Abstract: The management of Grid resources requires scheduling of both computation and communication tasks at various levels. In this study, we consider the two constituent sub-problems of Grid scheduling, namely: (i) the scheduling of computation tasks to processing resources and (ii) the routing and scheduling of the data movement in a Grid network. Regarding computation tasks, we examine two typical online task scheduling algorithms that employ advance reservations and perform full network simulation experiments to measure their performance when implemented in a centralized or distributed manner. Similarly, for communication tasks, we compare two routing and data scheduling algorithms that are implemented in a centralized or a distributed manner. We examine the effect network propagation delay has on the performance of these algorithms. Our simulation results indicate that a distributed architecture with an exhaustive resource utilization update strategy yields better average end-to-end delay performance than a centralized architecture.
Abstract: Many of the network security protocols employed today utilize symmetric block ciphers (DES, AES and CAST etc). The majority of the symmetric block ciphers implement the crucial substitution operation using look up tables, called substitution boxes. These structures should be highly nonlinear and have bit dispersal, i.e. avalanche, properties in order to render the cipher with resistant to cryptanalysis attempts, such as linear and differential cryptanalysis. Highly secure substitution boxes can be constructed using particular Boolean functions as components that have certain mathematical properties which enhance the robustness of the whole cryptoalgorithm. However, enforcing these properties on SBoxes is a highly computationally intensive task. In this paper, we present a distributed algorithm and its implementation on a computing cluster that accelerates the construction of secure substitution boxes with good security properties. It is fully parametric since it can employ any class of Boolean functions with algorithmically definable properties and can construct SBoxes of arbitrary sizes. We demonstrate the efficiency of the distributed algorithm implementation compared to its sequential counterpart, in a number of experiments.
Abstract: Wireless sensor networks are comprised of a vast number of
ultra-small autonomous computing, communication and sensing devices,
with restricted energy and computing capabilities, that co-operate
to accomplish a large sensing task. Such networks can be very useful
in practice, e.g.~in the local monitoring of ambient conditions and
reporting them to a control center. In this paper we propose a
distributed group key establishment protocol that uses mobile agents
(software) and is particularly suitable for energy constrained,
dynamically evolving ad-hoc networks. Our approach totally avoids
the construction and the maintenance of a distributed structure that
reflects the topology of the network. Moreover, it trades-off
complex message exchanges by performing some amount of additional
local computations in order to be applicable at dense and dynamic
sensor networks. The extra computations are simple for the devices
to implement and are evenly distributed across the participants of
the network leading to good energy balance. We evaluate the
performance of our protocol in a simulated environment and compare
our results with existing group key establishment protocols. The
security of the protocol is based on the Diffie-Hellman problem and
we used in our experiments its elliptic curve analog. Our findings
basically indicate the feasibility of implementing our protocol in
real sensor network devices and highlight the advantages and
disadvantages of each approach given the available technology and
the corresponding efficiency (energy, time) criteria.
Abstract: Wireless Sensor Networks are complex systems consisting of a number of relatively simple autonomous sensing devices spread on a geographical area. The peculiarity of these devices lies on the constraints they face in relation to their energy reserves and their computational, storage and communication capabilities. The utility of these sensors is to measure certain environmental conditions and to detect critical events in relation to these measurements. Those events thereupon have to be reported to a specific central station namely the “sink”. This data propagation generally has the form of a hop-by-hop transmission. In this framework we work on distributed data propagation protocols which are taking into account the energy reserves of the sensors. In particular following the work of Chatzigiannakis et al. on the Probabilistic Forwarding Protocol (PFR) we present the distributed probabilistic protocol EFPFR, which favors transmission from the less depleted sensors in addition to favor transmissions close to the “optimal line”. This protocol is simple and relies only on local information for propagation decisions. Its main goal is to limit the total amount of energy dissipated per event and therefore to extend the network’s operation duration.
Abstract: In this work, we study the propagation of influence and computation in dynamic distributed computing systems that are possibly disconnected at every instant. We focus on a synchronous message-passing communication model with broadcast and bidirectional links. Our network dynamicity assumption is a worst-case dynamicity controlled by an adversary scheduler, which has received much attention recently. We replace the usual (in worst-case dynamic networks) assumption that the network is connected at every instant by minimal temporal connectivity conditions. Our conditions only require that another causal influence occurs within every time window of some given length. Based on this basic idea, we define several novel metrics for capturing the speed of information spreading in a dynamic network. We present several results that correlate these metrics. Moreover, we investigate termination criteria in networks in which an upper bound on any of these metrics is known. We exploit our termination criteria to provide efficient (and optimal in some cases) protocols that solve the fundamental counting and all-to-all token dissemination (or gossip) problems.
