面向突发业务的云服务并发量应对策略研究
|
摘要
云服务突发并发量是导致服务质量(QoS)降级的重要因素.传统的突发并发量应对策略存在时效性差、资源利用率低的问题,会导致服务请求响应时间增长、请求违例率及拒绝率增大.针对上述问题,该文首先提出了基于多级队列的并发量分级缓存机制ATBM,可有效避免突发并发量对缓存队列的持续拥堵,确保缓存队列中请求的持续流通,进而提高整体的QoS.在此基础上,提出了一个主动式突发并发量应对策略GMAC,该策略能以较低的欠预测比例及规模预测用户并发量,得出欠分配规模较低的资源需求量,并根据系统当前的资源配置、并发量处理情况以及SLA(Service Level Agreement),全面、准确地量化资源可用量,最后通过对比资源需求量及可用量,提前制定及执行资源调整策略.该文采用真实并发量数据对ATBM在改善QoS方面的有效性进行验证,最后利用对比实验就GMAC的突发并发量应对效果进行评估.实验结果表明,ATBM下并发量的违例率仅为传统单级缓存队列下违例率的24.6%;相对于根据并发量峰值预留,GMAC可在保证服务不发生SLA违例的前提下,将系统资源利用率提高1.45倍;与基于并发量均值预留资源相比,GMAC能将系统的平均响应时间及拒绝率分别降低8.6%及16.9%.
Bursty workload is a crucial factor that deteriorates QoS(Quality of Service) of cloud services.However,traditional coping strategies for bursty workloads suffer disadvantages as poor timeliness and low resource utilization rate,which not only lengthens request response time but also increases request violation and rejection rate.To solve these problems,we first propose a hierarchical VM cache mechanism named as ATBM,and then give a proactive resource adjustment strategy named as GMAC,both of which improve QoS under cloud services when they are caught in bursty workloads.What's more,instead of using a single queue to cache requests within VMs,ATBM divides cache queues into two kinds:buffer queues and block queues.The former is set to cache requests,and the latter is applied to block requests.This multi-level cache mechanism efficiently avoids bursty workload's long-term congestions to cache queues,and ensures continuous circulation of requests as cached in queues.And on the foundation of ATBM,GMAC is designed which consists of two significant components:aprediction algorithm called as MGM for resources that are demanded to keep QoS,and a maximum available resource evaluation model named as ACM.Furthermore,MGM is based on a modified Grey Model which greatly inherits the high-robust lying in GM(Grey Model) that means having no strict limitation to the types of workloads.At the same time,MGM makes full use of residuals of GM to improve prediction results.Consequently,MGM is able to predict workload with low under-prediction rate and scale,which makes its prediction in resource also hold a small under-estimate scale and rate.On the other hand,ACM conclude its analysis in available resources by considering initial resource configurations of VMs,current requests processing conditions and SLA(Service Level Agreement) comprehensively as well as reasonably.As a result,GMAC depends on MGM to get the number of resources that demanded to ensure QoS,and utilize ACM to evaluate the available number of resource within current systems.Then,by means of comparing the quantities of demanded resources and available resources to get the number of resources calling for being adjusted,GAMC plans and carries out the resource adjustment strategy in advance through adding or removing VMs.Finally,we employ real workload data to detect the effectiveness of ATBM to enhance system QoS in the aspects of decreasing requests violation together with rejection rate,and optimizing the structure of request response that makes more requests to be responded within shorter time.And we also conduct comparison experiments to analyze the validity of GMAC in coping bursty workload.The results show that the violation rate of ATBM is only 24.6% of that from traditional single cache queue.Meanwhile,compared with resource reservation strategies based on workload peaks,GAMC increases resource utilization rate by 1.45 times without any SLA violations.And the average response time and rejection rate of the cloud service decrease by 8.6% and 16.9% with no violations,when it switches its resource adjustment strategy from reserving resources relying on the means of workloads to GMAC.
