Tag Archives: Performance Evaluation

15/11/2016 – Talk by Andrea Marin

Title: Fair workload distribution for multi-server systems with pulling strategies
Time: 13:00
Location: Meeting room, Building Zeta
Type: Research Result
Speaker: Andrea Marin
Abstract:
In this talk we present the paper that has received the best paper award at the conference Valuetools 2016.

We consider systems with a single queue and multiple parallel servers. Each server fetches a job from the queue immediately after completing its current work. We propose a pulling strategy that aims at achieving a fair distribution of the number of processed jobs among the servers. We show that if the service times are exponentially distributed then our strategy ensures that in the long run the expected difference among the processed jobs at each server is finite while maintaining a reasonable throughput.
We give the analytical expressions for the stationary distribution and the relevant stationary performance indices like the throughput and the system’s balance.
Interestingly, the proposed strategy can be used to control the join-queue length in fork-join queues and the analytical model gives the closed form expression of the performance indices in saturation.

08/11/2016 – Talk by Filippo Cavallin

Title: Multi-Step Discrete Time Process Algebra (MuPA)
Time: 13:00
Location: Meeting room, Building Zeta
Type: Research Result
Speaker: Filippo Cavallin
Abstract:
We consider Discrete Time Markov Chains with synchronous behaviour, i.e. models in which multiple events may occur at each recorded time in the model. This is based on a conceptual model in which the discrete times are sampling points when the system is observed and at each instant all events that have occurred since the previous observation point will be noted. We propose the first process algebra which is capable of capturing such systems and demonstrate its application to discrete time queues both in isolation and forming networks.

07/09/2016 – Talk by Mohsin Jafri

Title:  Simulating Depth-based routing in Underwater Networks
Time: 13:45
Location: Meeting room
Type: Research Result
Speaker: Mohsin Jafri
Abstract: In this talk, I will briefly discuss my ongoing research activities. I will present the implementation of a simulator for
studying Depth-Based Routing (DBR) in Underwater Wireless Sensor Networks (UWSNs). One of the major challenges for DBR is the configuration of a parameter called “holding time” which influences the network throughput, response time and energy consumption. I will show how it is possible to use the simulator to support the validation of a choice for a given value of the holding time, as well as other network parameters. The presentation will also discuss the important features that distinguish the implemented simulator with respect to the other ones available in the scientific literature and its validation against previous simulation models or measurements.

02/03/2016 – Talk by Filippo Cavallin

Title:  Cronization: a method to pass from Continuous Time to Discrete Time Automata
Time: 14:00
Location: Meeting room, building Zeta
Type: Research Result
Speaker: Filippo Cavallin
Abstract:
Discrete and Continuous Time automata are important formalisms to model and analyze systems such queuing and telecommunication networks. The stationary performance analysis of these automata requires the computation of the steady-state distribution of their underlying Markov chains. Sometimes for the analysis is more convenient to pass from continuous to discrete time. The main methods are focused on the underlying Markov chains and not on the automatas. To preserve the transitions with labels and the synchronization between them, we will introduce a method to pass from a continuous time stochastic automaton (SA) to a probabilistic Input/Output automaton (PIOA). This method is a bijection from continuous to discrete time and it allows us to switch between the two different types of automata and it preserves the parallel composition.

18/11/2015 – Talk by Gian-Luca dei Rossi

Title:  Evaluating the impact of eDoS attacks to cloud facilities
Time: 12:00
Location: Meeting Room, building Zeta
Type: Research Results
Speaker: Gian-Luca Dei Rossi
Abstract:
The complexity of modern cloud facilities requires attentive management policies that should encompass all aspects of the system. Security is a critical issue, as intrusions, misuse or denial of service attacks may damage both the users and the cloud provider including its reputation on the market.
Disruptive attacks happen fast, cause evident and short term damages and are usually the result of operations that are hard to disguise. On the other hand, Energy oriented Denial of Service (eDoS) attacks aim at producing continuous minor damages, eventually with long term consequences. These long lasting attacks are difficult to detect. In this tale we present a model of the behavior of a system under eDoS attack.
We study the impact in terms of cloud energy consumption of an attack strategy previously proposed in the literature and compare it with other strategies that we propose. Our findings show that the strategy previously proposed in the literature, based on keeping the cloud close to saturation, is not optimal (from the point of view of the attacker) in presence of non-constant workload and that there is a trade-off between the aggressiveness of the attacker and the duration of the attack in order to maximize the damage.

