Projekt

Zurück zur Übersicht

LoadOpt - Workload Characterization and Optimization for Multicore Systems

Gesuchsteller/in Binder Walter
Nummer 141002
Förderungsinstrument Projektförderung (Abt. I-III)
Forschungseinrichtung Facoltà di scienze informatiche Università della Svizzera italiana
Hochschule Università della Svizzera italiana - USI
Hauptdisziplin Informatik
Beginn/Ende 01.03.2013 - 28.02.2018
Bewilligter Betrag 497'124.00
Alle Daten anzeigen

Keywords (4)

multicores, workload characterization, server consolidation, performance modeling

Lay Summary (Deutsch)

Lead
In diesem Projekt werden neue Methoden erforscht, um das dynamische Verhalten von Anwendungen auf modernen Multicore Architekturen zu charakterisieren. Wir suchen neue Leistungsmodelle für lastabhängige Systeme, die mehrerere Klassen von Anwendungen ausführen. Wir erforschen neue Lastoptimierungsalgorithmen mit dem Ziel, eine optimale Konsolidierung von Anwendungsinstanzen für verschiedene Leistungsziele und Systemkapazitäten zu finden.
Lay summary
Multicore Systeme sind die grundlegenden Bausteine von Rechenzentren und Cloud Computing Plattformen. Multicore Architekturen eignen sich ideal für die Konsolidierung mehrerer Anwendungen auf derselben Maschine, um deren parallele Rechenleistung voll auszunutzen. Effektives Lastmanagement ist jedoch schwierig, weil die Interferenz zwischen konsolidierten Anwendungen deren Leistung erheblich beeinträchtigen kann.

In diesem Projekt werden neue Methoden erforscht, um das dynamische Verhalten von Anwendungen auf modernen Multicore Architekturen zu charakterisieren. Weiters suchen wir neue Leistungsmodelle für lastabhängige Systeme, die mehrerere Klassen von Anwendungen ausführen. Wir erforschen neue Lastoptimierungsalgorithmen mit dem Ziel, eine optimale Konsolidierung von Anwendungsinstanzen für verschiedene Leistungsziele und Systemkapazitäten zu finden. Um zeitlich variierende Lasten handzuhaben, werden wir die entwickelten Konsolidierungsalgorithmen erweitern, um die Migration von Lasten zu ermöglichen.

Dieses Projekt verspricht Resultate, die sowohl für die Wissenschaft als auch für die Industrie bedeutsam sind, da wir die Kapazitätsplanung und das Leistungsmanagement von verschiedenen Arten moderner Multicore Systeme verbessern. Die Ergebnisse des Projekts werden helfen, eine bessere Ausnutzung der verfügbaren Systemressourcen zu erreichen, was auch dazu beiträgt, den Energieverbrauch zu reduzieren.
Direktlink auf Lay Summary Letzte Aktualisierung: 18.02.2013