Bursty workload is a crucial factor that deteriorates QoS(Quality of Service) of cloud services.However,traditional coping strategies for bursty workloads suffer disadvantages as poor timeliness and low resource utilization rate,which not only lengthens request response time but also increases request violation and rejection rate.To solve these problems,we first propose a hierarchical VM cache mechanism named as ATBM,and then give a proactive resource adjustment strategy named as GMAC,both of which improve QoS under cloud services when they are caught in bursty workloads.What's more,instead of using a single queue to cache requests within VMs,ATBM divides cache queues into two kinds:buffer queues and block queues.The former is set to cache requests,and the latter is applied to block requests.This multi-level cache mechanism efficiently avoids bursty workload's long-term congestions to cache queues,and ensures continuous circulation of requests as cached in queues.And on the foundation of ATBM,GMAC is designed which consists of two significant components:aprediction algorithm called as MGM for resources that are demanded to keep QoS,and a maximum available resource evaluation model named as ACM.Furthermore,MGM is based on a modified Grey Model which greatly inherits the high-robust lying in GM(Grey Model) that means having no strict limitation to the types of workloads.At the same time,MGM makes full use of residuals of GM to improve prediction results.Consequently,MGM is able to predict workload with low under-prediction rate and scale,which makes its prediction in resource also hold a small under-estimate scale and rate.On the other hand,ACM conclude its analysis in available resources by considering initial resource configurations of VMs,current requests processing conditions and SLA(Service Level Agreement) comprehensively as well as reasonably.As a result,GMAC depends on MGM to get the number of resources that demanded to ensure QoS,and utilize ACM to evaluate the available number of resource within current systems.Then,by means of comparing the quantities of demanded resources and available resources to get the number of resources calling for being adjusted,GAMC plans and carries out the resource adjustment strategy in advance through adding or removing VMs.Finally,we employ real workload data to detect the effectiveness of ATBM to enhance system QoS in the aspects of decreasing requests violation together with rejection rate,and optimizing the structure of request response that makes more requests to be responded within shorter time.And we also conduct comparison experiments to analyze the validity of GMAC in coping bursty workload.The results show that the violation rate of ATBM is only 24.6% of that from traditional single cache queue.Meanwhile,compared with resource reservation strategies based on workload peaks,GAMC increases resource utilization rate by 1.45 times without any SLA violations.And the average response time and rejection rate of the cloud service decrease by 8.6% and 16.9% with no violations,when it switches its resource adjustment strategy from reserving resources relying on the means of workloads to GMAC.
引文
[1]Gong Z,Gu X,Wilkes J.PRESS:PRedictive Elastic ReSource scaling for cloud systems//Proceedings of the IEEEInternational Conference on Network and Service Management.Ontario,Canada,2011:9-16
[2]Qiu X,Dai Y,Xiang Y,Xing L.A hierarchical correlation model for evaluating reliability,performance,and power consumption of a cloud service.IEEE Transactions on Systems Man and Cybernetics Systems,2016,46(3):401-412
[3]Ali-Eldin A,Seleznjev O,Sj9stedt-de Luna S,et al.Measuring cloud workload burstiness//Proceedings of the IEEE/ACM,International Conference on Utility and Cloud Computing.London,UK,2014:566-572
[4]Mi N,Casale G,Cherkasova L,Smirni E.Burstiness in multi-tier applications:Symptoms,causes,and new models//Proceedings of the 9th ACM/IFIP/USENIX International Conference on Middleware.New York,USA,2008:265-286
