Dependable and Adaptive Distributed Systems11th DADS Track of the31st ACM Symposium on Applied Computing Previous years: | 10th DADS 2015 9th DADS 2014 8th DADS 2013 7th DADS 2012 6th DADS 2011 5th DADS 2010 4th DADS 2009 3rd DADS 2008 2nd DADS 2007 1st DADS 2006 |
 http://www.acm.org/conferences/sac/sac2016/ April 4 - 8, 2016 Pisa, Italy |
The Symposium on Applied Computing has been a primary gathering forum for applied computer scientists, computer engineers, software engineers, and application developers from around the world. SAC 2016 is sponsored by the ACM Special Interest Group on Applied Computing and is hosted by the University of Pisa and Scuola Superiore Sant’Anna University. The SRC Program is sponsored by Microsoft Research.
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 comprises the following session:
Stretching Multi-Ring Paxos
Samuel Benz, Leandro Pacheco de Sousa and Fernando Pedone
Internet-scale services rely on data partitioning and replication to provide scalable performance and high availability. Moreover, to reduce user-perceived response times and tolerate disasters (i.e., the failure of a whole datacenter), services are increasingly becoming geographically distributed. Data partitioning and replication, combined with local and geographical distribution, introduce daunting challenges, including the need to carefully order requests among replicas and partitions. One way to tackle this problem is to use group communication primitives that encapsulate order requirements. This paper presents a detailed performance evaluation of Multi-Ring Paxos, a scalable group communi- cation primitive. We focus our analysis on “extreme conditions” with deployments including high-end 10 Gbps networks, a large number of combined rings (i.e., independent Paxos instances), a large number of replicas in a ring, and a global deployment. We also report on the performance of recovery under peak load and present two novel extensions to boost Multi-Ring Paxos’s performance.
Planning the Transformation of Overlays
Young Yoon, Nathan Robinson, Vinod Muthusamy, Sheila McIlraith and Hans-Arno Jacobsen
Reconfiguring a topology is an important technique to sustain high efficiency and robustness of an overlay. However, the problem of transforming the overlay from an old topology to a newly refined topology, at runtime, has received relatively little attention. The key challenge is to minimize the disruption that can be caused by topology transformation operations. Excessive disruption can be costly and harmful and thus hamper the decision to migrate to a better topology. To address this issue, we solve a problem of finding an appropriate sequence of steps to transform a topology that incurs the least service disruption. We call this the incremental topology transformation (ITT) problem. ITT can be formulated well as an automated planning problem and can be solved with numerous off-the-shelf planning algorithms. However, we found that state-of-the-art domain-independent planning techniques can not scale to solve large ITT problem instances. This shortcoming motivated us to develop a suite of planners that use novel domain-specific heuristics to guide the search for a solution. Our empirical evaluation illustrates that our planners offer a viable solution to a diversity of ITT problems. We envision that our approach could eventually provide a compelling addition to the arsenal of techniques currently employed by the overlay administrators.
NATCloud: Cloud-Assisted NAT-Traversal Service
Hanna Kavalionak, Amir H. Payberah, Alberto Montresor and Jim Dowling
Although over the last decade large efforts have been done to design efficient peer-to-peer (P2P) protocols, very few of them have taken into account the problem of firewalls and network address translators (NAT). Most of the existing P2P systems do not work properly when a high percentage of nodes are behind NAT. While a few P2P systems tackled the NAT problem, all of them employ third party nodes to establish a connection towards nodes behind NAT, and these may become bottlenecks, menacing the health of the entire system. A possible solution to this problem is to rent extra resources from the cloud. This paper presents NATCLOUD, a cloud-assisted NAT-traversal service, where rented cloud resources are added on demand to the overlay, as third party nodes, to help other nodes to make connections to nodes behind NAT. We show the feasibility of integrating our approach with existing gossip- based peer sampling services and evaluate our solution by simulations, conducting extensive experiments under different network conditions.
Dynamic Adaptation of Geo-Replicated CRDTs
Carlos Bartolomeu, Manuel Bravo and Luis Rodrigues
Conflict-free Replicated Data Types (CRDTs) are high-level data types that can be replicated with minimal coordination among replicas due to its confluent semantics. This property makes CRDTs specially appealing for geo-replicated settings. Different approaches, such as state transfer and oper- ation forwarding, have been proposed to propagate updates among replicas, with different tradeoffs among the amount of network traffic generated and the staleness of local information. This paper proposes and evaluates techniques to automatically adapt a CRDT implementation, such that the best approach is used, based on the application needs (captured by a SLA) and the observed system configura- tion. Our techniques have been integrated in SwiftCloud, a state of the art geo-replicated store based on CRDTs.
Monitoring Service Level Workload and Adapting Highly Available Applications
Mehran Khan, Ferhat Khendek and Maria Toeroe
Elasticity is a key feature in the cloud while High-Availability (HA) is one of the challenges. Recently, an architecture has been proposed for managing HA in the cloud using the SA Forum middleware and an Elasticity Engine for triggering resource provisioning and de-provisioning at the application level while preserving HA. This Elasticity Engine requires the monitoring of the application service level workload in contrast to the monitoring of VM workload as done usually. In this paper we propose an approach for the monitoring of HA applications at the service level and its integration with the Elasticity Engine. The approach allows for the monitoring of application processes in the traditional manner and for the mapping of this workload to their combined service level workload. Resource usages are aggregated and mapped to the service level workload using a distributed client-server architecture. The approach allows for distinguishing between the different HA states, active and standby, a component can be assigned at runtime and it adapts to the situations where switchovers happen under the control of the SA Forum middleware due to failures for example.
