Imperial College London

MrZIFENGNIU

Faculty of EngineeringDepartment of Computing

Research Postgraduate
 
 
 
//

Contact

 

zifeng.niu19 Website

 
 
//

Location

 

Huxley BuildingSouth Kensington Campus

//

Summary

 

Publications

Publication Type
Year
to

6 results found

Dobre R-A, Niu Z, Casale G, 2024, Approximating Fork-Join Systems via Mixed Model Transformations, Workshop on Challenges in Performance Methods for Software Development (WOSP-C)

Conference paper

Niu Z, Casale G, 2024, Neural density estimation of response times in layered software systems, IEEE Transactions on Software Engineering, ISSN: 0098-5589

Layered queueing networks (LQNs) are a class of performance models for software systems in which multiple distributed resources may be possessed simultaneously by a job. Estimating response times in a layered system is an essential but challenging analysis dimension in Quality of Service (QoS) assessment. Current analytic methods are capable of providing accurate estimates of mean response times. However, accurately approximating response time distributions used in service-level objective analysis is a demanding task. This paper proposes a novel hybrid framework that leverages phase-type (PH) distributions and neural networks to provide accurate density estimates of response times in layered queueing networks. The core step of this framework is to recursively obtain response time distributions in the submodels that are used to analyse the network by means of decomposition. We describe these response time distributions as a mixture of density functions for which we learn the parameters through a Mixture Density Network (MDN). The approach recursively propagates MDN predictions across software layers using PH distributions and performs repeated moment-matching based refitting to efficiently estimate end-to-end response time densities. Extensive numerical experiment results show that our scheme significantly improves density estimations compared to the state-of-the-art.

Journal article

Niu Z, 2023, Graph learning based performance analysis for queueing networks, ACM SIGMETRICS Performance Evaluation Review, Vol: 51, Pages: 6-7, ISSN: 0163-5999

Queueing networks serve as a popular performance model in the analysis of business processes and computer systems [4]. Solving queueing network models helps the decision making of system designers. System response time and throughput are two key performance measures in queueing networks. The most widely used algorithms for solving these measures are mean value analysis (MVA) and its approximate extensions [3, §8-9]. However, conventional analytic methods are inaccurate at solving non-product form queueing networks that are frequently encountered in modeling most real systems. Approximation formulas typically rely on assumptions that may lead, on particular regions of the parameters, to inaccurate and misleading results. Simulation modeling is an accurate way, but it requires to be designed for each specific problem, and usually takes longer time to converge.

Journal article

Casale G, Gao Y, Niu Z, Zhu Let al., 2023, LN: a flexible algorithmic framework for layered queueing network analysis, ACM Transactions on Modeling and Computer Simulation, ISSN: 1049-3301

Layered queueing networks (LQNs) are an extension of ordinary queueing networks to model simultaneous resource possession andstochastic call graphs in distributed systems. Existing computational algorithms for LQNs have primarily focused on mean-valueanalysis. However, other solution paradigms, such as normalizing constant analysis and mean-field approximation, can improve thecomputation of LQN mean and transient performance metrics, state probabilities, and response time distributions. Motivated by thisobservation, we propose the first LQN meta-solver, called LN, that allows for the dynamic selection of the performance analysisparadigm to be iteratively applied to the submodels arising from layer decomposition. We report experiments where this addedflexibility helps us to reduce the LQN solution errors. We also demonstrate that the meta-solver approach eases the integration ofLQNs with other formalisms, such as caching models, enabling the analysis of more general classes of layered stochastic networks.Additionally, to support the accurate evaluation of the LQN submodels, we develop novel algorithms for homogeneous queueingnetworks consisting of an infinite server node and a set of identical queueing stations. In particular, we propose an exact method ofmoment algorithms, integration techniques for normalizing constants, and a fast non-iterative mean-value analysis technique.

Journal article

Niu Z, Casale G, 2022, A Mixture Density Network approach to predicting response times in layered systems, IEEE MASCOTS 2021, Publisher: IEEE, Pages: 1-8

Layering is a common feature in modern service-based systems. The characterization of response times in a layered system is an important but challenging analysis dimension inQuality of Service (QoS) assessment. In this paper, we develop anovel approach to estimate the mean and variance of response time in systems that may be abstracted as layered queueing networks. The core step of the method is to obtain the response time distributions in the submodels that are used to analyze the layered queueing networks by means of decomposition. We model the conditional response time distribution as a mixture of Gamma density functions for which we learn the parametersby means of a Mixture Density Network (MDN). The scheme recursively propagates the MDN predictions through the layersusing phase-type distributions and performs convolutions togain the approximation of the system delay. The experimental results show an accurate match between simulations and MDN predictions and also verify the effectiveness of the approach.

Conference paper

Casale G, Gao Y, Niu Z, Zhu Let al., 2022, LN: a meta-solver for layered queueing network analysis, International Conference on Quantitative Evaluation of SysTems (QEST 2022), Publisher: Springer, Pages: 232-254, ISSN: 0302-9743

We overview LN, a novel solver introduced in the LINE soft-ware package to analyze layered queueing network (LQN) models. Thenovelty of the LN solver lies in its capability to analyze LQNs with a user-defined combination of solution paradigms, including discrete-event andstochastic simulation, continuous-time Markov chain analysis (CTMC),normalizing constant evaluation (NC), matrix analytic methods (MAM),mean-field approximations (FLUID), and mean-value analysis (MVA).Being parametric in the solver used for each LQN layer, LN as a wholeenables the efficient computation of advanced performance metrics suchas marginal and joint state probabilities, response and passage time distributions, and transient measures, leveraging individual strengths of thesupported solution paradigms. We discuss in particular recent developments added to NC, the default layer solver of LN, which significantlyimprove the solution of queueing network models obtained using looselayering

Conference paper

This data is extracted from the Web of Science and reproduced under a licence from Thomson Reuters. You may not copy or re-distribute this data in whole or in part without the written consent of the Science business of Thomson Reuters.

Request URL: http://wlsprd.imperial.ac.uk:80/respub/WEB-INF/jsp/search-html.jsp Request URI: /respub/WEB-INF/jsp/search-html.jsp Query String: respub-action=search.html&id=01744817&limit=30&person=true