Program

The track provides a forum for scientists and engineers in academia and industry to present and discuss their latest research findings on selected topics in dependable and adaptive distributed systems. The track is structured in two sessions:

Thu, April 16, 3:00-4:50pm (session shared with WCN track)

• A Taxonomy of Reliable Request-Response Protocols
  Naghmeh Ivaki, Nuno Laranjeiro and Filipe Araujo

  Reliable request-response interactions, in which the server never executes a given request more than once, are being used to support business and safety-critical operations in diverse sectors, such as banking, E-commerce, or healthcare. This form of interactions can be quite difficult to implement, because the client, server, or communication channel may fail, potentially requiring diverse and complex recovery procedures. In this paper we address the following question: could we provide a meaningful taxonomy of reliable request-response protocols? We generate valid sequences of client and server actions, organize the generated sequences into a prefix tree, and classify them according to their reliability semantics and memory requirements. The tree reveals three families of protocols matching common real-world implementations that try to deliver exactly-once or at-most-once. The strict organization of the protocols provides a solid foundation for creating correct services, and we show that it also serves to easily identify fallacies and pitfalls of existing implementations.

• Workload characterization model for optimal resource allocation in cloud middleware
  Shruti Kunde and Tridib Mukherjee

  With increasing focus on inter-operability across cloud offerings to leverage their disparate capabilities, it has become more and more important to enable a flexible framework for sharing of heterogeneous resources in the cloud infrastructure. At the same time, it is imperative to be aware of the performance implications of hosting application workloads on different resources in order to guarantee Service Level Agreements (SLAs) to the applications. This paper focusses on experimental characterization of performance implications of different heterogeneous resources in hosting big-data analytics application workloads (one of the most critical applications in modern times). To create the knowledge, based on which the recommendations are provided, we benchmark the performance of big-data analytics applications, using a Hadoop cluster setup. Specifically, we study parameters of interest such as turnaround time and throughput, which are most likely to influence our choice of infrastructure for a particular application. Our experiments are conducted on varied platforms, both internal to Xerox and external cloud providers. We present a model based on our experiments, that facilitates the characterization of hetergeneous applications, thus enabling the cloud middleware to select an appropriate infrastructure and metrics in order to attain the the desired SLA.

Thu, April 16, 5:20-6:30pm

• Modeling Dependable Systems with Continuous Time Bayesian Networks
  Martin Größl

  In the domain of information systems modeling for depend- ability is an established method. Most approaches dealing with structural or probabilistic modeling do not consider time information and handle only discrete data. But in re- ality systems have a time varying behavior and numerous measures are continuous. In this paper an approach for modeling dependable informa- tion systems for fault prediction is presented. The method considers time behavior and continuous variables. The tech- nique is based on continuous time Bayesian networks (CTBN) which make assumptions for time to transition or time to failure feasible. A drawback of CTBN is that only discrete data is processed, thus continuous variables have to be dis- cretized. This is carried out by grouping measures with dis- tributions which are similar with the restriction that values from continuous range are nearby. Furthermore this tech- nique is capable of performing a data reduction such that subsequent computations can be done with moderate hard- ware resources. Based on such preprocessed data the struc- ture of the Bayesian network (BN) is learned by a Max- Min Hill-Climbing (MMHC) algorithm. Known misbehav- ior (e.g. faults) is incorporated into the BN by introduction of auxiliary variables. A structural model generated in this way forms the backbone of a continuous time Bayesian net- work. Henceforth CTBN parameter estimation (e.g. time characteristic) is doable by established learning methods.

• Evaluation of an adaptive framework for resilient executions of high demanding Monte Carlo applications
  Manuel Rodríguez-Pascual, Antonio Juan Rubio-Montero and Rafael Mayo-García

  Solving certain calculations in time is crucial for some industrial, medical or research areas. However, problems with high computational requirements are specially constrained to the availability and dependability of the computational resources. Distributed Computing Infrastructures have consolidated as the platform that can solve the issue in last decade. Grid and Cloud infrastructures can currently supply users with thousands of resources of different types. Nevertheless, despite the advances achieved, the nature of these infrastructures finally makes them un-predictable, especially grid. Consequently, users continuously experience failures and poor performance, and consequently, they result unfeasible for some calculations. An instrument to deal with this lack of dependability is to build adaptive algorithms specifically designed for increasing the reliability of certain types of applications on these heterogeneous and dynamic infrastructures. In this work, the suitability of the Montera2 framework is evaluated for Monte Carlo calculations. For this purpose, the proposed approach is compared with the basic tools offered by current middleware.

• Optimal Planning for Architecture-Based Self-Adaptation Via Model Checking of Stochastic Games
  Javier Camara, David Garlan, Bradley Schmerl and Ashutosh Pandey

  Architecture-based approaches to self-adaptation rely on architectural descriptions to reason about the best way of adapting the structure and behavior of complex software-intensive systems at runtime, either by choosing among aset of predefined adaptation strategies, or by automatically generating adaptation plans. While predefined strategy selection approaches have a low computational overhead and often facilitate dealing with uncertainty (e.g., by accounting explicitly for contingencies derived from unexpected outcomes of actions), they require additional designer effort regarding the specification of strategies and are unable to guarantee optimal solutions. In contrast, runtime plan generation is able to explore a richer solution space and provide optimal solutions in some cases, but are more limited when dealing with uncertainty, and incur higher computational overheads. In this paper, we propose an approach to optimal adaptation plan generation for architecture-based self-adaptation via model checking of stochastic multiplayer games (SMGs). Our approach enables: (i) trade-off analysis among different qualities by means of utility functions and preferences, and (ii) explicit modeling of uncertainty both as probabilistic outcomes of adaptation actions and by explicit modeling of the environment. Basing on the concepts embodied in the Rainbow framework for architecture-based self-adaptation, we illustrate the scalability of our approach in Znn.com, a benchmark case study that reproduces the typical infrastructure for a news website.

Poster session, Wed April 15, 2015

Track Program Co-Chairs

Important Dates