, and local machines clusters have also been used for building and using simulated cloud environments.
Figure 4: Studies Distribution with respect to Cloud Deployment Models RESERVOIR have been used for private clouds. Grids, cloud simulations toolkits, and local machines clusters have also been used for building and using simulated cloud environments.
Figure 5: Studies Distribution with respect to Solutions Abstraction
Figure 5: Studies Distribution with respect to Solutions Abstraction
Algorithms Used to Complement Architecture Based Solutions
Algorithms Used to Complement Architecture Based Solutions
Since software architecture design decisions play a significant role in support or inhibiting different quality attributes, we decided to draw a matrix of the research themes’ categories and the quality attributes on which different papers have focused. We believe that such a matrix can be useful to quickly figure out the quality attributes addressed in the papers that fall under certain category of the research themes. Figure 6 shows the distribution of the papers with respect to quality attributes that are shown on X-axis and the categories are shown on Y-axis. Scalability, performance, efficient resource utilization, CPU utilization and response time are the most frequently reported quality attributes in papers reporting architectural challenges and solutions for cloud-based systems. The bubble at each of the intersections of X-axis and Y-axis shows the number of corresponding papers. For example, the intersection of Performance and Resource and Service Management indicates that there are 36 papers that have reported challenges and solutions related to performance in the context of resources and service management issues. 3.9. Quality Assessment of the Studies
Since software architecture design decisions play a significant role in support or inhibiting different quality attributes, we decided to draw a matrix of the research themes’ categories and the quality attributes on which different papers have focused. We believe that such a matrix can be useful to quickly figure out the quality attributes addressed in the papers that fall under certain category of the research themes. Figure 6 shows the distribution of the papers with respect to quality attributes that are shown on X-axis and the categories are shown on Y-axis. Scalability, performance, efficient resource utilization, CPU utilization and response time are the most frequently reported quality attributes in papers reporting architectural challenges and solutions for cloud-based systems. The bubble at each of the intersections of X-axis and Y-axis shows the number of corresponding papers. For example, the intersection of Performance and Resource and Service Management indicates that there are 36 papers that have reported challenges and solutions related to performance in the context of resources and service management issues. 3.9. Quality Assessment of the Studies
8. |Research Themes Categories and Quality Attributes Matrix Chauhan M.A., Ali Babar M., Benatallah B.; Architecting Cloud-Enabled Systems, 2016.
8. |Research Themes Categories and Quality Attributes Matrix Chauhan M.A., Ali Babar M., Benatallah B.; Architecting Cloud-Enabled Systems, 2016.
Table 25 (Appendix B) shows the quality score for each of the papers in this SLR. All of the papers have value score 1 for Q1 as none of the reviewed papers had reported lessons learned or experience reports. The average scores for Q2, Q3 and Q4 are 0.99, 0.93 and 0.74 respectively. It shows that the aim and context of the research were clearly reported in all of the papers and the findings have been reported in satisfactory manner. An average value score of Q5 is 0.53, which means many of the selected papers have not provided clear directions for their future work and enhancements. The aggregated average quality assessment score is 4.2 out of 5, which indicates that the reviewed papers were of high quality. Table 8: Quality Assessment Criteria (a tailored version of the propositions from [19, 20]) setting in which the proposed solutions and the corresponding implementations were evaluated. The question 4 was used to evaluate whether or not a study had reported the data to support the evaluation results. The question 5 purported to assess whether or not a study clearly reported the limitations of the proposed solutions, and stated directions for improvement and future enhancements in the solutions. The answer to this question helped us to determine the future research scope of the problem areas discussed in a paper.
Table 25 (Appendix B) shows the quality score for each of the papers in this SLR. All of the papers have value score 1 for Q1 as none of the reviewed papers had reported lessons learned or experience reports. The average scores for Q2, Q3 and Q4 are 0.99, 0.93 and 0.74 respectively. It shows that the aim and context of the research were clearly reported in all of the papers and the findings have been reported in satisfactory manner. An average value score of Q5 is 0.53, which means many of the selected papers have not provided clear directions for their future work and enhancements. The aggregated average quality assessment score is 4.2 out of 5, which indicates that the reviewed papers were of high quality. Table 8: Quality Assessment Criteria (a tailored version of the propositions from [19, 20]) setting in which the proposed solutions and the corresponding implementations were evaluated. The question 4 was used to evaluate whether or not a study had reported the data to support the evaluation results. The question 5 purported to assess whether or not a study clearly reported the limitations of the proposed solutions, and stated directions for improvement and future enhancements in the solutions. The answer to this question helped us to determine the future research scope of the problem areas discussed in a paper.
