Max Lehn

Fossa: Learning ECA Rules for Adaptive Distributed Systems

The development of adaptive distributed systems is complex. Due to a large amount of interdependencies and feedback loops between network nodes and software components, distributed systems respond nonlinearly to changes in the environment and system adaptations. Although Event Condition Action (ECA) rules allow a crisp definition of the adaptive behavior and a loose coupling with the actual system implementation, defining concrete rules is nontrivial. It requires specifying the events and conditions which trigger adaptations, as well as the selection of appropriate actions leading to suitable new configurations. In this paper, we present the idea of Fossa, an ECA framework for adaptive distributed systems. Following a methodology that separates the adaptation logic from the actual application implementation, we propose learning ECA rules by automatically executing a multitude of tests. Rule sets are generated by algorithms such as genetic programming, and the results are evaluated using a utility function provided by the developer. Fossa therefore provides an automated offline learner that derives suitable ECA rules for a given utility function.

Channel-based Unidirectional Stream Protocol (CUSP)

This paper presents a novel transport protocol, CUSP, specifically designed with complex and dynamic network applications in mind. Peer-to-peer applications benefit in particular, as their requirements are met by neither UDP nor TCP. While other modern transports like SCTP or SST have also tried to combine the advantages of TCP and UDP, CUSP overcomes their technical and conceptual shortcomings. CUSP makes it possible to directly express application logic in the message flow. Modern applications need a mixture of request-response, request-multiple-response, publish-subscribe, and message-passing. All of these operations can be conveniently implemented using CUSPs unidirectional streams. We separate low-level packet management from streams into reusable channels. A channel connects two applications providing negotiation, congestion control, and cryptography. Developers operate on the stream level, sending messages as reliable and ordered byte-streams. Although they may share a common channel, a stall or loss in one stream does not block the others.

Towards a comparative performance evaluation of overlays for Networked Virtual Environments

Peer-to-peer overlays for Networked Virtual Environments have recently gained much research interest, resulting in a variety of different approaches for spatial information dissemination. Although designed for the same purpose, the evaluation methodologies used by particular authors differ widely. This makes any comparison of existing systems difficult, if not impossible. To overcome this problem we present a benchmarking methodology which allows for a fair comparison of those systems. We, therefore, define a common set of workloads and metrics. We demonstrate the feasibility of our approach by testing four typical systems for spatial information dissemination and discovering their specific performance profiles.

Transition-enabled event dissemination for pervasive mobile multiplayer games

Todays smartphones feature sophisticated computational capabilities as well as a plethora of sensors, enabling new kinds of applications. One popular example are pervasive mobile multiplayer games, where interactions with other users or physical objects in the real world are part of the gameplay. This induces a notion of locality, as a users actions affect other nearby users. However, this locality in the interaction is not reflected in the underlying communication system. Events triggered by users are sent via the cellular connection to a remote data center, where they are processed and then distributed to other players, again via the cellular link. In this work, we demonstrate an event dissemination system that utilizes the available local communication interfaces of todays smartphones to distribute events in groups of nearby players. The system executes transitions between different communication interfaces (e.g., Bluetooth or Wi-Fi Direct) and the utilized dissemination protocol depending on the size and density of the group of players, thereby exploiting the characteristics of the chosen interface. Attendees can experience the resulting benefits in terms of latency and update frequency by playing our mobile pervasive multiplayer game TowerWorld at the venue.

An online gaming testbed for peer-to-peer architectures

In this demo we present a testbed environment for Peer-to-Peer (P2P) game architectures. It is based on Planet PI4, an online multiplayer game whose gameplay provides a standard workload for a set of gaming-specific network interfaces. Its pluggable architecture allows for the evaluation and comparison of existing and new P2P networking approaches. Planet PI4 can run on a real network for prototypical evaluation as well as in a discrete-event simulator providing a reproducible environment.

Performance Evaluation of Peer-to-Peer Gaming Overlays

In this demo we present a performance evaluation testbed for peer-to-peer gaming overlays. It consists of a 3D first person shooter game that is designed to run in a simulated network environment as well as on a real network. Simulation with autonomous players (bots) guarantees scalability, a controlled workload, and reproducible results; a prototype deployment on a real network can then validate the simulation results. The information dissemination overlay pSense is implemented as a first subject for evaluation.

Towards a Common Interface for Overlay Network Simulators

Simulation has become an important evaluation method in the area of Peer-to-Peer (P2P) research due to the scalability limitations of evaluation test beds such as Planet Lab or G-Lab. Current simulators provide various abstraction levels for different underlay models, such that applications can be evaluated at different granularity. However, existing simulators suffer from a lack of interoperability and portability making the comparison of research results extremely difficult. To overcome this problem, we present an approach for a generic application interface for discrete-event P2P overlay network simulators. It enables porting of the same implementation of a targeted application once and then running it on various simulators as well as in a real network environment, thereby enabling a diverse and extensive evaluation. We established the feasibility of our approach and showed negligible memory and runtime overhead.

Generation of synthetic workloads for multiplayer online gaming benchmarks

We present an approach to the generation of realistic synthetic workloads for use in benchmarking of (massively) multiplayer online gaming infrastructures. Existing techniques are either too simple to be realistic or are too specific to a particular network structure to be used for comparing different networks with each other. Desirable properties of a workload are reproducibility, realism and scalability to any number of players. We achieve this by simulating a gaming session with AI players that are based on behavior trees. The requirements for the AI as well as its parameters are derived from a real gaming session with 16 players. We implemented the evaluation platform including the prototype game Planet PI4. A novel metric is used to measure the similarity between real and synthetic traces with respect to neighborhood characteristics. In our experiments, we compare real trace files, workload generated by two mobility models and two versions of our AI player. We found that our AI players recreate the real workload characteristics more accurately than the mobility models.

Designing benchmarks for P2P systems

In this paper we discuss requirements for peer-to-peer (P2P). benchmarking, and we present two exemplary approaches to benchmarks for Distributed Hashtables (DHT) and P2P gaming overlays. We point out the characteristics of benchmarks for P2P systems, focusing on the challenges compared to conventional benchmarks. The two benchmarks for very different types of P2P systems are designed applying a common methodology. This includes the definition of the system under test (SUT). and particularly its interfaces, the workloads and metrics. A set of common P2P quality metrics helps to achieve a comprehensive selection of workloads and metrics for each scenario.

Fossa: Using genetic programming to learn ECA rules for adaptive networking applications

Due to complex interdependencies and feedback loops between network layers and nodes, the development of adaptive applications is difficult. As networking applications respond nonlinearly to changes in the environment and adaptations, defining concrete adaptation rules is nontrivial. In this paper, we present the offline learner Fossa, which uses genetic programming to automatically learn suitable Event Condition Action (ECA) rules. Based on utility functions defined by the developer, the genetic programming learner generates a multitude of rule sets and evaluates them using simulations to obtain their utility. We show, for a concrete example scenario, how the genetic programming learner benefits from the clear model of the ECA rules, and that the methodology efficiently generates ECA rules which outperform nonadaptive and manually tuned solutions.