Abstract: In this chapter, our focus is on computational network analysis from a theoretical point of view. In particular, we study the \emph{propagation of influence and computation in dynamic distributed computing systems}. We focus on a \emph{synchronous message passing} communication model with bidirectional links. Our network dynamicity assumption is a \emph{worst-case dynamicity} controlled by an adversary scheduler, which has received much attention recently. We first study the fundamental \emph{naming} and \emph{counting} problems (and some variations) in
networks that are \emph{anonymous}, \emph{unknown}, and possibly dynamic. Network dynamicity is modeled here by the \emph{1-interval connectivity model}, in which communication is synchronous and a (worst-case) adversary
chooses the edges of every round subject to the condition that each instance is connected. We then replace this quite strong assumption by minimal \emph{temporal connectivity} conditions. These conditions only require that \emph{another causal influence occurs within every time-window of some given length}. Based on this basic idea we define several novel metrics for capturing the speed of information spreading in a dynamic network. We present several results that correlate these metrics. Moreover, we investigate \emph{termination criteria} in networks in which an upper bound on any of these metrics is known. We exploit these termination criteria to provide efficient (and optimal in some cases) protocols that solve the fundamental \emph{counting} and \emph{all-to-all token dissemination} (or \emph{gossip}) problems. Finally, we propose another model of worst-case temporal connectivity, called \emph{local
communication windows}, that assumes a fixed underlying communication network and restricts the adversary to allow communication between local neighborhoods in every time-window of some fixed length. We prove some basic properties and provide a protocol for counting in this model.
Abstract: We study the partially eponymous model of distributedcomputation, which simultaneously
generalizes the anonymous and the eponymous models. In this model, processors have
identities, which are neither necessarily all identical (as in the anonymous model) nor
necessarily unique (as in the eponymous model). In a decision problem formalized as a
relation, processors receive inputs and seek to reach outputs respecting the relation. We
focus on the partially eponymous ring, and we shall consider the computation of circularly
symmetric relations on it. We consider sets of rings where all rings in the set have the same
multiset of identity multiplicities.
We distinguish between solvability and computability: in solvability, processors are
required to always reach outputs respecting the relation; in computability, they must
do so whenever this is possible, and must otherwise report impossibility.
We present a topological characterization of solvability for a relation on a set of rings,
which can be expressed as an efficiently checkable, number-theoretic predicate.
We present a universal distributed algorithm for computing a relation on a set of
rings; it runs any distributed algorithm for constructing views, followed by local steps.
We derive, as our main result, a universal upper bound on the message complexity to
compute a relation on a set of rings; this bound demonstrates a graceful degradation
with the Least Minimum Base, a parameter indicating the degree of least possible
eponymity for a set of rings. Thereafter, we identify two cases where a relation can be
computed on a set of rings, with rings of size n, with an efficient number of O .n lg n/
messages.
Abstract: We work on an extension of the Population Protocol model of Angluin et al. that allows edges of the communication graph, G, to have states that belong to a constant size set. In this extension, the so called Mediated Population Protocol model (MPP), both uniformity and anonymity are preserved. We here study a simplified version of MPP, the Graph Decision Mediated Population Protocol model (GDM), in order to capture MPP's ability to decide (stably compute) graph languages (sets of communication graphs). To understand properties of the communication graph is an important step in almost any distributed system. We prove that any graph language is undecidable if we allow disconnected communication graphs. As a result, we focus on studying the computational limits of the GDM model in (at least) weakly connected communication graphs only and give several examples of decidable graph languages in this case. To do so, we also prove that the class of decidable graph languages is closed under complement, union and intersection operations. Node and edge parity, bounded out-degree by a constant, existence of a node with more incoming than outgoing neighbors and existence of some directed path of length at least k=O(1) are some examples of properties whose decidability is proven. To prove the decidability of graph languages we provide protocols (GDMs) for them and exploit the closure results. Finally, we prove the existence of symmetry in two specific communication (sub)graphs which we believe is the first step towards the proof of impossibility results in the GDM model. In particular, we prove that there exists no GDM, whose states eventually stabilize, to decide whether G contains some directed cycle of length 2 (2-cycle).
