4,138 Hits in 5.7 sec

Characterizing, modeling, and generating workload spikes for stateful services

Peter Bodik, Armando Fox, Michael J. Franklin, Michael I. Jordan, David A. Patterson
2010 Proceedings of the 1st ACM symposium on Cloud computing - SoCC '10  
A popular approach is to create a workload model and generate synthetic workload, however, there exists no characterization and model of stateful spikes.  ...  Evaluating the resiliency of stateful Internet services to significant workload spikes and data hotspots requires realistic workload traces that are usually very difficult to obtain.  ...  This suggests that "stress testing for spikes" (especially for stateful services) is more subtle and involves more degrees of freedom than previously thought.  ... 
doi:10.1145/1807128.1807166 dblp:conf/cloud/BodikFFJP10 fatcat:uccqhnnu5bflzk4cmp2mhthlle

Model-Switching: Dealing with Fluctuating Workloads in Machine-Learning-as-a-Service Systems

Jeff Zhang, Sameh Elnikety, Shuayb Zarar, Atul Gupta, Siddharth Garg
2020 USENIX Workshop on Hot Topics in Cloud Computing  
However, existing service ML serving systems have trouble dealing with fluctuating workloads and either drop requests or significantly expand hardware resources in response to load spikes.  ...  We implement Model-Switching within Clipper, a state-of-art DNN model serving system, and demonstrate its advantages over baseline approaches.  ...  Acknowledgements The authors would like to thank Jonathan Mace, as well as the anonymous reviewers for their time, suggestions, and valuable feedback.  ... 
dblp:conf/hotcloud/0001EZGG20 fatcat:xpcqmvjmcjby5k5kf36fghuvwe

Online Spike Detection in Cloud Workloads

Amardeep Mehta, Jonas Durango, Johan Tordsson, Erik Elmroth
2015 2015 IEEE International Conference on Cloud Engineering  
We investigate methods for detection of rapid workload increases (load spikes) for cloud workloads.  ...  Such rapid and unexpected workload spikes are a main cause for poor performance or even crashing applications as the allocated cloud resources become insufficient.  ...  RELATED WORK Previous efforts have been made on the topic of modeling and characterizing workloads and spikes, see [2] , [3] , [1] .  ... 
doi:10.1109/ic2e.2015.50 dblp:conf/ic2e/MehtaDTE15 fatcat:rteiitj3z5cppheosz7bnar3ja

An SLA-oriented capacity planning tool for streaming media services

L. Cherkasova, Wenting Tang, S. Singhal
2004 International Conference on Dependable Systems and Networks, 2004  
In this paper, we propose a new capacity planning framework for evaluating the resources needed for processing a given streaming media workload with specied performance requirements.  ...  This paper addresses the problem of mapping the requirements of a known media service workload into the corresponding system resource requirements and accurately sizing a media server cluster to handle  ...  Acknowledgments: We would like to thank our HPLabs colleagues Kim Keeton, Mustafa Uysal, Renato Santos, and Xiaoyun Zhu for stimulating discussions which helped in improving the presentation of the paper  ... 
doi:10.1109/dsn.2004.1311945 dblp:conf/dsn/CherkasovaTS04 fatcat:ummqp7beprbtnnyvocuwsgdc4u

Evaluating Server Capacity for Streaming Media Services [chapter]

Beomjoo Seo, Michele Covell, Mirjana Spasojevic, Sumit Roy, Roger Zimmermann, Leonidas Kontothanassis, Nina Bhatti
2008 Lecture Notes in Business Information Processing  
In this study, we propose a more general, procedural methodology that characterizes a single system's streaming capacity and derives a prediction model that is applicable for any type of workload imposed  ...  Some of these goals conflict and motivate the development of precise and accurate models that predict the system states under extremely diverse workloads on-the-fly.  ...  While such generalized workload model helps to understand the average behavior of a streaming server, it usually fails to detect whether the server is saturated.  ... 
doi:10.1007/978-3-540-77581-2_8 fatcat:jzwsr3t56zgsxddswouofzbwfa

CoSL: A coordinated statistical learning approach to measuring the capacity of multi-tier websites

Jia Rao, Cheng-Zhong Xu
2008 Proceedings, International Parallel and Distributed Processing Symposium (IPDPS)  
It uses a Bayesian network to correlate low level instrumentation data like system and user cpu time, available memory size, and I/O status that are collected at run-time to high level system states in  ...  It is because a website may have different capacities in terms of the maximum concurrent level when the characteristic of workload changes.  ...  Song Jiang and other group members for their comments and suggestions about this work. The authors are also grateful to the anonymous reviewers for their constructive comments.  ... 
doi:10.1109/ipdps.2008.4536232 dblp:conf/ipps/RaoX08 fatcat:jwnwqg3ovzb57h2cv7vatlcddu

Workload characterization

D.A. Menasce
2003 IEEE Internet Computing  
In addition to their role in capacity planning, we can use good and representative workload models to design synthetic workload generators that mimic the workload's behavior.  ...  planning -the process of predicting when site resources must be upgraded to provide adequate quality of service under increased workloadintensity levels. 5 Function-Layer Characterization Workload characterization  ...  Second, researchers continually seek data to facilitate insight into Internet routing and topology behavior, as well as to build better models of how various protocols and services behave on a local and  ... 
doi:10.1109/mic.2003.1232524 fatcat:yzxwxqxhtrhk5es4xqgzlec7eu

