Ulrich Lampe

Maximizing Cloud Provider Profit from Equilibrium Price Auctions

Auctioning constitutes a market-driven scheme for the allocation of cloud-based computing capacities. It is practically applied today in the context of Infrastructure as a Service offers, specifically, virtual machines. However, the maximization of auction profits poses a challenging task for the cloud provider, because it involves the concurrent determination of equilibrium prices and distribution of virtual machine instances to the underlying physical hosts in the data center. In the work at hand, we propose an optimal approach, based on linear programming, as well as a heuristic approach to tackle this Equilibrium Price Auction Allocation Problem (EPAAP). Through an evaluation based on realistic data, we show the practical applicability and benefits of our contributions. Specifically, we find that the heuristic approach reduces the average computation time to solve an EPAAP by more than 99.9%, but still maintains a favorable average solution quality of 96.7% in terms of cloud provider profit, compared to the optimal approach.

Cost-Driven Optimization of Complex Service-Based Workflows for Stochastic QoS Parameters

The challenge of optimally selecting services from a set of functionally appropriate ones under Quality of Service (QoS) constraints -- the Service Selection Problem -- has been extensively addressed in the literature based on deterministic parameters. In practice, however, Quality of Service QoS parameters rather follow a stochastic distribution. In the work at hand, we present an integrated approach which addresses the Service Selection Problem for complex workflows in conjunction with stochastic Quality of Service parameters. Accounting for penalty cost which accrue due to Quality of Service violations, our approach reduces the impact of stochastic QoS behavior on total cost significantly.

Adaptive matchmaking for RESTful services based on hRESTS and MicroWSMO

Matchmaking -- i.e., the task of finding functionally suitable service offers based on a service request -- has only been addressed in the context of WS-* Web services. However, RESTful services are gaining increasing attraction and have been adopted by major companies, thus increasing the need for suitable matchmaking solutions. This paper introduces XAM4SWS, an adaptive matchmaker for semantic Web services that supports multiple service description formats, including hRESTS and MicroWSMO for RESTful services. XAM4SWS adapts existing methodologies from WS-* matchmaking and extends them through the inclusion of REST-specific service features. A prototypical implementation of the matchmaker is evaluated with respect to multiple information retrieval metrics using an adapted semantic Web service test collection.

LOG4SWS.KOM: Self-Adapting Semantic Web Service Discovery for SAWSDL

In recent years, a number of approaches to semantic Web service matchmaking have been proposed. Most of these proposals are based on discrete and thus relatively coarse Degrees of Match (DoMs). However, different basic assumptions regarding the generalization and specialization of semantic concepts in ontologies and their subsequent rating in matchmaking exist. Hence, most matchmakers are only properly suitable if these assumptions are met. In this paper, we present an approach for mapping subsumption reasoning-based DoMs to a continuous scale. Instead of determining the numerical equivalents of the formerly discrete DoMs manually, these values are automatically derived using a linear regression model. This permits not only easy combination with other numerical similarity measures, but also allows to adapt matchmaking to different basic assumptions. These notions are implemented and tested in LOG4SWS.KOM -- a matchmaker for SAWSDL that provides very good evaluation results with respect to Information Retrieval metrics such as precision and recall.

Concurrent negotiations in cloud-based systems

Utilizing cloud-based services, consumers gain a high level of flexibility, but they cannot obtain individual Quality of Service guarantees or request service compositions according to their specific business needs. Therefore, appropriate mechanisms for an automated negotiation of Quality of Service parameters are required that do not only consider the individual business objectives and strategies of the negotiation partners involved, but do also account for the dependencies between the different services and service tiers in cloud computing. This enables enterprises to increase the quality and flexibility of their business processes and lays the foundation for market-based complex service provisioning. In this paper, we present one such negotiation approach and evaluate the application of different negotiation strategies.