Details see SAC poster page.
An Adaptive Multi-Criteria Ranking of Security Countermeasures
Nicola Nostro, Ilaria Matteucci, Francesco Santini, Andrea Ceccarelli, Felicita Di Giandomenico, Fabio Martinelli and Andrea Bondavalli
We propose a multi-criteria framework for ranking controlling strategies according to several weights, such as delay-time, resource cost, and success-probability of attacks defined via quantitative threat analysis. Therefore, by assigning a different priority to weight-dimensions, we can consequently rank controllers in an adaptive way. To achieve this, we exploit an algebraic structure (i.e., semirings) with the purpose to have a parametric system of costs, and formal properties to compose/optimize them. To exemplify our approach, we adopt a real world use case: the Customer Energy Management System (CEMS), that is part of an electrical power system. Indeed, acting as an interface among different systems, CEMS is exposed to attacks, such as the Man in the Middle (MiM) and the Denial of Service (DoS), both considered as attack examples in our approach.
Virtualization Technologies for the Big Data Environment
Aymen JLASSI and Patrick MARTINEAU
Today, the "big data" paradigm is known by researchers and professionals. Many services are proposed on the cloud in order to fill the customers’ needs at storage and calculation. The "big data" tools require a lot of resources to run in a non continuous period of time. Thus, performance is rapidly becoming ubiquitous concerning the deployment of big data and cloud infrastructures. The consumers request, nowadays, virtual resources like CPU, RAM, disk (etc.) taken from service providers (like Amazon) and they pay on a "pay-as-you-go" basis. The supervisors adopt virtualization technologies, which allow multiple users and applications to share a physical machine. These technologies optimize resources usage and limit the operating cost. The virtualization technologies is classified in two categories. The first one concerns the heavy vir- tualization, which is based on virtual machines (VM) concept. Each VM emulates hardware and embeds its own operating system (OS) that is completely isolated from the host OS. The second one con- cerns the light virtualization, which is based on the management of containers. The containers share the host OS kernel while ensuring isolation. In this paper, we benchmark the performance and the energy consumption of a big data infrastructure that is based on the software Hadoop regarding the two technologies of virtualization. We analyze the influence of the provisioning variation on the performance using MapReduce benchmarks. At first, we will identify the points to be improved concerning Hadoop performances and then we will reduce the deployment cost on the cloud. Second, the Hadoop community finds an in-depth study of the resources consumption depending on the environment of deployment. Our experiments are based on the comparison of the Docker technology (light virtualization) and VMware technology® (heavy virtualization). We come to the point that in most experiments the light technology offers better performances in completion time of workloads. Besides, the performances of Hadoop clusters vary according to the number of slave machines (containers/VMs) per physical host and per the size and the configuration of the hardware.
Analysis of Checkpointing Overhead in Parallel State Machine Replication
Odorico Mendizabal, Fernando Dotti and Fernando Pedone
A well-established technique used to design fault-tolerant systems is State Machine Replication. In part, this is explained by the simplicity of the approach and its strong consistency guarantees. Recently, several proposals have suggested parallelizing the execution of state machine replicas to achieve higher throughput. Concurrent execution of commands has many implications, including the procedure to recover replicas from failures. Conventional checkpointing techniques, for example, must be revisited in parallelized models. In this paper, we review parallel variations of state machine replication and discuss how checkpointing procedures apply to these models. Moreover, we evaluate the impact caused by checkpointing techniques on recovery through simulations.
Deadlock Models in Distributed Computation: Foundations, Design, and Computational Complexity
Valmir Barbosa, Alan Carneiro, Fabio Protti and Ueverton Souza
Distributed systems consist of a set of independent processors interconnected by a communication network that supports resource sharing. A deadlock occurs in a distributed system when a group of processes waits indefinitely for resources from each other. Distributed systems are usually represented by wait-for graphs, where the behavior of a process is determined by a deadlock model. In this paper, we revisit deadlock model concepts, and present a new deadlock model as a simpler alternative to the And/Or model. Using also computational complexity and circuit complexity aspects, we provide a novel analysis of the hierarchy of classical deadlock models, where we identify how expressive each model is from the point of view of polynomial computations. Finally we present a generic graph structure to characterize deadlock situations.
Karl M. Göschka (Main contact chair)
University of Applied Sciences Technikum Wien
Embedded Systems Institute
Hoechstaedtplatz 6
A-1200 Vienna, Austria
phone: +43 664 180 6946
fax: +43 664 188 6275
dads@dedisys.org
goeschka (at) technikum-wien dot at
Rui Oliveira
Universidade do Minho
Computer Science Department
Campus de Gualtar
4710-057 Braga, Portugal
phone: +351 253 604 452 / Internal: 4452
fax: +351 253 604 471
rco (at) di dot uminho dot pt
Peter Pietzuch
Imperial College London
Department of Computing
South Kensington Campus
180 Queen's Gate
London SW7 2AZ, United Kingdom
phone: +44 (20) 7594 8314
fax: +44 (20) 7581 8024
prp (the at sign goes here) doc (dot) ic (dot) ac (dot) uk
Giovanni Russello
University of Auckland
Department of Computer Science
Private Bag 92019
Auckland 1142, New Zealand
phone: +64 9 373 7599 ext. 86137
g dot russello at auckland dot ac dot nz
| September 28, 2015 (11:59PM Pacific Time) - extended | Paper submission |
| November 23, 2015 | Author notification |
| December 7, 2015 | Camera-ready papers |
For general information about SAC, please visit: http://www.acm.org/conferences/sac/sac2016/
If you have further questions, please do not hesitate to contact us: dads@dedisys.org