Verantw. Gesuchsteller/in und weitere Gesuchstellende

Mitarbeitende

Publikationen

Publikation
AutoBench: Finding Workloads That You Need Using Pluggable Hybrid Analyses
Zheng Yudi, Rosà Andrea, Salucci Luca, Li Yao, Sun Haiyang, Javed Omar, Bulej Lubomir, Chen Lydia Y., Qi Zhengwei, Binder Walter (2016), AutoBench: Finding Workloads That You Need Using Pluggable Hybrid Analyses, in IEEE SANER 2016, Osaka, JapanIEEE, Osaka, Japan.
Comprehensive Multi-platform Dynamic Program Analysis for Java and Android
Zheng Yudi, Kell Stephen, Bulej Lubomír, Sun Haiyang, Binder Walter (2016), Comprehensive Multi-platform Dynamic Program Analysis for Java and Android, in IEEE Software, 33(4), 55-63.
Analyzing Distributed Multi-platform Java and Android Applications with ShadowVM
Sun Haiyang, Zheng Yudi, Bulej Lubomír, Kell Stephen, Binder Walter (2015), Analyzing Distributed Multi-platform Java and Android Applications with ShadowVM, in 13th Asian Symposium on Programming Languages and Systems, Pohang, KoreaSpringer International Publishing, Pohang, Korea.
Custom Full-Coverage Dynamic Program Analysis for Android
Sun Haiyang, Zheng Yudi, Bulej Lubomír, Kell Stephen, Binder Walter (2015), Custom Full-Coverage Dynamic Program Analysis for Android, in Companion Proceedings of SPLASH 2015, Pittsburgh, USACM, Pittsburgh, US.
Join the Best Queue: Reducing Performance Variability in Heterogeneous Systems
Spicuglia Sebastiano, Chen Lydia, Binder Walter (2013), Join the Best Queue: Reducing Performance Variability in Heterogeneous Systems, in 2013 IEEE Sixth International Conference on Cloud Computing (CLOUD'13), Santa Clara, CA, USAIEEE, Santa Clara, CA, USA.
On Load Balancing: A Mix-aware Algorithm for Heterogeneous Systems
Spicuglia Sebastiano, Björkqvist Mathias, Chen Lydia, Serazzi Giuseppe, Binder Walter, Smirni Evgenia (2013), On Load Balancing: A Mix-aware Algorithm for Heterogeneous Systems, in Proceedings of the 4th ACM/SPEC International Conference on Performance Engineering (ICPE'13), Prague, Czech RepublicACM, Prague, Czech Republic.
QoS-aware Service VM Provisioning in Clouds: Experiences, Models, and Cost Analysis
Björkqvist Mathias, Spicuglia Sebastiano, Chen Lydia, Binder Walter (2013), QoS-aware Service VM Provisioning in Clouds: Experiences, Models, and Cost Analysis, in 11th International Conference on Service-Oriented Computing (ICSOC'13), Berlin, GermanySpringer Berlin Heidelberg, Berlin, Germany.
An Endpoint Communication Profiling Tool for Distributed Computing Frameworks
Rosà Andrea, Chen Lydia Y., Binder Walter (2016), An Endpoint Communication Profiling Tool for Distributed Computing Frameworks, in 36th IEEE International Conference on Distributed Computing Systems (ICDCS 2016), Nara, JapanIEEE, Nara, Japan.
Extended Code Coverage for AspectJ-based Runtime Verification Tools