[5]Mohamed M,Amziani M,Bela6d D,et al.An autonomic approach to manage elasticity of business processes in the cloud.Future Generation Computer Systems,2015,50(C):49-61
[6]Shen Z,Subbiah S,Gu X,Wilkes J.CloudScale:Elastic resource scaling for multi-tenant cloud systems//Proceedings of the ACM Symposium on Cloud Computing.New York,USA,2011:5
[7]Ali-Eldin A,Tordsson J,Elmroth E.An adaptive hybrid elasticity controller for cloud infrastructures//Proceedings of the IEEE Network Operations and Management Symposium.HI,USA,2012:204-212
[8]Hasan M Z,Magana E,Clemm A,Tucker L.Integrated and autonomic cloud resource scaling.IEEE Network Operations and Management Symposium,2012,104(5):1327-1334
[9]Calheiros R,Masoumi E,Ranjan R,Buyya R.Workload prediction using ARIMA model and its impact on cloud applications’QoS.IEEE Transactions on Cloud Computing,2015,3(4):449-458
[10]Zhang S,Qian Z,Luo Z,et al.Burstiness-aware resource reservation for server consolidation in computing clouds.IEEE Transactions on Parallel and Distributed Systems,2016,27(4):964-977
[11]Richman J S,Moorman J R.Physiological time-series analysis using approximate entropy and sample entropy.AJPHeart and Circulatory Physiology,2000,278(6):2039-2049
[12]Casale G,Mi N,Smirni E.Model-driven system capacity planning under workload burstiness.IEEE Transactions on Computers,2009,59(1):66-80
[13]Jiang J,Lu J,Zhang G,Long G.Optimal cloud resource auto-scaling for web applications//Proceedings of the IEEE/ACM International Symposium on Cluster,Cloud and Grid Computing.Delft,Netherlands,2013:58-65
[14]Ali-Eldin A,Kihl M,Tordsson J,Elmroth E.Efficient provisioning of bursty scientific workloads on the cloud using adaptive elasticity control//Proceedings of the ACM Workshop on Scientific Cloud Computing DATE.New York,USA,2012:31-40
[15]Zou B X,Liu Q.ARMA-based traffic prediction and overload detection of network.Journal of Computer Research and Development,2002,39(12):1645-1652
[16]Roy N,Dubey A,Gokhale A.Efficient autoscaling in the cloud using predictive models for workload forecasting//Proceedings of the IEEE International Conference on Cloud Computing.Washington,USA,2011:500-507
[17]Tran V G,Debusschere V,Bacha S.Hourly server workload forecasting up to 168hours ahead using seasonal ARIMAmodel//Proceedings of the IEEE International Conference on Industrial Technology.Athens,Greece,2012:1127-1131
[18]Li S,Wang Y,Qiu X,et al.A workload prediction-based multi-VM provisioning mechanism in cloud computing//Proceedings of the IFIP/IEEE,15th Asia-Pacific Network Operations and Management Symposium.Hiroshima,Japan,2013:1-6
[19]Baldan F J,Ramirezgallego S,Bergmeir C,et al.A forecasting methodology for workload forecasting in cloud systems.IEEE Transactions on Cloud Computing,2016,(99):1-1
[20]Wei G,Ling Y,Guo B,et al.Prediction-based data aggregation in wireless sensor networks:Combining grey model and Kalman filter.Computer Communications,2011,34(6):793-802
[21]Kayacan E,Ulutas B,Kaynak O.Grey system theory-based models in time series prediction.Expert Systems with Applications,2010,37(2):1784-1789
[22]Jheng J J,Tseng F H,Chao H C,Chou L D.A novel VMworkload prediction using grey forecasting model in cloud data center//Proceedings of the IEEE International Conference on Information Networking.Phuket,Thailand,2014:40-45
[23]Tseng F M,Yu H C,Tzeng G H.Applied hybrid grey model to forecast seasonal time series.Technological Forecasting and Social Change,2001,67(2-3):291-302
[24]Islam S,Keung J,Lee K,Liu A.Empirical prediction models for adaptive resource provisioning in the cloud.Future Generation Computer Systems,2012,28(1):155-162
[25]Imam M T,Miskhat S F,Rahman R M,Amin M A.Neural network and regression based processor load prediction for efficient scaling of grid and cloud resources//Proceedings of the International Conference on Computer and Information Technology.Dhaka,Bangladesh,2011:333-338