Abstract: Wireless sensor networks are comprised of a vast number of ultra-small autonomous computing, communication and sensing devices, with restricted energy and computing capabilities, that co-operate to accomplish a large sensing task. Such networks can be very useful in practice, e.g.~in the local monitoring of ambient conditions and reporting them to a control center. In this paper we propose a new lightweight, distributed group key establishment protocol suitable for such energy constrained networks. Our approach basically trade-offs complex message exchanges by performing some amount of additional local computations. The extra computations are simple for the devices to implement and are evenly distributed across the participants of the network leading to good energy balance. We evaluate the performance our protocol in comparison to existing group key establishment protocols both in simulated and real environments. The intractability of all protocols is based on the Diffie-Hellman problem and we used its elliptic curve analog in our experiments. Our findings basically indicate the feasibility of implementing our protocol in real sensor network devices and highlight the advantages and disadvantages of each approach given the available technology and the corresponding efficiency (energy, time) criteria.
Abstract: Wireless sensor networks are comprised of a vast number of ultra-small autonomous computing, communication and sensing devices, with restricted energy and computing capabilities, that co-operate to accomplish a large sensing task. Such networks can be very useful in practice, e.g.~in the local monitoring of ambient conditions and reporting them to a control center. In this paper we propose a new lightweight, distributed group key establishment protocol suitable for such energy constrained networks. Our approach basically trade-offs complex message exchanges by performing some amount of additional local computations. The extra computations are simple for the devices to implement and are evenly distributed across the participants of the network leading to good energy balance. We evaluate the performance our protocol in comparison to existing group key establishment protocols both in simulated and real environments. The intractability of all protocols is based on the Diffie-Hellman problem and we used its elliptic curve analog in our experiments. Our findings basically indicate the feasibility of implementing our protocol in real sensor network devices and highlight the advantages and disadvantages of each approach given the available technology and the corresponding efficiency (energy, time) criteria.
Abstract: Wireless sensor networks are comprised of a vast number of ultra-small fully autonomous computing, communication and sensing devices, with very restricted energy and computing capabilities, which co-operate to accomplish a large sensing task. Such networks can be very useful in practice in applications that require fine-grain monitoring of physical environment subjected to critical conditions (such as inaccessible terrains or disaster places). Very large numbers of sensor devices can be deployed in areas of interest and use self-organization and collaborative methods to form deeply networked environments. Features including the huge number of sensor devices involved, the severe power, computational and memory limitations, their dense deployment and frequent failures, pose new design and implementation aspects. The efficient and robust realization of such large, highly-dynamic, complex, non-conventional environments is a challenging algorithmic and technological task. In this work we consider certain important aspects of the design, deployment and operation of distributed algorithms for data propagation in wireless sensor networks and discuss some characteristic protocols, along with an evaluation of their performance.
Abstract: Andrews et al. [Automatic method for hiding latency in high bandwidth networks, in: Proceedings of the ACM Symposium on Theory of Computing, 1996, pp. 257–265; Improved methods for hiding latency in high bandwidth networks, in: Proceedings of the Eighth Annual ACM Symposium on Parallel Algorithms and Architectures, 1996, pp. 52–61] introduced a number of techniques for automatically hiding latency when performing simulations of networks with unit delay links on networks with arbitrary unequal delay links. In their work, they assume that processors of the host network are identical in computational power to those of the guest network being simulated. They further assume that the links of the host are able to pipeline messages, i.e., they are able to deliver P packets in time O(P+d) where d is the delay on the link.
In this paper we examine the effect of eliminating one or both of these assumptions. In particular, we provide an efficient simulation of a linear array of homogeneous processors connected by unit-delay links on a linear array of heterogeneous processors connected by links with arbitrary delay. We show that the slowdown achieved by our simulation is optimal. We then consider the case of simulating cliques by cliques; i.e., a clique of heterogeneous processors with arbitrary delay links is used to simulate a clique of homogeneous processors with unit delay links. We reduce the slowdown from the obvious bound of the maximum delay link to the average of the link delays. In the case of the linear array we consider both links with and without pipelining. For the clique simulation the links are not assumed to support pipelining.
The main motivation of our results (as was the case with Andrews et al.) is to mitigate the degradation of performance when executing parallel programs designed for different architectures on a network of workstations (NOW). In such a setting it is unlikely that the links provided by the NOW will support pipelining and it is quite probable the processors will be heterogeneous. Combining our result on clique simulation with well-known techniques for simulating shared memory PRAMs on distributed memory machines provides an effective automatic compilation of a PRAM algorithm on a NOW.
