Michal Kucharzak

Heuristic algorithms for optimization of task allocation and result distribution in peer-to-peer computing systems

Abstract Recently, distributed computing system have been gaining much attention due to a growing demand for various kinds of effective computations in both industry and academia. In this paper, we focus on Peer-to-Peer (P2P) computing systems, also called public-resource computing systems or global computing systems. P2P computing systems, contrary to grids, use personal computers and other relatively simple electronic equipment (e.g., the PlayStation console) to process sophisticated computational projects. A significant example of the P2P computing idea is the BOINC (Berkeley Open Infrastructure for Network Computing) project. To improve the performance of the computing system, we propose to use the P2P approach to distribute results of computational projects, i.e., results are transmitted in the system like in P2P file sharing systems (e.g., BitTorrent). In this work, we concentrate on offline optimization of the P2P computing system including two elements: scheduling of computations and data distribution. The objective is to minimize the system OPEX cost related to data processing and data transmission. We formulate an Integer Linear Problem (ILP) to model the system and apply this formulation to obtain optimal results using the CPLEX solver. Next, we propose two heuristic algorithms that provide results very

ILP model and algorithms for restoration of anycast flows in Elastic Optical Networks

Elastic Optical Network (EON) is a promising technology proposed recently to provide flexible and multi-bitrate data transmission in the optical layer. EONs utilize various modulation format conversion and smaller spectrum granularity, what provide more efficient use of spectrum resources comparing to currently used fix-grid WDM networks. Concurrently, growing popularity of network services provided by Data Centers located in different network nodes and related to cloud computing and Content Delivery Network applications triggers the need to address anycast transmissions as an effective way to improve network performance. This paper focuses on modeling and optimization of anycast flows in EONs in the context of network restoration. Specifically, we formulate an original NP-hard Routing and Spectrum Allocation for Restoration of Anycast Flows (RSA/RAF) problem. In order to solve the problem, we design, implement and evaluate four greedy algorithms and a metaheuristic derived from Simulated Annealing. Extensive numerical experiments prove high practical applicability of presented algorithms.

An improved annealing algorithm for throughput maximization in static overlay-based multicast systems

Overlay-based multicast has been proposed as an efficient solution to ensure multicast communication in the Internet. In contrast to IP multicast, it uses a virtual topology created in an overlay network. For a few past years overlay-based multicast has attracted research community and many new protocols have been proposed. Thereby, a lot of research has been focusing on various aspects of such systems, however most of previous works concern on minimization of flow costs. In this paper, we take a look at the problem of modelling and optimization of overlay-based multicast with a special focus on maximization of systems throughput. We employ an integer program of fractional tree packing problem and propose an improved annealing algorithm for solving the problem. Efficiency of the algorithm is compared to optimal results as well as random ones and a basic simulated annealing.

Maximum flow trees in overlay multicast: Modeling and optimization

Recently, peer-to-peer (P2P) multicast has become more and more popular approach of simultaneous content delivery to a group of users. It is especially valuable for multimedia applications realized in the Internet. Apart from network-based IP multicast, P2P multicast forms an overlay structure of routing where end-hosts actively contribute to the network by sharing their data streams across other receivers. The overlay structure is routed then via IP network using well-defined unicast flows. This paper focuses on modeling and optimization of content routing in overlay multicast systems. Having a set of overlay nodes interested in joining the same, single source multicast transmission, where every node is subject to limited upload and download access link capacity, we aim at providing an optimal overlay multicast in order to maximize the total system throughput and we formulate maximum flow trees problem related to multicast in overlay network. We also present an original Greedy Algorithm and Golden Ratio Heuristic for solving the problem. These algorithms are evaluated in relation to optimal results yielded by CPLEX solver and random benchmarks.