Javed Omar, Zheng Yudi, Rosà Andrea, Sun Haiyang, Binder Walter (2016), Extended Code Coverage for AspectJ-based Runtime Verification Tools, in Proceedings of the 16th International Conference on Runtime Verification (RV), Madrid, SpainSpringer, Madrid, Spain.
Adaptable Runtime Monitoring for the Java Virtual Machine
Rosà Andrea, Zheng Yudi, Sun Haiyang, Javed Omar, Binder Walter (2016), Adaptable Runtime Monitoring for the Java Virtual Machine, in ISoLA 2016, Corfu, GreeceSpringer, Corfu, Greece.
Profiling Actor Utilization and Communication in Akka
Rosà Andrea, Chen Lydia Y., Binder Walter (2016), Profiling Actor Utilization and Communication in Akka, in Proceedings of the 15th ACM SIGPLAN Erlang Workshop (Erlang), Nara, JapanACM, Nara, Japan.
Efficient Profiling of Actor-based Applications in Parallel and Distributed Systems
Rosà Andrea, Chen Lydia Y., Binder Walter (2016), Efficient Profiling of Actor-based Applications in Parallel and Distributed Systems, in ICOOOLPS 2016, Rome, ItalySpringer, Rome, Italy.
AkkaProf: a Profiler for Akka Actors in Parallel and Distributed Applications
Rosà Andrea, Chen Lydia Y., Binder Walter (2016), AkkaProf: a Profiler for Akka Actors in Parallel and Distributed Applications, in Proceedings of the 14th Asian Symposium on Programming Languages and Systems (APLAS), Hanoi, VietnamSpringer, Hanoi, Vietnam.
Actor Profiling in Virtual Execution Environments
Rosà Andrea, Chen Lydia Y., Binder Walter (2016), Actor Profiling in Virtual Execution Environments, in 15th International Conference on Generative Programming: Concepts & Experience (GPCE'16), Amsterdam, The NetherlandsACM, Amsterdam, The Netherlands.
What to Expect when you are Consolidating: Effective Prediction Models of Application Performance on Multicores
Chen Lydia, Serazzi Giuseppe, Ansaloni Danilo, Smirni Evgenia, Binder Walter (2013), What to Expect when you are Consolidating: Effective Prediction Models of Application Performance on Multicores, in Cluster Computing, 17(1), 19-37.
Parallelism Profiling and Wall-time Prediction for Multi-threaded Applications
Peternier Achille, Binder Walter, Yokokawa Akira, Chen Lydia (2013), Parallelism Profiling and Wall-time Prediction for Multi-threaded Applications, in Proceedings of the 4th ACM/SPEC International Conference on Performance Engineering (ICPE'13), Prague, Czech RepublicACM, Prague, Czech Republic.
Optimizing Capacity Allocation for Big Data Applications in Cloud Datacenters
Spicuglia Sebastiano, Chen Lydia Y., Birke Robert, Binder Walter (2015), Optimizing Capacity Allocation for Big Data Applications in Cloud Datacenters, in 2015 IFIP/IEEE International Symposium on Integrated Network Management (IM-2015), Ottawa, ON, CanadaIEEE, Ottawa, ON, Canada.
Catching the Response Time Tail in the Cloud
Spicuglia Sebastiano, Björkqvist Mathias, Chen Lydia Y., Binder Walter (2015), Catching the Response Time Tail in the Cloud, in 2015 IFIP/IEEE International Symposium on Integrated Network Management (IM-2015), Ottawa, ON, CanadaIEEE, Ottawa, ON, Canada.