Abstract: The simplex method has been successfully used in solving linear programming problems for many years. Parallel approaches have also extensively been studied due to the intensive computations required, especially for the solution of large linear problems (LPs). In this paper we present a highly scalable parallel implementation framework of the standard full tableau simplex method on a highly parallel (distributed memory) environment. Specifically, we have designed and implemented a suitable column distribution scheme as well as a row distribution scheme and we have entirely tested our implementations over a considerably powerful distributed platform (linux cluster with myrinet interface). We then compare our approaches (a) among each other for variable number of problem size (number of rows and columns) and (b) to other recent and valuable corresponding efforts in the literature. In most cases, the column distribution scheme performs quite/much better than the row distribution scheme. Moreover, both schemes (even the row distribution scheme over large-scale problems) lead to particularly high speedup and efficiency values, which are considerably better in all cases than the ones achieved in other similar research efforts and implementations. Moreover, we further evaluate our basic parallelization scheme over very large LPs in order to validate more reliably the high efficiency and scalability achieved.
Abstract: This paper addresses the efficient processing of
top-k queries in wide-area distributed data
repositories where the index lists for the attribute
values (or text terms) of a query are distributed
across a number of data peers and the
computational costs include network latency,
bandwidth consumption, and local peer work.
We present KLEE, a novel algorithmic
framework for distributed top-k queries,
designed for high performance and flexibility.
KLEE makes a strong case for approximate top-k
algorithms over widely distributed data sources.
It shows how great gains in efficiency can be
enjoyed at low result-quality penalties. Further,
KLEE affords the query-initiating peer the
flexibility to trade-off result quality and expected
performance and to trade-off the number of
communication phases engaged during query
execution versus network bandwidth
performance. We have implemented KLEE and
related algorithms and conducted a
comprehensive performance evaluation. Our
evaluation employed real-world and synthetic
large, web-data collections, and query
benchmarks. Our experimental results show that
KLEE can achieve major performance gains in
terms of network bandwidth, query response
times, and much lighter peer loads, all with small
errors in result precision and other result-quality
measures
Abstract: In this paper we propose an energy-aware broadcast algorithm for wireless networks. Our algorithm is based on the multicost approach and selects the set of nodes that by transmitting implement broadcasting in an optimally energy-efficient way. The energy-related parameters taken into account are the node transmission power and the node residual energy. The algorithm{\^a}€™s complexity however is non-polynomial, and therefore, we propose a relaxation producing a near-optimal solution in polynomial time. We also consider a distributed information exchange scheme that can be coupled with the proposed algorithms and examine the overhead introduced by this integration. Using simulations we show that the proposed algorithms outperform other solutions in the literature in terms of energy efficiency. Moreover, it is shown that the near-optimal algorithm obtains most of the performance benefits of the optimal algorithm at a smaller computational overhead.
Abstract: Wireless Sensor Networks (WSNs) constitute a recent and promising new
technology that is widely applicable. Due to the applicability of this
technology and its obvious importance for the modern distributedcomputational world, the formal scientific foundation of its inherent laws
becomes essential. As a result, many new computational models for WSNs
have been proposed. Population Protocols (PPs) are a special category of
such systems. These are mainly identified by three distinctive
characteristics: the sensor nodes (agents) move passively, that is, they
cannot control the underlying mobility pattern, the available memory to
each agent is restricted, and the agents interact in pairs. It has been
proven that a predicate is computable by the PP model iff it is
semilinear. The class of semilinear predicates is a fairly small class. In
this work, our basic goal is to enhance the PP model in order to improve
the computational power. We first make the assumption that not only the
nodes but also the edges of the communication graph can store restricted
states. In a complete graph of n nodes it is like having added O(n2)
additional memory cells which are only read and written by the endpoints
of the corresponding edge. We prove that the new model, called Mediated
Population Protocol model, can operate as a distributed nondeterministic
Turing machine (TM) that uses all the available memory. The only
difference from a usual TM is that this one computes only symmetric
languages. More formally, we establish that a predicate is computable by
the new model iff it is symmetric and belongs to NSPACE(n2). Moreover, we
study the ability of the new model to decide graph languages (for general
graphs). The next step is to ignore the states of the edges and provide
another enhancement straight away from the PP model. The assumption now is
that the agents are multitape TMs equipped with infinite memory, that can
perform internal computation and interact with other agents, and we define
space-bounded computations. We call this the Passively mobile Machines
model. We prove that if each agent uses at most f(n) memory for f(n)={\`U}(log
n) then a predicate is computable iff it is symmetric and belongs to
NSPACE(nf(n)). We also show that this is not the case for f(n)=o(log n).
