Thursday, 14 December 2017

Session 1: Performance Models and Techniques - Part 1

Computing First Passage Time Distributions in Stochastic Well-Formed Nets

Authors:

Gianfranco Balbo (Università di Torino)
Marco Beccuti (Università di Torino)
Massimiliano De Pierro (Università di Torino)
Giuliana Franceschinis (Università del Piemonte Orientale)

Abstract:

The increasing demand for customer centric evaluation of systems, mostly related with the assessment of the quality of service that they can deliver, requires the development of techniques properly designed to model and to study the movement of specific entities generically referred to as "customers". Stochastic Well-Formed Net(SWN) are naturally suited for the representation of systems in which "customers" of different categories compete for the use of common resources. Color classes of SWN are easily associated with these different categories, leaving to the peculiar features of the formalism the possibility of exploiting all the symmetries existing into the representation for the efficient and effective computation of the measures of interest. Within this application context, the computation of first passage time distribution measures in Stochastic Well-Formed Net (SWN) is becoming of primary interest. Customers however are not primitive entities in the formalism and an approach similar to that previously developed for Generalized Stochastic Petri Nets (GSPN) is suggested to overcome this problem in which P-semiflows are used to identify the circulating "customers". In this paper we propose an original algorithm for computing some P-semiflows of colored PNs (in particular of SWNs) in parametric form by exploiting the peculiarities of the objective of this investigation, and extend the customer centric first passage time computation approach previously developed for GSPNs, to make it suitable for SWN models. Moreover, the paper proposes an enhancement of the SWN notation in order to provide a way to ease the modeler in the specification of customer scheduling policies that may affect the computation of first passage time distributions. This extension, inspired by Queueing Petri Nets, adds to SWN some "syntactic sugar" that allows to include in the model queueing places which are automatically replaced by appropriate submodels, before solving the model.

DOI: 10.1145/1958746.1958754

Full text: PDF

[#][]

Detection and Solution of Software Performance Antipatterns in Palladio Architectural Models

Authors:

Catia Trubiani (Università degli Studi dell'Aquila)
Anne Koziolek (Karlsruhe Institute of Technology)

Abstract:

Antipatterns are conceptually similar to patterns in that they document recurring solutions to common design problems. Performance Antipatterns document, from a performance perspective, common mistakes made during software development as well as their solutions. The definition of performance antipatterns concerns software properties that can include static, dynamic, and deployment aspects. Currently, such knowledge is only used by domain experts; the problem of automatically detecting and solving antipatterns within an architectural model had not yet been empirically addressed. In this paper we present an approach to automatically detect and solve software performance antipatterns within the Palladio architectural models: the detection of an antipattern provides a software performance feedback to designers, since it suggests the architectural alternatives to overcome specific performance problems. We implemented the approach and a case study is presented to demonstrate its validity. The system performance under study has been improved by 50% with the use of antipatterns' solutions.

DOI: 10.1145/1958746.1958755

Full text: PDF

[#][]

An Approach for Scalability-Bottleneck Solution: Identification and Elimination of Scalability Bottlenecks in a DBMS

Authors:

Takashi Horikawa (NEC Corporation)

Abstract:

ACID-compliant DBMSs are said to be difficult to scale their performance by using many more processors, which means that they are difficult to enjoy the benefits of recent many-core systems that wide-spread use of multi-core processors has made practicable. Since DBMSs are indispensable in most of IT systems, scalability issues should be addressed to fulfill the demand of handling large quantity of data. This paper proposes a viable approach for solving scalability issue, in which lock-related bottleneck will be identified from event trace based measurements and scalability will be improved by replacing the bottleneck-lock with fine-grained locks. This paper also describes a case study on the application of the proposed method, in which the scalability of a many-core system in executing DBT-1 transactions with MySQL adopting the InnoDB storage engine has been successfully improved. Since applying the proposed method produced the increase in maximum throughput of the 16-CPU system by 1.6 times, the method is promising, as long as lock-related bottlenecks are of concern.