Table 9: Resource Provisioning and Management - Quality Attribute
Table 9: Resource Provisioning and Management - Quality Attribute
Table 10: Resource Provisioning and Management — Monitoring and Deployment
Table 10: Resource Provisioning and Management — Monitoring and Deployment
Table 11: Pervasive Embedded Networks Solutions: Two studies, [PS22] and [PS23], present a Knowledge Aware and Service Oriented (KASO) middleware platform, which carries out common tasks by sharing resources. The middleware platform has been implemented following the Perceptual Reasoning Agents (PRAs) paradigm. Zhu et al. [PS109] present a framework for human centric context aware flowable services. The framework takes advantage of users’ context information to support proactive human activities and service integration. Lounis et al. [PS68] present a secure architecture for collecting and processing arge volumes of medical sensor data. The security mechanism guarantees confidentiality, integrity and fine-grained access to the data. A multi-layered cloud platform for ubiquitous healthcare services to satisfy high
Table 11: Pervasive Embedded Networks Solutions: Two studies, [PS22] and [PS23], present a Knowledge Aware and Service Oriented (KASO) middleware platform, which carries out common tasks by sharing resources. The middleware platform has been implemented following the Perceptual Reasoning Agents (PRAs) paradigm. Zhu et al. [PS109] present a framework for human centric context aware flowable services. The framework takes advantage of users’ context information to support proactive human activities and service integration. Lounis et al. [PS68] present a secure architecture for collecting and processing arge volumes of medical sensor data. The security mechanism guarantees confidentiality, integrity and fine-grained access to the data. A multi-layered cloud platform for ubiquitous healthcare services to satisfy high
Table 12: Federated Cloud Solutions: To address the challenges associated with cloud federation, Villegas et al. [PS99] propose to provide inter-cloud federation only at corresponding layers of NIST reference architecture model. This approach defines different federation methods at each layer to allow astic and fault-tolerant behavior at different stages without restricting them to a given cloud nvironment and deployment model. Walraven et al. [PS100] present a middleware platform amed PaaSHopper for developing interoperable and multi-tenant cloud services. Celesti e . [PS114] provide a layered architecture to gain economies of scale on federated clouds tha onsists of services discovery agents, matchmaking agent, and authentication agents. Rezaei al. [PS85] present a service-oriented architecture to provide interoperability among the ouds by autonomously converting semantic information to service syntax information and ag a 5 oqag0 8 Challenges: Cloud federation is referred to a collection of cloud-enabled resources collaborating with each other [22]. The papers classified in this subcategory focus on the challenges associated with cloud federation to facilitate cloud interoperability, avoid vendor lock-in and utilize resources from other clouds. Table 12 lists the challenges and indicates their relevance to IaaS, PaaS, SaaS, Essential or Complementary quality attributes. TL. 1%. Dina nwetn dy Qiin<wxa
Table 12: Federated Cloud Solutions: To address the challenges associated with cloud federation, Villegas et al. [PS99] propose to provide inter-cloud federation only at corresponding layers of NIST reference architecture model. This approach defines different federation methods at each layer to allow astic and fault-tolerant behavior at different stages without restricting them to a given cloud nvironment and deployment model. Walraven et al. [PS100] present a middleware platform amed PaaSHopper for developing interoperable and multi-tenant cloud services. Celesti e . [PS114] provide a layered architecture to gain economies of scale on federated clouds tha onsists of services discovery agents, matchmaking agent, and authentication agents. Rezaei al. [PS85] present a service-oriented architecture to provide interoperability among the ouds by autonomously converting semantic information to service syntax information and ag a 5 oqag0 8 Challenges: Cloud federation is referred to a collection of cloud-enabled resources collaborating with each other [22]. The papers classified in this subcategory focus on the challenges associated with cloud federation to facilitate cloud interoperability, avoid vendor lock-in and utilize resources from other clouds. Table 12 lists the challenges and indicates their relevance to IaaS, PaaS, SaaS, Essential or Complementary quality attributes. TL. 1%. Dina nwetn dy Qiin<wxa