Catching Failures of Failures at Big-Data Clusters: A Two-Level Neural Network Approach
Rosà Andrea, Chen Lydia Y., Binder Walter (2015), Catching Failures of Failures at Big-Data Clusters: A Two-Level Neural Network Approach, in Proceedings of the 2015 IEEE 23rd International Symposium of Quality of Service (IWQoS), Portland, OR, USAIEEE, Portland, OR, USA.
When Things Turn Sour at Big Data Clusters: Understanding Unsuccessful Executions
Rosà Andrea, Chen Lydia Y., Binder Walter (2014), When Things Turn Sour at Big Data Clusters: Understanding Unsuccessful Executions, ACM Symposium on Cloud Computing, Seattle, WA.
Quantifying the Brown Side of Priority Schedulers: Lessons from Big Clusters
Çavdar Derya, Rosà Andrea, Chen Lydia Y., Binder Walter, Alagöz Fatih (2014), Quantifying the Brown Side of Priority Schedulers: Lessons from Big Clusters, in SIGMETRICS Perform. Eval. Rev., 42(3), 76-81.
Predicting and Mitigating Jobs Failures in Big Data Clusters
Rosà Andrea, Chen Lydia Y., Binder Walter (2015), Predicting and Mitigating Jobs Failures in Big Data Clusters, in Proceedings of the 2015 15th IEEE/ACM International Symposium on Cluster, Cloud and Grid Computing, Shenzhen, ChinaIEEE, Shenzhen, China.
Understanding Unsuccessful Executions in Big-Data Systems
Rosà Andrea, Chen Lydia Y., Binder Walter (2015), Understanding Unsuccessful Executions in Big-Data Systems, in Proceedings of the 2015 15th IEEE/ACM International Symposium on Cluster, Cloud and Grid Computing, Shenzhen, ChinaIEEE, Shenzhen, China.
Understanding the Dark Side of Big Data Clusters: An Analysis beyond Failures
Rosà Andrea, Chen Lydia Y., Binder Walter (2015), Understanding the Dark Side of Big Data Clusters: An Analysis beyond Failures, in 2015 45th Annual IEEE/IFIP International Conference on Dependable Systems and Networks (DSN), Rio de Janeiro, BrazilIEEE, Rio de Janeiro, Brazil.
ParSim: A Tool for Workload Modeling and Reproduction of Parallel Applications
Rosà Andrea, Binder Walter, Chen Lydia Y., Gribaudo Marco, Serazzi Giuseppe (2014), ParSim: A Tool for Workload Modeling and Reproduction of Parallel Applications, in 22nd Int.Simp. on Modelling, Analysis Simulation of Computer and Telecommunication Systems (MASCOTS), Paris, FranceIEEE, Paris, France.
Demystifying Casualties of Evictions in Big Data Priority Scheduling
Rosà Andrea, Chen Lydia Y., Birke Robert, Binder Walter (2015), Demystifying Casualties of Evictions in Big Data Priority Scheduling, in SIGMETRICS Perform. Eval. Rev., 42(4), 12-21.
When Virtual Meets Physical at the Edge: A Field Study on Datacenters' Virtual Traffic
Birke Robert, Björkqvist Mathias, Minkenberg Cyriel, Schmatz Martin, Chen Lydia Y. (2015), When Virtual Meets Physical at the Edge: A Field Study on Datacenters' Virtual Traffic, in 2015 ACM SIGMETRICS International Conference on Measurement and Modeling of Computer Systems, Portland, ORACM, Portland, OR.