DOI: 10.1145/1958746.1958756

Full text: PDF

[#][]

Experience Building Non-Functional Requirement Models of a Complex Industrial Architecture

Authors:

Daniel Dominguez Gouvêa (Chemtech a Siemens Company)
Cyro de A. Assis D. Muniz (Chemtech a Siemens Company)
Gilson A. Pinto (Chemtech a Siemens Company)
Alberto Avritzer (Siemens Corporate Research)
Rosa Maria Meri Leão (Federal University of Rio de Janeiro)
Edmundo de Souza e Silva (Federal University of Rio de Janeiro)
Morganna Carmem Diniz (Federal University of the State of Rio de Janeiro)
Luca Berardinelli (University of L' Aquila)
Julius C. B. Leite (Universidade Federal Fluminense)
Daniel Mossé (University of Pittsburgh)
Yuanfang Cai (Drexel University)
Mike Dalton (Drexel University)
Lucia Kapova (Karlsruhe Institute of Technology)
Anne Koziolek (Karlsruhe Institute of Technology)

Abstract:

In this paper, we report on our experience with the application of validated models to assess performance, reliability, and adaptability of a complex mission critical system that is being developed to dynamically monitor and control the position of an oil-drilling platform. We present real-time modeling results that show that all tasks are schedulable. We performed stochastic analysis of the distribution of tasks execution time as a function of the number of system interfaces. We report on the variability of task execution times for the expected system configurations. In addition, we have executed a system library for an important task inside the performance model simulator. We report on the measured algorithm convergence as a function of the number of vessel thrusters. We have also studied the system architecture adaptability by comparing the documented system architecture and the implemented source code. We report on the adaptability findings and the recommendations we were able to provide to the system's architect. Finally, we have developed models of hardware and software reliability. We report on hardware reliability results based on the evaluation of the system architecture. As a topic for future work, we report on an approach that we recommend be applied to evaluate the system under study software reliability.

DOI: 10.1145/1958746.1958757

Full text: PDF

[#][]

Relative Roles of Instruction Count and Cycles Per Instruction in WCET Estimation

Authors:

Archana Ravindar (Indian Institute of Science)
Y. N. Srikant (Indian Institute of Science)

Abstract:

Most of the existing WCET estimation methods directly estimate execution time, ET, in cycles. We propose to study ET as a product of two factors, ET = IC * CPI, where IC is instruction count and CPI is cycles per instruction. Considering directly the estimation of ET may lead to a highly pessimistic estimate since implicitly these methods may be using worst case IC and worst case CPI. We hypothesize that there exists a functional relationship between CPI and IC such that CPI=f(IC). This is ascertained by computing the covariance matrix and studying the scatter plots of CPI versus IC. IC and CPI values are obtained by running benchmarks with a large number of inputs using the cycle accurate architectural simulator, Simplescalar on two different architectures. It is shown that the benchmarks can be grouped into different classes based on the CPI versus IC relationship. For some benchmarks like FFT, FIR etc., both IC and CPI are almost a constant irrespective of the input. There are other benchmarks that exhibit a direct or an inverse relationship between CPI and IC. In such a case, one can predict CPI for a given IC as CPI=f(IC). We derive the theoretical worst case IC for a program, denoted as SWIC, using integer linear programming(ILP) and estimate WCET as SWIC*f(SWIC). However, if CPI decreases sharply with IC then measured maximum cycles is observed to be a better estimate. For certain other benchmarks, it is observed that the CPI versus IC relationship is either random or CPI remains constant with varying IC. In such cases, WCET is estimated as the product of SWIC and measured maximum CPI. It is observed that use of the proposed method results in tighter WCET estimates than Chronos, a static WCET analyzer, for most benchmarks for the two architectures considered in this paper.

DOI: 10.1145/1958746.1958758

Full text: PDF

[#][]