[26]Qiu F,Zhang B,Guo J.A deep learning approach for VMworkload prediction in the cloud//Proceedings of the IEEE/ACM International Conference on Software Engineering,Artificial Intelligence,Networking and Parallel/distributed Computing.Shanghai,China,2016:319-324
[27]Messias V R,Estrella J C,Ehlers R,et al.Combining time series prediction models using genetic algorithm to autoscaling Web applications hosted in the cloud infrastructure.Neural Computing and Applications,2016,27(8):2383-2406
[28]Ghelardoni L,Ghio A,Anguita D.Energy load forecasting using empirical mode decomposition and support vector regression.IEEE Transactions on Smart Grid,2013,4(1):549-556
[29]Bankole A A,Ajila S A.Cloud client prediction models for cloud resource provisioning in a multitier web application environment//Proceedings of the IEEE International Symposium on Service Oriented System Engineering.Redwood City,USA,2013:156-161
[30]Grechanik M,Luo Q,Poshyvanyk D,Portere A.Enhancing rules for cloud resource provisioning via learned software performance models//Proceedings of the 7th ACM/SPEC on International Conference on Performance Engineering.New York,USA,2016:209-214
[31]Yang J,Yu T,Jian L R,et al.An extreme automation framework for scaling cloud applications.IBM Journal of Research and Development,2011,55(6):410-421
[32]Sun J,Chen H,Yin Z.AERS:An autonomic and elastic resource scheduling framework for cloud applications//Proceedings of the IEEE International Conference on Services Computing.San Francisco,USA,2016:66-73
[33]Coutinho E F,Gomes D G,Souza J N D.An autonomic computing-based architecture for cloud computing elasticity//Proceedings of the Network Operations and Management Symposium.Joao Pessoa,Brazil,2015:111-112
[34]Wager S,Hastie T,Efron B.Confidence intervals for random forests:The jackknife and the infinitesimal jackknife.Journal of Machine Learning Research,2014,15(1):1625-1651
[35]Olinsky A,Kristin K,Brayton K B.Assessing gradient boosting in the reduction of misclassification error in the prediction of success for actuarial majors.Csbigs Cases in Business Industry Government&Government Statistics,2014,5(1):12-16
[36]Ali-Eldin A,Rezaie A,Mehta A,et al.How will your workload look like in 6Years?analyzing Wikimedia’s workload//Proceedings of the IEEE International Conference on Cloud Engineering.Boston,USA,2014:349-354
[37]Sladescu M,Fekete A,Lee K,Liu.A.A polymorphic model for event associated workload bursts//Proceedings of the IEEE Institute Conference on Distributed Computing Systems Workshops.Philadelphia,USA,2013:119-125
[38]Huang C J,Guan C T,Chen H M,et al.An adaptive resource management scheme in cloud computing.Engineering Applications of Artificial Intelligence,2013,26(1):382-389
[39]Yin J,Lu X,Chen H,et al.System resource utilization analysis and prediction for cloud based applications under bursty workloads.Information Sciences,2014,279:338-357
[40]Yin J,Lu X,Zhao X,et al.BURSE:A bursty and self-similar workload generator for cloud computing.IEEETransactions on Parallel and Distributed Systems,2015,26(3):668-680
[41]Gusella R.Characterizing the variability of arrival processes with indexes of dispersion.IEEE Journal on Selected Areas in Communications,2006,9(2):203-211
[42]Casale G,Mi N,Smirni E.Model-driven system capacity planning under workload burstiness.IEEE Transactions on Computers,2009,59(1):66-80
[43]Meng Yu,Zhang Bin,Guo Jun.Interval forecasting model of concurrent volume of cloud service users in cloud computing environment.Chinese Journal of Computers,2017,40(2):378-396(in Chinese)(孟煜,张斌,郭军.云计算环境下云服务用户并发量的区间预测模型.计算机学报,2017,40(2):378-396)
[44]Tai J,Zhang J,Li J,et al.ArA:Adaptive resource allocation for cloud computing environments under bursty workloads//Proceedings of the IEEE International Performance Computing and Communications Conference.Orlando,USA,2011:1-8
(1)http://stats.grok.se/en/200906/Michael%20Jackson
(2)请求违例率指请求响应时间超过SLA规定值的用户并发量数占总用户并发量数的比例.