Recouping Energy Costs From Cloud Tenants: Tenant Demand Response Aware Pricing Design
Wang Cheng, Nasiriani Neda, Kesidis George, Urgaonkar Bhuvan, Wang Qian, Chen Lydia Y., Gupta Aayush, Birke Robert (2015), Recouping Energy Costs From Cloud Tenants: Tenant Demand Response Aware Pricing Design, in 2015 ACM Sixth International Conference on Future Energy Systems (e-Energy), Bangalore, IndiaACM, Bangalore, India.
Analyzing the Impact of CPU Pinning and Partial CPU Loads on Performance and Energy Efficiency
Podzimek Andrej, Bulej Lubomír, Chen Lydia Y., Binder Walter, Tůma Petr (2015), Analyzing the Impact of CPU Pinning and Partial CPU Loads on Performance and Energy Efficiency, in 2015 15th IEEE/ACM International Symposium on Cluster, Cloud and Grid Computing (CCGrid), Shenzhen, ChinaIEEE, Shenzhen, China.
A simulation framework for priority scheduling on heterogeneous clusters
Çavdar Derya, Birke Robert, Chen Lydia Y., Alagöz Fatih (2015), A simulation framework for priority scheduling on heterogeneous clusters, in Future Generation Computer Systems, 52(C), 37-48.
On Fair Attribution of Costs Under Peak-based Pricing to Cloud Tenants
Nasiriani Neda, Wang Cheng, Kesidis George, Urgaonkar Bhuvan, Chen Lydia Y., Birke Robert (2015), On Fair Attribution of Costs Under Peak-based Pricing to Cloud Tenants, in 23rd Int.Simp. on Modelling, Analysis Simulation of Computer and Telecommunication Systems (MASCOTS), Atlanta, GA, USAIEEE, Atlanta, GA, USA.
Recouping Energy Costs from Cloud Tenants: Tenant Demand Response Aware Pricing Design
Wang Cheng, Nasiriani Neda, Kesidis George, Urgaonkar Bhuvan, Wang Qian, Chen Lydia Y., Gupta Aayush, Birke Robert (2015), Recouping Energy Costs from Cloud Tenants: Tenant Demand Response Aware Pricing Design, in 6th ACM International Conference on Future Energy Systems (ACM e-Energy), Bangalore, IndiaACM, Bangalore, India.
Failure Analysis and Prediction for Big-Data Systems
Rosà Andrea, Chen Lydia Y., Binder Walter (accepted), Failure Analysis and Prediction for Big-Data Systems, in IEEE Transactions on Services Computing, 1-14.
Priority Scheduling for Heterogeneous Workloads: Tradeoff Between Evictions and Response Time
Çavdar Derya, Chen Lydia Y., Alagoz Fatih (accepted), Priority Scheduling for Heterogeneous Workloads: Tradeoff Between Evictions and Response Time, in IEEE Systems Journal, 1-12.
ADRENALIN-RV: Android Runtime Verification using Load-time Weaving
Sun Haiyang, Rosà Andrea, Javed Omar, Binder Walter (accepted), ADRENALIN-RV: Android Runtime Verification using Load-time Weaving, in 10th IEEE International Conference on Software Testing, Verification and Validation (ICST), Tokyo, JapanIEEE, Tokyo, Japan.
Optimizing for Tail Sojourn Times of Cloud Clusters
Björkqvist Mathias, Gautam Natarajan, Birke Robert, Chen Lydia Y., Binder Walter (accepted), Optimizing for Tail Sojourn Times of Cloud Clusters, in IEEE Transactions on Cloud Computing, 1-14.

