1. What have the authors contributed in "University of groningen utility-based decision making for migrating cloud-based applications" ?
However, the heterogeneity of cloud services and providers has progressively become a challenge for migrating applications to the cloud, as ( i ) the performance of cloud services typically fluctuates and varies w. r. t. the type of cloud service and provider ( Gómez Sáez et al. 2015 ), and ( ii ) the fluctuation of the application workload has an impact on the application ’ s overall performance ( Gómez Sáez et al. 2014 ).. This work focuses on migrating applications that follow the layered architectural pattern ( Fowler 2002 ), such as the three-layered Web shop application evaluated in Andrikopoulos et al. ( 2014 ).. In particular, this work materializes the decision making mechanism in SCARF, by using utility theory for the costand performanceefficient distribution of cloud-based applications.. Utility has also been utilized for optimizing the allocation of computational ( Minarolli and Freisleben 2011a ) and storage resources ( Strunk et al. 2008 ), and in this work has one major goal: assisting business and IT experts in the decision making tasks when spanning their applications among multiple and heterogeneous cloud services and providers.. The contributions of this article are ( 1 ) an extension of the SCARF lifecycle introduced in their previous work ( Gómez Sáez et al. 2016 ), which focuses on the performanceand cost-efficient distribution of applications, by means of incorporating the decision making phases and using utility theory as the basis ; ( 2 ) a formal utility model serving as the underlying decision making mechanism to assist in the distribution of cloud-based applications consuming different cloud services ; and ( 3 ) a first evaluation of such a model using a two-layered MediaWiki application, the Wikipedia realistic workload and data, its financial reports, and under different singleprovider distribution scenarios.. The remainder of this article 1MediaWiki: https: //www. mediawiki. org.. Section 2 summarizes relevant concepts this work builds upon.. The utility model for optimizing the distribution of cloud-based applications is presented in Section 3. 1, which are subsequently evaluated in Section 4.. Section 5 presents the limitations of their approach, Section 6 introduces related works, and Section 7 concludes with future research works.. This work builds upon two pillars: ( i ) the design and distribution of cloud-based applications, and ( ii ) the economic models used for decision making of cloud-based applications.. Toward empowering the reusability of topology models among applications, the authors proposed to model an application topology as a typed topology graph model, which can be partitioned into a graph model that depicts the application-specific ( α-topology ) and nonapplication specificγ -topology ( Andrikopoulos et al. 2014 ).. Α-topologies represent the components that are unique and specific for each application, e. g., the frontand back-ends of the Web shop in ACM Transactions on Internet Technology, Vol. 18, No. 2, Article 22.. So far, the authors introduced a topology model comprising the functional aspects of the application.. For this, a costand performance-aware topology model incorporating the notions of cost and performance was introduced in Gómez Sáez et al. ( 2016 ).. The introduced topology model is fundamental for the utility-based analysis carried on in the remainder of this article, as it is used as the basis for the decision making tasks related to optimally selecting a specific distribution of an application ( μtopology ) ( Gómez Sáez et al. 2014, 2016 ).. The distribution of applications in cloud environments is typically a non-trivial task, due to the diversity of cloud offerings, providers, and the heterogeneous characteristics among them.. The first ingredient in SCARF consists of ( i ) a lifecycle ( extended from Gómez Sáez et al. ( 2016 ) ) depicted in Figure 2 and ( ii ) the SCARM introduced in Gómez Sáez et al. ( 2016 ).. Since migration of applications to the cloud often involves the interaction of business leaders and IT professionals ( Laverty et al. 2014 ), the authors consider two main actors in the design and development of distributed cloud-based applications: the ( i ) Application Architect, responsible for the architectural design and planning of the application distribution, and the application profile, and the ( ii ) Business Architect, ACM Transactions on Internet Technology, Vol. 18, No. 2, Article 22.. During the production phase of the application, the following synergistic decision making tasks take place: the ( vi ) Monitoring task captures real performance metrics, which are leveraged by application architects to ( vii ) analyze the evolution of the application performance and workload behavior in the Evolution Analysis task.. This work focuses on the utility-based evaluation step of SCARM, i. e., ( iv ), by means of using utility theory as the underlying model for the decision making mechanism for distributing cloud-based applications.. This section presents a formal utility model geared toward the profitable distribution of cloudbased applications in SCARF, using the Web shop application previously introduced as an example and depicting further distribution alternatives ( see Figure 1 ).. For the scope of this work, the SCARF utility model quantitatively represents the monetary cost and performance tradeoff of a ACM Transactions on Internet Technology, Vol. 18, No. 2, Article 22.. Since a cloud-based application viable μ-topology is decomposed into an application specific α-topology and multiple non-application specific ( and reusable ) γ -topology models ( see Section 2. 1 ), let us define the following: —T α as the set of application specific α-topology models { T α 1,.... One possible definition of the sat ( Ψ, W ) can be realized ACM Transactions on Internet Technology, Vol. 18, No. 2, Article 22.. Γ subtopologies are intended to be used in further application topologies, e. g., using the Apache Web server γ -topology of Figure 1 in another PHP-based application.. Workload models are defined as a probabilistic model representing different potential workload behaviors, each depicting the arrival rate of users and transactions that impact the application state ( Gómez Sáez et al. 2016 ).. Focusing on calculating the utility for the lifetime of the application Ψ, denoted as multiple time intervals Ψm, the authors can define the utility of the application w. r. t. its lifetime as
read more
2. What future works have the authors mentioned in the paper "University of groningen utility-based decision making for migrating cloud-based applications" ?
Future works focus on continuing the evaluations, by ( i ) driving an extended sensitivity analysis on the model, and by evaluating ( ii ) multi-cloud deployment and ( iii ) redistribution scenarios.. Moreover, further types of application architectures, e. g., workflowbased or built as micro-services, are planned to be considered.
read more