adaressed by the research reported in this Category Of papers. Solutions: Two studies, [PS39] and [PS77], report middleware-based solutions to address cache management challenges. To tackle the issues of high throughput and managing cache for stateful and transactional applications, an elastic multi-tier architecture has been devised [PS77]. The proposed architecture uses SI-Cache to attain high performance by using snapshot isolation and extends the basic cache functionality by adding support for replication. Han et al. [PS39] presents an architecture that focuses on minimizing application’s cost while maintaining QoS through a cost-aware and workload-adaptive cache-based scaling approach. Tahle 12> Cache Manacement
adaressed by the research reported in this Category Of papers. Solutions: Two studies, [PS39] and [PS77], report middleware-based solutions to address cache management challenges. To tackle the issues of high throughput and managing cache for stateful and transactional applications, an elastic multi-tier architecture has been devised [PS77]. The proposed architecture uses SI-Cache to attain high performance by using snapshot isolation and extends the basic cache functionality by adding support for replication. Han et al. [PS39] presents an architecture that focuses on minimizing application’s cost while maintaining QoS through a cost-aware and workload-adaptive cache-based scaling approach. Tahle 12> Cache Manacement
Interoperability between services is provided with the help of Interoperability Engine, which generates adapters to facilitate interoperability. Zixue et al. [PS111] present a three-tier architecture to offload portion of computing tasks from wearable devices to cloud nodes. The offloading strategy maximizes the number of tasks that can be executed on the devices while guaranteeing acceptable response rate to the users. Hu et al. [PS44] present a cyber-physical system (CPS) named as Vita to support mobile users while performing crowd-sensing tasks. The system has been designed using SOA and supports intelligent distribution of tasks. The platform addresses the challenges of low computing and communication overhead on mobile devices. Zhang et al. [105] present an OSGi-based pervasive cloud infrastructure for services- offloading to the cloud. The infrastructure uses an elastic open service gateway to migrate services from small mobile devices to powerful cloud nodes. An architecture named Cooperative Terminals Service Environment (CTSE) for a Cloudlet amiaiteam hac haan nrancantad kx: Cine atk al TDWCO71] CTCD tanenncnc hanAiwn Ath «afGrnsanmncxy fae Table 14: Support for Mobile Devices.
Interoperability between services is provided with the help of Interoperability Engine, which generates adapters to facilitate interoperability. Zixue et al. [PS111] present a three-tier architecture to offload portion of computing tasks from wearable devices to cloud nodes. The offloading strategy maximizes the number of tasks that can be executed on the devices while guaranteeing acceptable response rate to the users. Hu et al. [PS44] present a cyber-physical system (CPS) named as Vita to support mobile users while performing crowd-sensing tasks. The system has been designed using SOA and supports intelligent distribution of tasks. The platform addresses the challenges of low computing and communication overhead on mobile devices. Zhang et al. [105] present an OSGi-based pervasive cloud infrastructure for services- offloading to the cloud. The infrastructure uses an elastic open service gateway to migrate services from small mobile devices to powerful cloud nodes. An architecture named Cooperative Terminals Service Environment (CTSE) for a Cloudlet amiaiteam hac haan nrancantad kx: Cine atk al TDWCO71] CTCD tanenncnc hanAiwn Ath «afGrnsanmncxy fae Table 14: Support for Mobile Devices.
Table 15: High Performance Computing Solutions: Prodan et al. [PS80] present a generic master slave framework to implement computing intensive algorithms on Google App Engine (GAE). The framework consists of a master application that manages logic and parallelization algorithm that can split a task into several parallel jobs. The slave applications perform the jobs. The results from slave jobs are collected by the master application and are combined together. The framework presented by Wong and Goscinski [PS101] is used to deploy and expose High Performance Computing (HPC) application on cloud as services. The framework sets up an execution environment by resolving dependencies (library decencies and searchable program paths) and provides a unique interface for constructing application services as SaaS. Somasundaram and Govindarajan [PS94] present CLOUDBR, which consists of a layered architecture and Particle Swarm Optimization (PSO)-based resource allocation and job scheduling mechanism. The framework guarantees deadline specific job completion of scientific applications.
