Marvin Ferber

Resource Allocation for Cloud-Assisted Mobile Applications

Mobile devices such as netbooks, smartphones, and tablets have made computing ubiquitous. However, such battery powered devices often have limited computing power for the benefit of an extended runtime. Nevertheless, despite the reduced processing power, users expect to perform the same types of operations as they could do using their desktop or laptop computers. We address mobile devicess lack of computing power by leveraging cloud computing resources. We present a middleware that relocates computing-intensive parts of Java applications to cloud re-sources. Consequently, our middleware enables the execution of computing-intensive applications on mo-bile devices. We present a case study on which we adapt Sunflow, an open-source ray tracing application, to use our middleware and show the results obtained by deploying it on Amazon EC2. We show, via simulations, a cost analysis of using the different resource allocation strategies available on our solution.

Combining Object-Oriented Design and SOA with Remote Objects over Web Services

Current approaches of accessing stateful resources via SOAP Web services do not provide a standardized way to use program objects (classes). In this article, we show how the interface of an object-oriented class can be expressed using WSDL. This approach enables a program object to be used in a distributed environment by accessing its Web service interface. The set of Web services associated with a class define a Remote Object over Web Service (ROWS). A ROWS object can be used to facilitate the simultaneous use of the concepts of Service Oriented Architectures (SOA) and Distributed Object Architectures (DOA). In this context, we show how ROWS objects can be utilized in BPEL. Beyond that, we present a ROWS implementation that is suitable to serve as a standalone distributed object middleware. In a case study, we show how classes that were written in Java can be made accessible remotely through the ROWS technology.

Load Balancing Concurrent BPEL Processes by Dynamic Selection of Web Service Endpoints

Business workflows implemented as BPEL processes play an important role for many business applications. BPEL is used to orchestrate a series of Web service calls. Which provider is used for a specific Web service request is statically defined as endpoint of a Web service call within the BPEL code. If a Web service is offered by more than one provider an intelligent choice of a Web service provider can improve the throughput of a BPEL engine. In this article we study how a scheduling mechanism can be injected into a standard BPEL process by an automatic transformation of the BPEL process. The scheduling mechanism helps the BPEL engine to select between different Web service providers, e. g., according to their current workload. Three different strategies for scheduling Web service calls are examined in practical experiments, considering a homogeneous and a heterogeneous collection of Web service providers. We show that in most cases the dynamic scheduling approach leads to a smaller average execution time (makespan) of concurrently executed BPEL processes.

Mobile cloud computing in 3g cellular networks using pipelined tasks

Network latency is often high on mobile devices due to wireless access, e. g., via 3G cellular networks. To better use the ubiquitously available 3G network connections, we propose a pipelining task concept on a single encrypted channel between a mobile device and a cloud resource. This does not only increases wireless bandwidth occupation, it also makes wireless communication more predictable by assuring a high throughput even for small messages. Constantly high throughput allows for a better data transfer time estimation and can thus lead to a more adequate cloud resource selection to assist the mobile application. In an experimental evaluation using streaming image processing, we investigate the performance and applicability of our approach and compare it to the widely used HTTP.

Look-up Table Design for Deep Sub-threshold through Full-Supply Operation

Field programmable gate arrays (FPGAs) are the implementation platform of choice when it comes to design flexibility. However, the high power consumption of FPGAs (which arises due to their flexible structure), make them less appealing for extreme low power applications. In this paper, we present a design of an FPGA look-up table (LUT), with the goal of seamless operation over a wide band of supply voltages. The same LUT design has the ability to operate at sub-threshold voltage when low power is required, and at higher voltages whenever faster performance is required. The results show that operating the LUT in sub-threshold mode yields a (~80×) lower power and a (~4×) lower energy than full supply voltage operation, for a 6-input LUT implemented in a 22nm predictive technology. The key drawback of sub-threshold operation is its susceptibility to process, temperature, and supply voltage (PVT) variations. This paper also presents the design and experimental results for a closed-loop adaptive body biasing mechanism to dynamically cancel global (spacial) as well as local (random) PVT variations. For the same 22nm technology, we demonstrate that the closed-loop adaptive body biasing circuits can allow the FPGA LUT to operate over an operating frequency range that spans an order of magnitude (40 MHz to 1300 MHz). We also show that the closed-loop adaptive body biasing circuits can cancel delay variations due to supply voltage changes, and reduce the effect of process variations on setup and hold times by 1.8X and 2.9× respectively. The dynamic body biasing circuits incur a 3.49% area overhead when designed to each drive a cluster of 25 LUTs.

BPEL Remote Objects: Integrating BPEL Processes into Object-Oriented Applications

Service-orientation and object-oriented design are common practice in the field of business application development. Business process execution languages help to facilitate the orchestration of Web services in service-oriented architectures (SOA). However, using business processes from within object-oriented and event-driven applications is difficult as asynchronous event handling is missing in workflow-based business process modeling languages. The present article presents an approach for integrating BPEL business processes into object-oriented applications. We propose BPEL remote objects (BPELROs) that can be accessed asynchronously in an object-oriented manner. We present a method how state-based business processes can be implemented using BPELROs. It is shown how to apply BPELROs for software modernization tasks and we also evaluate the performance of BPELROs on different BPEL engines.

Reducing the class coupling of legacy code by a metrics-based relocation of class members

With the rapid growth of the complexity of software systems, the problem of integrating and maintaining legacy software is more relevant than ever. To overcome this problem, many methods for refactoring legacy code have already been proposed such as renaming classes or extracting interfaces. To perform a real modularization, methods have to be moved between classes. However, moving a single method is often not possible due to code dependencies. In this article we present an approach to modularize legacy software by moving multiple related class members. It is shown how to identify groups of class members with similar concerns. We present two different code patterns that the related members and their dependent classes must match to allow a relocation of the related members. We also demonstrate how our pattern-based approach for automated modularization of legacy software can be applied to two open source projects.