(3)请求违例率指由于VM无可用缓存空间而被拒绝的用户并发量数占总用户并发量数的比例.
(1)Amazon Elastic Computer Cloud.http://aws.amazon.com/ec2/
(1)http://www.cloudbus.org/cloudsim/
(2)https://github.com/JaneWuNEU/cloudsim
(1)基于预测窗口内的数据对下一时刻点的并发量进行预测,例如当MGM的预测窗口设置为10时,算法会基于{ti|i=0,…,9}内数据预测t10时刻的并发量.
(2)2015年第49天为2月19日-春节,“淘宝”于该天停止发货,故“淘宝”对应的百度指数陡降.
[2]Qiu X,Dai Y,Xiang Y,Xing L.A hierarchical correlation model for evaluating reliability,performance,and power consumption of a cloud service.IEEE Transactions on Systems Man and Cybernetics Systems,2016,46(3):401-412
[3]Ali-Eldin A,Seleznjev O,Sj9stedt-de Luna S,et al.Measuring cloud workload burstiness//Proceedings of the IEEE/ACM,International Conference on Utility and Cloud Computing.London,UK,2014:566-572
[4]Mi N,Casale G,Cherkasova L,Smirni E.Burstiness in multi-tier applications:Symptoms,causes,and new models//Proceedings of the 9th ACM/IFIP/USENIX International Conference on Middleware.New York,USA,2008:265-286
[5]Mohamed M,Amziani M,Bela6d D,et al.An autonomic approach to manage elasticity of business processes in the cloud.Future Generation Computer Systems,2015,50(C):49-61
[6]Shen Z,Subbiah S,Gu X,Wilkes J.CloudScale:Elastic resource scaling for multi-tenant cloud systems//Proceedings of the ACM Symposium on Cloud Computing.New York,USA,2011:5
[7]Ali-Eldin A,Tordsson J,Elmroth E.An adaptive hybrid elasticity controller for cloud infrastructures//Proceedings of the IEEE Network Operations and Management Symposium.HI,USA,2012:204-212
[8]Hasan M Z,Magana E,Clemm A,Tucker L.Integrated and autonomic cloud resource scaling.IEEE Network Operations and Management Symposium,2012,104(5):1327-1334
[9]Calheiros R,Masoumi E,Ranjan R,Buyya R.Workload prediction using ARIMA model and its impact on cloud applications’QoS.IEEE Transactions on Cloud Computing,2015,3(4):449-458
[10]Zhang S,Qian Z,Luo Z,et al.Burstiness-aware resource reservation for server consolidation in computing clouds.IEEE Transactions on Parallel and Distributed Systems,2016,27(4):964-977
[11]Richman J S,Moorman J R.Physiological time-series analysis using approximate entropy and sample entropy.AJPHeart and Circulatory Physiology,2000,278(6):2039-2049
[12]Casale G,Mi N,Smirni E.Model-driven system capacity planning under workload burstiness.IEEE Transactions on Computers,2009,59(1):66-80
[13]Jiang J,Lu J,Zhang G,Long G.Optimal cloud resource auto-scaling for web applications//Proceedings of the IEEE/ACM International Symposium on Cluster,Cloud and Grid Computing.Delft,Netherlands,2013:58-65
[14]Ali-Eldin A,Kihl M,Tordsson J,Elmroth E.Efficient provisioning of bursty scientific workloads on the cloud using adaptive elasticity control//Proceedings of the ACM Workshop on Scientific Cloud Computing DATE.New York,USA,2012:31-40
[15]Zou B X,Liu Q.ARMA-based traffic prediction and overload detection of network.Journal of Computer Research and Development,2002,39(12):1645-1652
[16]Roy N,Dubey A,Gokhale A.Efficient autoscaling in the cloud using predictive models for workload forecasting//Proceedings of the IEEE International Conference on Cloud Computing.Washington,USA,2011:500-507