Position paper

Markus Klems, Hoàng Anh Lê
2013 Proceedings of the 2013 international workshop on Hot topics in cloud services - HotTopiCS '13  
Creating system setups for controlled performance evaluation experiments of distributed systems is time-consuming and expensive.  ...  In this paper, we present an experiment automation approach for evaluating distributed systems in compute cloud environments.  ...  We thank the reviewers for their valuable feedback and suggestions which have helped to improve the quality of this paper.  ... 
doi:10.1145/2462307.2462322 dblp:conf/wosp/KlemsL13 fatcat:ko67e4uw6fdgroznimmo47lnja

Capacity planning an essential tool for managing Web services

V.A.F. Almeida, D.A. Menasce
2002 IT Professional Magazine  
the workload, modeling the workload, validating and calibrating the models, predicting the performance, analyzing the costperformance plans, and suggesting actions.  ...  This article introduces capacity planning as an essential tool for managing quality of service on the Web and presents a methodology, where the main steps are: understanding the environment, characterizing  ...  Steps 2 and 3 generate the workload model, which is a synthetic and compact representation of the workload seen by a Web service. 4.  ... 
doi:10.1109/mitp.2002.1046642 fatcat:wrmkxqoigrcbbd6626kiiyjuwq

Measuring Cloud Workload Burstiness

Ahmed Ali-Eldin, Oleg Seleznjev, Sara Sjostedt-de Luna, Johan Tordsson, Erik Elmroth
2014 2014 IEEE/ACM 7th International Conference on Utility and Cloud Computing  
Workload burstiness and spikes are among the main reasons for service disruptions and decrease in the Quality-of-Service (QoS) of online services.  ...  In this paper, we review the state-of-the-art in online identification of workload spikes and quantifying burstiness.  ...  We have surveyed a large body of work on signal analysis, time-series analysis, workload analysis, spike detection and modeling, workload generation and autonomic management.  ... 
doi:10.1109/ucc.2014.87 dblp:conf/ucc/Ali-EldinSLTE14 fatcat:lfuhmiicpzdofahi6hflojmq3u

WASP: Workload Adaptive Energy-Latency Optimization in Server Farms Using Server Low-Power States

Fan Yao, Jingxin Wu, Suresh Subramaniam, Guru Venkataramani
2017 2017 IEEE 10th International Conference on Cloud Computing (CLOUD)  
In this paper, we present WASP, where we explore techniques that make smart use of the processor and system low-power states, and orchestrate their use with workload adaptivity for more effective energy  ...  At the same time, application workloads have Quality of Service (QoS) constraints that need to be satisfied.  ...  ACKNOWLEDGMENT This material is based upon work supported in part by the National Science Foundation under CAREER Award CCF-1149557 and CNS-1320226.  ... 
doi:10.1109/cloud.2017.30 dblp:conf/IEEEcloud/YaoWSV17 fatcat:ozrkfse72bg7lecziry6ew7dzm

Modeling user activities in a large IPTV system

Tongqing Qiu, Zihui Ge, Seungjoon Lee, Jia Wang, Jun Xu, Qi Zhao
2009 Proceedings of the 9th ACM SIGCOMM conference on Internet measurement conference - IMC '09  
We then combine theses models to design an IPTV user activity workload generation tool called SIMULWATCH, which takes a small number of input parameters and generates synthetic workload traces that mimic  ...  We further generalize the findings and develop a series of models for capturing both the probability distribution and time-dynamics of user activities.  ...  In this paper, we model a wider range of aspects of user activities, and design a workload generator.  ... 
doi:10.1145/1644893.1644945 dblp:conf/imc/QiuGLWXZ09 fatcat:h5dlfxjcjja7xgx2mdv5zkfzvi

Predictive topology refinements in distributed stream processing system

Muhammad Hanif, Choonhwa Lee, Sumi Helal, Rashid Mehmood
2020 PLoS ONE  
Predictions are made through a model based on a combination of SVR, autoregressive, and moving average model with a feedback mechanism.  ...  Good quality of service (QoS) is a must for the enterprises, as they strive to survive in a competitive environment.  ...  We generated synthetic workload to mimic different patterns of spikes and show how the system adapts itself accordingly in section 4.  ... 
doi:10.1371/journal.pone.0240424 pmid:33151974 fatcat:mfq6n4x2xbhn3bdggs3t6s6v2a


Michael Brown, Jose Renau
2011 Performance Evaluation Review  
, and service agreement penalties.  ...  The ReRack optimization component finds the best ratio of renewable energy sources for a given location and workload.  ...  ReRack model can be divided into a data center energy consumption model, workload, service level agreement, energy generation for wind and solar which is dependent on the weather patterns, battery model  ... 
doi:10.1145/2160803.2160865 fatcat:ks2cpmmymbaxxke3m55bxkp6oa


Huajing Li, Wang-Chien Lee, Anand Sivasubramaniam, Lee Giles
2007 Proceedings of the 2007 conference on Digital libraries - JCDL '07  
In this paper, we analyze the usage logs of CiteSeer, a scientific literature digital library and search engine, to characterize workloads for both robots and users.  ...  Based on the findings, we propose SearchGen, a synthetic workload generator to output traces for scientific literature digital libraries and search engines.  ...  ACKNOWLEDGEMENTS We gratefully acknowledge partial support from Microsoft Research and NSF.  ... 
doi:10.1145/1255175.1255203 dblp:conf/jcdl/LiLSG07 fatcat:mhkejxtkjbazrdxfaregdq5ssq
« Previous Showing results 1 — 15 out of 4,138 results