Based on these, we show that for f(n)={\`U}(log n) there exists a space
hierarchy like the one for classical symmetric TMs. We also show that the
latter is not the case for f(n)=o(loglog n), since here the corresponding
class collapses in the class of semilinear predicates and finally that for
f(n)={\`U}(loglog n) the class becomes a proper superset of semilinear
predicates. We leave open the problem of characterizing the classes for
f(n)={\`U}(loglog n) and f(n)=o(log n).
Abstract: In this paper we present an efficient general simulation strategy for
computations designed for fully operational BSP machines of n ideal processors,
on n-processor dynamic-fault-prone BSP machines. The fault occurrences are failstop
and fully dynamic, i.e., they are allowed to happen on-line at any point of the
computation, subject to the constraint that the total number of faulty processors
may never exceed a known fraction. The computational paradigm can be exploited
for robust computations over virtual parallel settings with a volatile underlying
infrastructure, such as a NETWORK OF WORKSTATIONS (where workstations may be
taken out of the virtual parallel machine by their owner).
Our simulation strategy is Las Vegas (i.e., it may never fail, due to backtracking
operations to robustly stored instances of the computation, in case of locally
unrecoverable situations). It adopts an adaptive balancing scheme of the workload
among the currently live processors of the BSP machine.
Our strategy is efficient in the sense that, compared with an optimal off-line
adversarial computation under the same sequence of fault occurrences, it achieves an O
¡
.log n ¢ log log n/2¢
multiplicative factor times the optimal work (namely, this
measure is in the sense of the “competitive ratio” of on-line analysis). In addition,
our scheme is modular, integrated, and considers many implementation points.
We comment that, to our knowledge, no previous work on robust parallel computations
has considered fully dynamic faults in the BSP model, or in general distributed
memory systems. Furthermore, this is the first time an efficient Las Vegas
simulation in this area is achieved.
Abstract: In emerging pervasive scenarios, data is collected by sensing devices in streams that occur at several distributed points of observation. The size of the data typically far exceeds the storage and computational capabilities of the tiny devices that have to collect and process them. A general and challenging task is to allow (some of) the nodes of a pervasive network to collectively perform monitoring of a neighbourhood of interest by issuing continuous aggregate queries on the streams observed in its vicinity. This class of algorithms is fully decentralized and diffusive in nature: collecting all the data at a few central nodes of the network is unfeasible in networks of low capability devices or in the presence of massive data sets. Two main problems arise in this scenario: (i) the intrinsic complexity of maintaining statistics over a data stream whose size greatly exceeds the capabilities of the device that performs the computation; (ii) composing the partial outcomes computed at different points of observation into an accurate, global statistic over a neighbourhood of interest, which entails coping with several problems, last but not least the receipt of duplicate information along multiple paths of diffusion.
Streaming techniques have emerged as powerful tools to achieve the general goals described above, in the first place because they assume a computational model in which computational and storage resources are assumed to be far exceeded by the amount of data on which computation occurs. In this contribution, we review the main streaming techniques and provide a classification of the computational problems and the applications they effectively address, with an emphasis on decentralized scenarios, which are of particular interest in pervasive networks
Abstract: We extend the population protocol model with a cover-time service that informs a walking state every time it covers the whole network. This represents a known upper bound on the cover time of a random walk. The cover-time service allows us to introduce termination into population protocols, a capability that is crucial for any distributed system. By reduction to an oracle-model we arrive at a very satisfactory lower bound on the computational power of the model: we prove that it is at least as strong as a Turing Machine of space log n with input commutativity, where n is the number of nodes in the network. We also give a log n-space, but nondeterministic this time, upper bound. Finally, we prove interesting similarities of this model to linear bounded automata.
Abstract: We extend the population protocol model with a cover-time service that informs a walking state every time it covers the whole network. This is simply a known upper bound on the cover time of a random walk. This allows us to introduce termination into population protocols, a capability that is crucial for any distributed system. By reduction to an oracle-model we arrive at a very satisfactory lower bound on the computational power of the model: we prove that it is at least as strong as a Turing Machine of space logn with input commutativity, where n is the number of nodes in the network. We also give a logn-space, but nondeterministic this time, upper bound. Finally, we prove interesting similarities of this model to linear bounded automata.