[17]Tran V G,Debusschere V,Bacha S.Hourly server workload forecasting up to 168hours ahead using seasonal ARIMAmodel//Proceedings of the IEEE International Conference on Industrial Technology.Athens,Greece,2012:1127-1131
[18]Li S,Wang Y,Qiu X,et al.A workload prediction-based multi-VM provisioning mechanism in cloud computing//Proceedings of the IFIP/IEEE,15th Asia-Pacific Network Operations and Management Symposium.Hiroshima,Japan,2013:1-6
[19]Baldan F J,Ramirezgallego S,Bergmeir C,et al.A forecasting methodology for workload forecasting in cloud systems.IEEE Transactions on Cloud Computing,2016,(99):1-1
[20]Wei G,Ling Y,Guo B,et al.Prediction-based data aggregation in wireless sensor networks:Combining grey model and Kalman filter.Computer Communications,2011,34(6):793-802
[21]Kayacan E,Ulutas B,Kaynak O.Grey system theory-based models in time series prediction.Expert Systems with Applications,2010,37(2):1784-1789
[22]Jheng J J,Tseng F H,Chao H C,Chou L D.A novel VMworkload prediction using grey forecasting model in cloud data center//Proceedings of the IEEE International Conference on Information Networking.Phuket,Thailand,2014:40-45
[23]Tseng F M,Yu H C,Tzeng G H.Applied hybrid grey model to forecast seasonal time series.Technological Forecasting and Social Change,2001,67(2-3):291-302
[24]Islam S,Keung J,Lee K,Liu A.Empirical prediction models for adaptive resource provisioning in the cloud.Future Generation Computer Systems,2012,28(1):155-162
[25]Imam M T,Miskhat S F,Rahman R M,Amin M A.Neural network and regression based processor load prediction for efficient scaling of grid and cloud resources//Proceedings of the International Conference on Computer and Information Technology.Dhaka,Bangladesh,2011:333-338
[26]Qiu F,Zhang B,Guo J.A deep learning approach for VMworkload prediction in the cloud//Proceedings of the IEEE/ACM International Conference on Software Engineering,Artificial Intelligence,Networking and Parallel/distributed Computing.Shanghai,China,2016:319-324
[27]Messias V R,Estrella J C,Ehlers R,et al.Combining time series prediction models using genetic algorithm to autoscaling Web applications hosted in the cloud infrastructure.Neural Computing and Applications,2016,27(8):2383-2406
[28]Ghelardoni L,Ghio A,Anguita D.Energy load forecasting using empirical mode decomposition and support vector regression.IEEE Transactions on Smart Grid,2013,4(1):549-556
[29]Bankole A A,Ajila S A.Cloud client prediction models for cloud resource provisioning in a multitier web application environment//Proceedings of the IEEE International Symposium on Service Oriented System Engineering.Redwood City,USA,2013:156-161
[30]Grechanik M,Luo Q,Poshyvanyk D,Portere A.Enhancing rules for cloud resource provisioning via learned software performance models//Proceedings of the 7th ACM/SPEC on International Conference on Performance Engineering.New York,USA,2016:209-214
[31]Yang J,Yu T,Jian L R,et al.An extreme automation framework for scaling cloud applications.IBM Journal of Research and Development,2011,55(6):410-421