Table 15: High Performance Computing Solutions: Prodan et al. [PS80] present a generic master slave framework to implement computing intensive algorithms on Google App Engine (GAE). The framework consists of a master application that manages logic and parallelization algorithm that can split a task into several parallel jobs. The slave applications perform the jobs. The results from slave jobs are collected by the master application and are combined together. The framework presented by Wong and Goscinski [PS101] is used to deploy and expose High Performance Computing (HPC) application on cloud as services. The framework sets up an execution environment by resolving dependencies (library decencies and searchable program paths) and provides a unique interface for constructing application services as SaaS. Somasundaram and Govindarajan [PS94] present CLOUDBR, which consists of a layered architecture and Particle Swarm Optimization (PSO)-based resource allocation and job scheduling mechanism. The framework guarantees deadline specific job completion of scientific applications.
Table 16: Security Management in Multi-tenant application Almorsy et al. [PS112] provide an aspect-based architecture (to inject security into applications) that consists of multiple modules including application architecture description database, security management console, multi-tenant security requirement database and system wrapper. Moens et al. [PS70] present a feature-based cloud resource management model using product line engineering methods to share instances among multiple tenants. The presented cost effective resource allocation model analyses cost of failure to place services on cloud instances and cost of using the instances in terms of price and energy consumption. Fehling et al. [PS117] provide tenant context based router and load balancer to distribute enants on hybrid cloud resources with the help of a scheduling algorithm. Goldschmidt et al. [PS119] provide a layered architecture that consists of an observer for scalability corresponding to specific needs of tenants’ data to be processed. Sousa and Machado [PS132] present QoSDBCCoordinator that consists of monitoring, scheduling, balancing and provisioning services that monitor the resources, schedule their provisioning, balance the provisioning schedule according to available storage services and provision data stores.
Table 16: Security Management in Multi-tenant application Almorsy et al. [PS112] provide an aspect-based architecture (to inject security into applications) that consists of multiple modules including application architecture description database, security management console, multi-tenant security requirement database and system wrapper. Moens et al. [PS70] present a feature-based cloud resource management model using product line engineering methods to share instances among multiple tenants. The presented cost effective resource allocation model analyses cost of failure to place services on cloud instances and cost of using the instances in terms of price and energy consumption. Fehling et al. [PS117] provide tenant context based router and load balancer to distribute enants on hybrid cloud resources with the help of a scheduling algorithm. Goldschmidt et al. [PS119] provide a layered architecture that consists of an observer for scalability corresponding to specific needs of tenants’ data to be processed. Sousa and Machado [PS132] present QoSDBCCoordinator that consists of monitoring, scheduling, balancing and provisioning services that monitor the resources, schedule their provisioning, balance the provisioning schedule according to available storage services and provision data stores.
Table 17: Data protection
Table 17: Data protection
Chauhan M.A., Ali Babar M., Benatallah B.; Architecting Cloud-Enabled Systems, 2016. Solutions: Itani et al. [PS49] present a PaaS security framework, SNUGE, for supporting confidentiality, integrity and authenticity of enterprise application’s data in PaaS. The framework supports decoupling among security domains to prevent spread of vulnerability from one domain to another. Zissis and Lekkas [PS110] identify the threats in the traditiona security implementations and proposes the use of a certified hardened operating system (OS on a bootable media that is open to extensive audits. The paper proposes a physical security mechanism, which eliminates the threat of installing malicious software on a system and provides elastic cloud architecture to prevent Distributed Denial of Service (DDoS) attacks. Jammalamadaka et al. [PS50] provide a client side interoperable security middleware, named iDataGuard, to facilitate adoption of heterogeneous interfaces of data storage facilities on the Internet. The middleware consists of a security model that provides confidentiality and integrity of outsourced data by indexing encrypted data, and allows searching on the encrypted data. Ferretti et al. [PS28] present an architecture-centric solution and a forma model that combines data encryption, key management, and authentication and authorization solutions. The presented formal model guarantees enforcement of confidential access contro to meta-data and corresponding data. Samanthula et al. [PS91] present a Secure Data Sharing (SDS) framework to prevent leakage of data from clouds. The approach specifically addresses cases when an unauthorized or