Zusammenarbeit

Gruppe / Person Land
Felder der Zusammenarbeit
Politecnico di Milano (Europa)
- vertiefter/weiterführender Austausch von Ansätzen, Methoden oder Resultaten
- Publikation
University Texas A&M (Nordamerika)
- vertiefter/weiterführender Austausch von Ansätzen, Methoden oder Resultaten
- Publikation
- Austausch von Mitarbeitern
College of William and Mary (Nordamerika)
- vertiefter/weiterführender Austausch von Ansätzen, Methoden oder Resultaten
- Publikation
- Austausch von Mitarbeitern
Software Technology Group, TU Darmstadt (Europa)
- vertiefter/weiterführender Austausch von Ansätzen, Methoden oder Resultaten
- Publikation
Computer Systems Group, Penn State University (Nordamerika)
- vertiefter/weiterführender Austausch von Ansätzen, Methoden oder Resultaten
- Austausch von Mitarbeitern

Wissenschaftliche Veranstaltungen

Aktiver Beitrag

Titel Art des Beitrags Titel des Artikels oder Beitrages Datum Ort Beteiligte Personen
14th Asian Symposium on Programming Languages and Systems (APLAS'16) Vortrag im Rahmen einer Tagung AkkaProf: a Profiler for Akka Actors in Parallel and Distributed Applications. 21.11.2016 Hanoi, Vietnam Binder Walter; Rosà Andrea
15th International Conference on Generative Programming: Concepts & Experience (GPCE'16) Vortrag im Rahmen einer Tagung Actor Profiling in Virtual Execution Environments 31.10.2016 Amsterdam, Niederlande Binder Walter; Rosà Andrea
7th International Symposium on Leveraging Applications of Formal Methods, Verification and Validation (ISoLA'16) Vortrag im Rahmen einer Tagung Adaptable Runtime Monitoring for the Java Virtual Machine 10.10.2016 Corfu, Griechenland Binder Walter
16th International Conference on Runtime Verification (RV'16) Vortrag im Rahmen einer Tagung Extended Code Coverage for AspectJ-based Runtime Verification Tools 23.09.2016 Madrid, Spanien Binder Walter; Sun Haiyang
15th ACM SIGPLAN Erlang Workshop (Erlang'16) Vortrag im Rahmen einer Tagung Profiling Actor Utilization and Communication in Akka 18.09.2016 Nara, Japan Binder Walter; Rosà Andrea
11th Workshop on Implementation, Compilation, Optimization of Object-Oriented Languages, Programs and Systems (ICOOOLPS'16) Vortrag im Rahmen einer Tagung Efficient Profiling of Actor-based Applications in Parallel and Distributed Systems 18.07.2016 Rome, Italien Rosà Andrea
36th IEEE International Conference on Distributed Computing Systems Poster An Endpoint Communication Profiling Tool for Distributed Computing Frameworks 27.06.2016 Nara, Japan Chen Lydia Yiyu; Binder Walter; Rosà Andrea
Seminar talk at the Kyoto University Einzelvortrag AutoBench: Finding Workloads That You Need Using Pluggable Hybrid Analyses 22.03.2016 Kyoto, Japan Rosà Andrea; Binder Walter
23rd IEEE International Conference on Software Analysis, Evolution, and Reengineering Vortrag im Rahmen einer Tagung AutoBench: Finding Workloads That You Need Using Pluggable Hybrid Analyses 15.03.2016 Osaka, Japan Binder Walter; Rosà Andrea
Seminar talk at the Tokyo Institute of Technology Einzelvortrag AutoBench: Finding Workloads That You Need Using Pluggable Hybrid Analyses 11.03.2016 Tokyo, Japan Binder Walter; Rosà Andrea
Seminar talk at the University of Tokyo Einzelvortrag AutoBench: Finding Workloads That You Need Using Pluggable Hybrid Analyses 10.03.2016 Tokyo, Japan Rosà Andrea; Binder Walter
13th Asian Symposium on Programming Languages and Systems Vortrag im Rahmen einer Tagung Analyzing Distributed Multi-platform Java and Android Applications with ShadowVM 30.11.2015 Pohang, Korea, Republik (Südkorea) Binder Walter; Sun Haiyang
Companion Proceedings of SPLASH 2015 Vortrag im Rahmen einer Tagung Custom Full-Coverage Dynamic Program Analysis for Android 23.10.2015 Pittsburgh, Vereinigte Staaten von Amerika Sun Haiyang
1st Truffle/Graal Languages Workshop Vortrag im Rahmen einer Tagung Accurate Bytecode-level Profiling of Dynamically Optimized Code with Graal 07.07.2015 Prague, Tschechische Republik Binder Walter; Sun Haiyang