Enabling cost-efficient Software Service Distribution in infrastructure clouds at run time

Because Infrastructure as a Service in the form of virtual machines only provides a limited supply of resources per discrete instance, approaches are required to compute cost-efficient distribution strategies for Software as a Service instances. Due to its computational complexity, this Software Service Distribution Problem is difficult to address at run time where time constraints apply. In the work at hand, we propose a Knapsack-based heuristic approach; an extensive evaluation, based on data from actual cloud systems and software services, shows that the heuristic is able to solve the problem with more than 99.7 % reduction in computation time and only marginal degradation in solution quality compared to a locally optimal approach. Thus, through the minimization of infrastructure leasing costs, our proposed approach enables Software as a Service providers to achieve higher profit margins or more competitive pricing of their services in the market.

QoS-Based Optimization of Service Compositions for Complex Workflows

In Service-oriented Architectures, business processes can be realized by composing loosely coupled services. If services in the Internet of Services with comparable functionalities but varying quality levels are available at different costs on service marketplaces, service requesters can decide, which services from which service providers to select. The work at hand addresses computing an optimal solution to this service-selection-problem considering complex workflow patterns. For this, a linear optimization problem is formulated, which can be solved by applying integer linear programming techniques.

Will Mobile Cloud Gaming Work? Findings on Latency, Energy, and Cost

Mobile cloud gaming is a new and promising concept for the cloud-based delivery of video games to mobile devices in a platform-independent manner. In our ongoing work, we examine potential challenges of this service model concerning latency, energy consumption, and cost. The paper outlines our approach and presents initial results, which indicate that mobile cloud gaming could work well in wireless networks, specifically on modern smartphones, but may suffer from latency and cost issues in cellular networks.Software solution providers face numerous pitfalls when using standardized SOA-based solutions. One cause is variability between customer organizations—for example, variability occurs in local e-government because of municipalities’ autonomy when it comes to implementing national laws. This autonomy results in variations in business processes, which then cause variability in the information systems that implement them. This article presents pitfalls the authors observed in Dutch e-government, an example of a variability-intensive environment. They also present best practices that help deal with these pitfalls and organize them in an architecture pattern. Applying the pattern at more than 20 Dutch municipalities led to successful SOA implementations. Because such pitfalls are not just applicable to e-government, the proposed pattern can also be applied to other domains.Software solution providers face numerous pitfalls when using standardized SOA-based solutions. Once cause for pitfalls is variability between customer organizations. For example, in local e-government, variability occurs because of the autonomy of municipalities when implementing national laws. This autonomy results in variations in business processes in different municipalities, which then cause variability in information systems that implement business processes. In this article we report pitfalls that we observed in Dutch e-government, an example of a variability-intensive environment. Furthermore, we present best practices that help deal with these pitfalls and organize the best practices in an architecture pattern. Applying the pattern in more than 20 Dutch municipalities led to successful SOA implementations. As pitfalls are not only applicable to e-government, the proposed pattern can also be applied in other domains to deal with the pitfalls.

Let the clouds compute: cost-efficient workload distribution in infrastructure clouds

With cloud computing, a virtually inexhaustible pool of computing capacity has become available to IT users. However, given the large number of Infrastructure as a Service offers with differing pricing options, the cost-efficient distribution of workloads poses a complex challenge. In this work-in-progress paper, we formally describe the Cloud-oriented Workload Distribution Problem and propose an exact as well as a heuristic optimization approach. Through an evaluation that is based on realistic data from the cloud market, we examine the performance and practical applicability of these approaches.

Optimizing the Distribution of Software Services in Infrastructure Clouds

In the public perception, cloud computing is frequently associated with almost unlimited elasticity and scalability of computing capacity. However, Infrastructure as a Service in the form of virtual machines provides limited supplies of virtual resources, due to restrictions of the underlying physical hardware. At the same time, the execution of Software as a Service instances leads to a specific demand for these resources. Based on this observation, I introduce the Software Service Distribution Problem, i.e., the challenge of (cost-)efficiently distributing the execution of software service instances across available cloud infrastructure providers and virtual machine types under resource constraints. I outline my research approach, which aims at the development of optimization algorithms in the context of an integrated Software Service Distribution Broker, and report the progress made to date.