Chauhan M.A., Ali Babar M., Benatallah B.; Architecting Cloud-Enabled Systems, 2016. Solutions: Itani et al. [PS49] present a PaaS security framework, SNUGE, for supporting confidentiality, integrity and authenticity of enterprise application’s data in PaaS. The framework supports decoupling among security domains to prevent spread of vulnerability from one domain to another. Zissis and Lekkas [PS110] identify the threats in the traditiona security implementations and proposes the use of a certified hardened operating system (OS on a bootable media that is open to extensive audits. The paper proposes a physical security mechanism, which eliminates the threat of installing malicious software on a system and provides elastic cloud architecture to prevent Distributed Denial of Service (DDoS) attacks. Jammalamadaka et al. [PS50] provide a client side interoperable security middleware, named iDataGuard, to facilitate adoption of heterogeneous interfaces of data storage facilities on the Internet. The middleware consists of a security model that provides confidentiality and integrity of outsourced data by indexing encrypted data, and allows searching on the encrypted data. Ferretti et al. [PS28] present an architecture-centric solution and a forma model that combines data encryption, key management, and authentication and authorization solutions. The presented formal model guarantees enforcement of confidential access contro to meta-data and corresponding data. Samanthula et al. [PS91] present a Secure Data Sharing (SDS) framework to prevent leakage of data from clouds. The approach specifically addresses cases when an unauthorized or
Table 18: Security in Enterprise Service Bus Solutions: The architecture presented by Ryan [PS90] focuses on the security of Enterprise Service Bus (ESB) on the cloud. To tackle the problem of protecting internal application, a solution to include a hardware-based and application-aware appliance to provide protection from attacks (e.g. XML-based, message level and field level attacks) has been proposed. Message privacy, integrity and access control policies are enforced using a consistent Challenges: Providing standard end-point abstraction and protecting publicly exposed application interfaces are the challenges to consider for providing cloud-based enterprise service bus. Table 18 lists the challenges that have motivated the solutions reported in the studies that have been classified in this subcategory. The listed challenges signify the importance of security in ESB.
Table 18: Security in Enterprise Service Bus Solutions: The architecture presented by Ryan [PS90] focuses on the security of Enterprise Service Bus (ESB) on the cloud. To tackle the problem of protecting internal application, a solution to include a hardware-based and application-aware appliance to provide protection from attacks (e.g. XML-based, message level and field level attacks) has been proposed. Message privacy, integrity and access control policies are enforced using a consistent Challenges: Providing standard end-point abstraction and protecting publicly exposed application interfaces are the challenges to consider for providing cloud-based enterprise service bus. Table 18 lists the challenges that have motivated the solutions reported in the studies that have been classified in this subcategory. The listed challenges signify the importance of security in ESB.
Table 19: Data Intensive Architecture Challenges To handle data intensive computations, an architecture based on an extension to MapReduce has been presented by Gunarathne et al. [PS38] by adding a merge step and additional input parameters to Map and Reduce APIs to support the loop variant delta inputs on Windows Azure Platform named Twister4Azure. The Merge function is introduced as a new step tc MapReduce for supporting iterative MapReduce computation [PS38]. The framework also supports re-execution of the failed tasks until the completion of the iterative computations. Solutions: Chao et al. [PS19] present a FPGA-based acceleration solution using MapReduce framework. The framework combines hardware-based and software-based based acceleration to process data of massive scale and complexity. Kramer et al. [PS60] present a modular and flexible architecture that can encompass multiple big data processing algorithms for processing geospatial data. The data processing tasks are defined as workflows using domain specific language and are processed using cloud resources. parameters to Map and Reduce APIs to support the loop variant delta inputs on Windows
Table 19: Data Intensive Architecture Challenges To handle data intensive computations, an architecture based on an extension to MapReduce has been presented by Gunarathne et al. [PS38] by adding a merge step and additional input parameters to Map and Reduce APIs to support the loop variant delta inputs on Windows Azure Platform named Twister4Azure. The Merge function is introduced as a new step tc MapReduce for supporting iterative MapReduce computation [PS38]. The framework also supports re-execution of the failed tasks until the completion of the iterative computations. Solutions: Chao et al. [PS19] present a FPGA-based acceleration solution using MapReduce framework. The framework combines hardware-based and software-based based acceleration to process data of massive scale and complexity. Kramer et al. [PS60] present a modular and flexible architecture that can encompass multiple big data processing algorithms for processing geospatial data. The data processing tasks are defined as workflows using domain specific language and are processed using cloud resources. parameters to Map and Reduce APIs to support the loop variant delta inputs on Windows