The 802.11g relaying MAC that saves energy

In this paper we focus on energy efficient opportunistic relaying in WLANs. We propose and examine an extension to IEEE 802.11g MAC protocol that enables self-enforced relaying, where the main idea behind the relaying is to decrease overall energy expenses in WLANs. Since frame transmission and reception require spending some energy by wireless nodes as well as access points, it is desirable for every node in a wireless system to get access to the channel as fast as possible and occupy it for a reasonably short period. In our proposal the protocol extension uses RTS/CTS mechanisms to precede every transmission and employs intermediate nodes which may serve as relays. Potential relays in the network are self-motivated to perform as helpers since remaining in idle mode or waiting for channel access also requires energy. Therefore, by helping other ongoing transmissions, relays may decrease total transmission time and increase their possibility of being awarded faster access to wireless medium, which eventually yields decrease in energy consumptions. Experimentation results prove that opportunistic relaying provides energy savings in 802.11g-based WLANs.

A mixed integer formulation for multicast flow assignment in multilayer networks

Multicast has been being developed as a very promising technique to data delivery to a group of destinations simultaneously. Except of first implementation for multicasting in the Internet aka IP Multicast, for past few years some systems defined for end-hosts overlay network have been successfully designed and implemented. Additionally, overlay multicast defined for end-host based networks has been gaining an advantage over IP Multicast during recent years. It derives from the fact that the overlay architecture for multicast streaming provides potential scalability and easy deployment of new protocols independent of the network layer solutions at relatively low costs. Thus overlay multicast and network-aware multicast have been developed concurrently. On the one hand, IP Multicast provide pure network aware multicasting and the lowest packet redundancy on physical links, on the other hand, overlay multicasting creates logical multicast topology which is realized using unicast flows on network level. In this paper we combine advantages of both end-host systems and network awareness and formulate a mixed integer program for optimization of multicast tree topology and its network level unicast realization. The proposed formulation can be applied for deriving either lower bound of flows costs in existing systems or for designing new cooperative multilayer protocols for effective multicast transmission.

Integer programming, Constraint programming, and metaheuristic approaches for static optimization of anycast flows in elastic optical networks

Elastic optical network EON is a novel optical technology introduced recently to provide flexible and multibitrate data transmission in the optical layer. Since many new network services including cloud computing and content delivery networks are provisioned with the use of specialized data centers located in different network nodes, in place of one-to-one unicast transmission, the anycast transmission defined as one-to-one-of-many gains much popularity as a quite simple way to improve network performance. Therefore, this article focuses on modeling and static optimization of anycast flows in EONs. In particular, a NP-hard Routing and Spectrum Allocation for Restoration of Anycast Flows RSA/RAF problem is formulated. Next, various optimization approaches are proposed to solve this problem, namely, integer linear programming ILP using branch and bound algorithm, constraint programming CP, and various heuristic approaches. Extensive numerical experiments are run to evaluate and compare all proposed methods. The main conclusion is that in some cases the CP approach is more efficient than the ILP modeling. Moreover, the results show that the SA algorithm significantly outperforms other heuristic methods.

On modeling of minimum cost multicast topology with multiple static streams in overlay communication networks

Overlay communication networks create logical structures at the top of the Internet. Such a concept is also an efficient approach in the context of multicast applications. Moreover, the overlay multicast has actually outperformed network layer-based multicast provided by the IP protocol and attracted developers of multimedia systems, recently. In this paper, we take a look at modeling techniques and the problem of optimal flow assignment in the meaning of creating a minimum cost multicast topology based on multiple multicast static streams in overlay communication networks. We formulate the problem as an ILP (integer linear programming) model and additionally we develop a new model based on a constraint programming paradigm (CP). In order to compare and evaluate effectiveness of both modeling approaches, we use IBM ILOG CPLEX Optimization Studio that provides an efficient way to build optimization models and solve the problem.

Shortest Multilayered Path for Dimensioning in Two-Layer Connection-Oriented Networks

Multilayered model of network describes recent computers networks compounded of different layers of resources where different topologies in each layer are possible. It raises many new questions and changes most of well understood optimization problems on a single layer network into far from trivial. In this paper we focus attention on flow allocation in a multilayer network and we investigate dimensioning of network involving two layers of resources. In dimensioning problem, Flat Cost strategy allows to reconstruct network consists of many layers to single layer-like network. However considering modularization of capacity in adjacent layers, we indicate that depending on the volume of demand, strategies based on Shortest Multilayered Path and Temporary Flat Cost provide improved results.