45th Annual IEEE/IFIP International Conference on Dependable Systems and Networks Vortrag im Rahmen einer Tagung Understanding the Dark Side of Big Data Clusters: An Analysis beyond Failures 22.06.2015 Rio de Janeiro, Brasilien Chen Lydia Yiyu; Rosà Andrea
2015 ACM SIGMETRICS International Conference on Measurement and Modeling of Computer Systems Vortrag im Rahmen einer Tagung When Virtual Meets Physical at the Edge: A Field Study on Datacenters' Virtual Traffic 15.06.2015 Portland, OR, Vereinigte Staaten von Amerika Rosà Andrea; Chen Lydia Yiyu
23rd IEEE International Symposium of Quality of Service Vortrag im Rahmen einer Tagung Contention Detection by Throttling: a Black-box On-line Approach 15.06.2015 Portland, OR, Vereinigte Staaten von Amerika Chen Lydia Yiyu; Rosà Andrea
23rd IEEE International Symposium of Quality of Service Vortrag im Rahmen einer Tagung Catching Failures of Failures at Big-Data Clusters: A Two-Level Neural Network Approach 15.06.2015 Portland, OR, Vereinigte Staaten von Amerika Rosà Andrea; Chen Lydia Yiyu
23rd IEEE International Symposium of Quality of Service Vortrag im Rahmen einer Tagung Catching Failures of Failures at Big-Data Clusters: a Two-Level Neural Network Approach 15.06.2015 Portland, OR, Vereinigte Staaten von Amerika Rosà Andrea; Chen Lydia Yiyu
2015 IFIP/IEEE International Symposium on Integrated Network Management (IM-2015) Vortrag im Rahmen einer Tagung Optimizing Capacity Allocation for Big Data Applications in Cloud Datacenters 11.05.2015 Ottawa, Kanada Spicuglia Sebastiano
2015 IFIP/IEEE International Symposium on Integrated Network Management (IM-2015) Vortrag im Rahmen einer Tagung Catching the Response Time Tail in the Cloud 11.05.2015 Ottawa, Kanada Spicuglia Sebastiano
15th IEEE/ACM International Symposium on Cluster, Cloud and Grid Computing Vortrag im Rahmen einer Tagung Predicting and Mitigating Jobs Failures in Big Data Clusters 06.05.2015 Shenzhen, China Binder Walter; Chen Lydia Yiyu; Rosà Andrea
15th IEEE/ACM International Symposium on Cluster, Cloud and Grid Computing Poster Understanding Unsuccessful Executions in Big-Data Systems 06.05.2015 Shenzhen, China Binder Walter; Chen Lydia Yiyu; Rosà Andrea
15th IEEE/ACM International Symposium on Cluster, Cloud and Grid Computing Vortrag im Rahmen einer Tagung Analyzing the Impact of CPU Pinning and Partial CPU Loads on Performance and Energy Efficiency 04.05.2015 Shenzhen, China Chen Lydia Yiyu; Binder Walter
ACM Symposium on Cloud Computing 2014 Poster When Things Turn Sour at Big Data Clusters: Understanding Unsuccessful Executions 03.11.2014 Seattle, WA, Vereinigte Staaten von Amerika Rosà Andrea; Chen Lydia Yiyu
IEEE 22nd International Symposium on Modelling, Analysis & Simulation of Computer and Telecommunication Systems Vortrag im Rahmen einer Tagung ParSim: a Tool for Workload Modeling and Reproduction of Parallel Applications 09.09.2014 Paris, Frankreich Rosà Andrea
2014 ACM SIGMETRICS International Conference on Measurement and Modeling of Computer Systems Vortrag im Rahmen einer Tagung Quantifying the Brown Side of Priority Schedulers: Lessons from Big Clusters 16.06.2014 Austin, TX, Vereinigte Staaten von Amerika Rosà Andrea
11th International Conference on Service Oriented Computing Vortrag im Rahmen einer Tagung QoS-Aware Service VM Provisioning in Clouds: Experiences, Models, and Cost Analysis 02.12.2013 Berlin, Deutschland Spicuglia Sebastiano
IEEE 6th International Conference on Cloud Computing Vortrag im Rahmen einer Tagung Join the Best Queue: Reducing Performance Variability in Heterogeneous Systems 27.06.2013 Santa Clara, Vereinigte Staaten von Amerika Spicuglia Sebastiano
4th ACM/SPEC International Conference on Performance Engineering Vortrag im Rahmen einer Tagung On Load Balancing: A Mix-Aware Algorithm for Heterogeneous Systems 21.04.2013 Prague, Tschechische Republik Spicuglia Sebastiano; Chen Lydia Yiyu; Binder Walter