27/10/2015 – Talk by Jean-Michel Fourneau (Université de Versailles Saint Quintin)

Title: Discrete Time Stochastic Automata Network with Steady-State Product Form distribution
Time: 13:00
Location: Meeting Room, building Zeta
Type: Research result
Speaker: Jean-Michel Fourneau
Abstract: We present some sufficient conditions for a discrete time Stochastic Automata Networks (SAN) to have a steady-state distribution which has a multiplicative form. The proofs are based on algebraic properties of the tensor operations associated with SAN. Some examples are given.
Bio sketch: J.M. Fourneau is Professor of Computer Science at the University of Versailles St. Quentin, France. He was formerly with Ecole Nationale des Telecommunications, Paris and University of Paris XI Orsay as an Assistant Professor. He graduated in Statistics and Economics from Ecole Nationale de la Statistique et de l’Administation Economique, Paris and he obtained is Ph.D. and his habilitation in Computer Science at the University of Paris XI Orsay in 1987 and 1991 respectively. He is the Head of the Performance Evaluation team within PRiSM laboratory at Versailles University and his recent research interests are algorithmic performance evaluation, Stochastic Automata Networks, G-networks, stochastic bounds, and application to high speed networks, and all optical networks.

06/05/2015: Talk by Benny Van Houdt (University of Antwerp)

Title: Mean field models for SSD garbage collection
Time: 14:00
Location: Conference Room, building Alpha
Type: Research result
Speaker: Benny Van Houdt
Abstract: In this talk we discuss some mean field models for a broad class of garbage collection algorithms for flash-based solid state drives (SSDs) and as well as the insights they provide. We start with a basic introduction on mean field models and SSDs. Next we take a detailed look at the mean field model in the most basic setting (uniform random writes) and discuss its implications. Finally, if time permits, some of the new insights provided by more advanced models with hot and cold data or hot data identification will be presented.

11/03/2015 – Talk by Robbe Block

Title: Spatial Fairness in Multi-Channel CSMA Line Networks
Time: 14:00
Location: Meeting Room, building Zeta
Type: Research Result
Speaker: Robbe Block (University of Antwerp)
Abstract:
In this talk we will consider a line of wifi-hotspots placed in a single line. Starting from earlier results, we will show the steady state of this system has a product form solution and using this product form, a fast method to calculate the per-hotspot throughput will be given. Furthermore, for a single channel setting it was shown in previous work that a simple formula could be used to achieve fairness among the different nodes in the network. It will be shown that the same formula still achieves fairness in the multi-channel setting in either heavy or low traffic conditions, but no such simple formula exists in general. However, through numerical experiments, we see that the fairness index of the multi-channel system still remains very close to one, meaning the simple formula eliminates most of the unfairness in the network.

04/03/2015 Talk by Filippo Cavallin

Title: Single TTL Cache with Reset under the Independent Reference Model
Time: 14:00
Location: Conference room, Building Alfa
Type: Research Result
Speaker: Filippo Cavallin
Abstract:
Cache systems based on a Time To Live (TTL) policy are becoming increasingly important especially in the field of Information-Centric Network design. In this kind of cache an element is kept in it unless it is not used (i.e., it does not receive any write/read request) for a certain period.
This model is less effective if there are some periods of inactivity of the system.
In this talk, after some brief explanations about G-Networks and PEPA language, we will evaluate the performance of a single TTL cache equipped with a refresh system that rejuvenates the timers of the elements stored. In this way we try to avoid the problem of under-utilization of the cache when traffic is low. We will show also that the equilibrium distribution function has a product-form expression under the Independent Reference Model. This allows us to efficiently compute the performance indices.

11/06/2014 – Talk by Ivan Stojic

Title: Optimisation of servers with different quality of services
Time: 14:00
Location: Meeting room
Type: Research Result
Speaker: Ivan Stojic
Abstract:

In many applications involving client/server or equivalent architecture, it is possible to vary the quality of services provided by the servers. Dynamic control of the quality of services is an effective mechanism for control of the performance of the system in conditions of varying system load. One of the basic problems involved in applying this mechanism is optimisation of the controlling policy that determines the level of quality of services based on the state of the system. In this seminar, after a brief introduction to Markov chains and basic queueing models, a Markov chain based model of a multiserver queue with processor-sharing discipline and different quality of services is presented. Calculation of the performance indices from the model and the prohibitive complexity of the exact solution are discussed next, motivating the introduction of a simpler model that heuristically approximates the original model and allows for more efficient model solution and optimisation of the controlling policy.