[32]Sun J,Chen H,Yin Z.AERS:An autonomic and elastic resource scheduling framework for cloud applications//Proceedings of the IEEE International Conference on Services Computing.San Francisco,USA,2016:66-73
[33]Coutinho E F,Gomes D G,Souza J N D.An autonomic computing-based architecture for cloud computing elasticity//Proceedings of the Network Operations and Management Symposium.Joao Pessoa,Brazil,2015:111-112
[34]Wager S,Hastie T,Efron B.Confidence intervals for random forests:The jackknife and the infinitesimal jackknife.Journal of Machine Learning Research,2014,15(1):1625-1651
[35]Olinsky A,Kristin K,Brayton K B.Assessing gradient boosting in the reduction of misclassification error in the prediction of success for actuarial majors.Csbigs Cases in Business Industry Government&Government Statistics,2014,5(1):12-16
[36]Ali-Eldin A,Rezaie A,Mehta A,et al.How will your workload look like in 6Years?analyzing Wikimedia’s workload//Proceedings of the IEEE International Conference on Cloud Engineering.Boston,USA,2014:349-354
[37]Sladescu M,Fekete A,Lee K,Liu.A.A polymorphic model for event associated workload bursts//Proceedings of the IEEE Institute Conference on Distributed Computing Systems Workshops.Philadelphia,USA,2013:119-125
[38]Huang C J,Guan C T,Chen H M,et al.An adaptive resource management scheme in cloud computing.Engineering Applications of Artificial Intelligence,2013,26(1):382-389
[39]Yin J,Lu X,Chen H,et al.System resource utilization analysis and prediction for cloud based applications under bursty workloads.Information Sciences,2014,279:338-357
[40]Yin J,Lu X,Zhao X,et al.BURSE:A bursty and self-similar workload generator for cloud computing.IEEETransactions on Parallel and Distributed Systems,2015,26(3):668-680
[41]Gusella R.Characterizing the variability of arrival processes with indexes of dispersion.IEEE Journal on Selected Areas in Communications,2006,9(2):203-211
[42]Casale G,Mi N,Smirni E.Model-driven system capacity planning under workload burstiness.IEEE Transactions on Computers,2009,59(1):66-80
[43]Meng Yu,Zhang Bin,Guo Jun.Interval forecasting model of concurrent volume of cloud service users in cloud computing environment.Chinese Journal of Computers,2017,40(2):378-396(in Chinese)(孟煜,张斌,郭军.云计算环境下云服务用户并发量的区间预测模型.计算机学报,2017,40(2):378-396)
[44]Tai J,Zhang J,Li J,et al.ArA:Adaptive resource allocation for cloud computing environments under bursty workloads//Proceedings of the IEEE International Performance Computing and Communications Conference.Orlando,USA,2011:1-8
(1)http://stats.grok.se/en/200906/Michael%20Jackson
(2)请求违例率指请求响应时间超过SLA规定值的用户并发量数占总用户并发量数的比例.
(3)请求违例率指由于VM无可用缓存空间而被拒绝的用户并发量数占总用户并发量数的比例.
(1)Amazon Elastic Computer Cloud.http://aws.amazon.com/ec2/
(1)http://www.cloudbus.org/cloudsim/
(2)https://github.com/JaneWuNEU/cloudsim
(1)基于预测窗口内的数据对下一时刻点的并发量进行预测,例如当MGM的预测窗口设置为10时,算法会基于{ti|i=0,…,9}内数据预测t10时刻的并发量.
(2)2015年第49天为2月19日-春节,“淘宝”于该天停止发货,故“淘宝”对应的百度指数陡降.
© 2004-2018 中国地质图书馆版权所有 京ICP备05064691号 京公网安备11010802017129号 地址：北京市海淀区学院路29号 邮编：100083 电话：办公室：(+86 10)66554848；文献借阅、咨询服务、科技查新：66554700 |