Selber organisiert

Titel Datum Ort
3rd Virtual Machine Meetup (VMM '16) 01.09.2016 Lugano, Schweiz
13th International Conference on Modularity 22.04.2014 Lugano, Schweiz

Auszeichnungen

Titel Jahr
ACM Student Scholarship to attend the ACM Symposium on Cloud Computing 2014 (SoCC 2014) 2014
ACM Student Scholarship to attend the 27th ACM Symposium on Parallelism in Algorithms and Architectures (SPAA 2015) 2015
ACM Student Scholarship to attend the 45th Annual IEEE/IFIP International Conference on Dependable Systems and Networks (DSN 2015) 2015
Best paper runner-up award for "Analyzing the Impact of CPU Pinning and Partial CPU Loads on Performance and Energy Efficiency" 2015
IEEE Student Scholarship to attend the 36th IEEE International Conference on Distributed Computing Systems (ICDCS 2016) 2016
ACM SIGPLAN PAC Student Scholarship to attend the 21st ACM SIGPLAN International Conference on Functional Programming (ICFP 2016) 2016
ACM SIGPLAN PAC Student Scholarship to attend the 2016 ACM SIGPLAN conference on Systems, Programming, Languages and Applications: Software for Humanity (SPLASH) 2016

Verbundene Projekte

Nummer Titel Start Förderungsinstrument
153560 Fundamentals of Parallel Programming for Platform-as-a-Service Clouds 01.11.2014 Projektförderung (Abt. I-III)
136225 FAN: Foundations of dynamic program ANalysis 01.04.2012 Sinergia
127386 SOSOA: Self-Organizing Service-Oriented Architectures 01.02.2010 Sinergia

Abstract

Multicore architectures have become the standard both for general-purpose PCs and for high-end servers. Such platforms provide an environment that is ideally suited for consolidating multiple applications within the same machine to fully take advantage of the available parallelism. As multicores become the basic building blocks of data centers and Cloud computing platforms, the problem of effective workload management becomes very challenging because the interference between collocated workloads may significantly affect their performance. Moreover, workloads are often observed to be executed with different processing rates for different system resources (e.g., CPU and disk) depending on the overall system load. Empirical work on characterizing load-dependent overhead focuses on a specific layer or component (e.g., considering only caches) while executing a single instance of an application in isolation. Existing profiling approaches lack the holistic view of the overall system and are not directly applicable for analyzing collocated applications. Most modeling work for predicting the performance of multicore systems aims at capturing low-level performance metrics such as cache misses. To obtain application-level metrics such as response time, queueing analysis considering multiple classes of workloads (e.g., CPU-bound or I/O-bound workloads) offers practical methods, but requires proper assumptions on the workload distribution and on the processing characteristics of resources. Regarding workload consolidation strategies, existing research work mainly considers hardware virtualization using one or multiple system resources independently, whereas the spatial dependency of multiple applications on different resources is not taken into account. Clearly, there is a lack of adequate methods for optimizing applications in multicore systems that experience performance interference of collocated workloads and load-dependent processing rates. This project will address these issues from two ends. First, we will explore novel methods to characterize the dynamic behavior of applications running on modern multicores. We will develop practical methods to profile the spatial and temporal characteristics of applications. We will vertically profile load-dependent application behavior across the entire system stack, paying also attention to overheads due to virtualization. For the temporal characterization, we will devise sampling approaches that automatically adjust the sampling frequency to extract desired statistical properties and stochastic processes. We will leverage and refine existing dimension reduction and clustering methods to identify predominant load-dependent resources and representative workload classes. Our goal is to provide the statistics necessary to capture the temporal and spatial workload characteristics as well as to build suitable performance models. Second, we will investigate novel performance models of load-dependent systems executing multiple classes of workloads. Based on our performance analysis, we will explore new load optimization algorithms aiming at optimally collocating application instances taking various performance objectives and system capacities into account. Our aim is to provide a set of optimal rules in collocating application instances in a wide range of system scenarios, applying rigorous mathematical optimization. To manage time-varying workloads, we will extend the developed workload consolidation algorithm into a dynamic framework, focusing on admission control and migration of workloads. Furthermore, we will explore virtual resource provisioning to optimize the system for a given workload at runtime. This project promises to have a significant impact both in academia and in industry. It will improve capacity planning and performance management for various kinds of modern multicore systems. The results of the project will help avoid expensive over-provisioning of hardware resources and achieve better utilization of the available system resources, which in turn helps reduce energy consumption. This project is supported by an IBM Shared University Research (SUR) grant, and the project team will have access to a wide range of multicore systems managed by IBM. This project asks for the funding of two PhD students.