Table 21: QoS Aware Services and Data Management
Table 21: QoS Aware Services and Data Management
Table 22: QoS-aware Resource Discovery, Monitoring and Management
Table 22: QoS-aware Resource Discovery, Monitoring and Management
Chauhan M.A., Ali Babar M., Benatallah B.; Architecting Cloud-Enabled Systems, 2016. Solutions: Chen et al. [PS21] provide a solution to satisfy consistency requirements of the application when cloud services are replicated. The proposed architecture introduces the notion of consistency regions and proposes a Region-Based Election Protocol (REP) to elastically balance workload among the regions. The architecture that is described by Kertesz et al. [PS57] focuses on service virtualization following SLA and supports interoperable service execution in a diverse and distributed virtua ized service system. MAPE (Monitor, Analyze, Plan and Execute) pattern is used for processing SLA-based Service Virtualization SSV) architecture. A behavioral-based resource m resources across multiple cloud layers (i.e. infrastruc application terms of use and utilizes this information QoS constraints using self-adaptation. ure, platform, and software), cost of satisfying QoS requirements and optimize infrastructure capacity has been by Kousiouris et al. [PS59]. The proposed approach analyzes information related to the o estimate low-level resource using time series analysis method and Artificial Neural Networks. Casalicchio and [PS16] analyze the problem from the perspective of an Application Service Provider (ASP) and uses a cloud infrastructure to achieve scalable provisioning of its services with respect to anagement approach that can manage minimize presented attributes Silvestri Krebs et al. [PS61] present metrics to quantify performance isolation in cloud-based systems.
Chauhan M.A., Ali Babar M., Benatallah B.; Architecting Cloud-Enabled Systems, 2016. Solutions: Chen et al. [PS21] provide a solution to satisfy consistency requirements of the application when cloud services are replicated. The proposed architecture introduces the notion of consistency regions and proposes a Region-Based Election Protocol (REP) to elastically balance workload among the regions. The architecture that is described by Kertesz et al. [PS57] focuses on service virtualization following SLA and supports interoperable service execution in a diverse and distributed virtua ized service system. MAPE (Monitor, Analyze, Plan and Execute) pattern is used for processing SLA-based Service Virtualization SSV) architecture. A behavioral-based resource m resources across multiple cloud layers (i.e. infrastruc application terms of use and utilizes this information QoS constraints using self-adaptation. ure, platform, and software), cost of satisfying QoS requirements and optimize infrastructure capacity has been by Kousiouris et al. [PS59]. The proposed approach analyzes information related to the o estimate low-level resource using time series analysis method and Artificial Neural Networks. Casalicchio and [PS16] analyze the problem from the perspective of an Application Service Provider (ASP) and uses a cloud infrastructure to achieve scalable provisioning of its services with respect to anagement approach that can manage minimize presented attributes Silvestri Krebs et al. [PS61] present metrics to quantify performance isolation in cloud-based systems.
Table 23: Energy Awareness The reviewed papers included in energy awareness category provide solutions for resource optimization and business process management according to energy efficiency requirement. This category is further classified into two subcategories as described in the following sections. Table 23 lists the reported challenges. By incorporating the challenges in the cloud solutions, energy footprint can be reduced by turning off idle nodes and placing virtual machines on the hardware with respect to energy constraints.
Table 23: Energy Awareness The reviewed papers included in energy awareness category provide solutions for resource optimization and business process management according to energy efficiency requirement. This category is further classified into two subcategories as described in the following sections. Table 23 lists the reported challenges. By incorporating the challenges in the cloud solutions, energy footprint can be reduced by turning off idle nodes and placing virtual machines on the hardware with respect to energy constraints.
Appendix B Table 24: Data Extraction Form Table 25: Detailed Quality Assessment Score of Selected Primary Studies
Appendix B Table 24: Data Extraction Form Table 25: Detailed Quality Assessment Score of Selected Primary Studies
Chauhan M.A., Ali Babar M., Benatallah B.; Architecting Cloud-Enabled Systems, 2016. Following is the list of the selected studies from primary searches that have been included in this SLR.
Chauhan M.A., Ali Babar M., Benatallah B.; Architecting Cloud-Enabled Systems, 2016. Following is the list of the selected studies from primary searches that have been included in this SLR.
580 California St., Suite 400
San Francisco, CA, 94104
© 2025 Academia. All rights reserved