As dependence on Content Delivery Networks (CDNs) increases, there is a growing need for innovative solutions to optimize cache performance amid increasing traffic and complicated cache-sharing workloads. Allocating exclusive resources to applications in CDNs boosts the overall cache hit ratio (OHR), enhancing efficiency. However, the traditional method of creating the miss ratio curve (MRC) is unsuitable for CDNs due to the diverse sizes of items and the vast number of applications, leading to high computational overhead and performance inconsistency. To tackle this issue, we propose a l ightweight and a daptive cache p artitioning scheme called $varepsilon$ -LAP. This scheme uses a corresponding shadow cache for each partition and sorts them based on the average hit numbers on the granularity unit in the shadow caches. During partition resizing, $varepsilon$ -LAP transfers storage capacity, measured in units of granularity, from the $(N-k+1)$ -th ( $kleq frac{N}{2}$ ) partition to the $k$ -th partition. A learning threshold parameter, i.e., $varepsilon$ , is also introduced to prudently determine when to resize partitions, improving caching efficiency. This can eliminate about 96.8% of unnecessary partition resizing without compromising performance. $varepsilon$ -LAP, when deployed in PicCloud at Tencent , improved OHR by 9.34% and reduced the average user access latency by 12.5 ms. Experimental results show that $varepsilon$ -LAP outperforms other cache partitioning schemes in terms of both OHR and access latency, and it effectively adapts to workload variations.
{"title":"$varepsilon$ɛ-LAP: A Lightweight and Adaptive Cache Partitioning Scheme With Prudent Resizing Decisions for Content Delivery Networks","authors":"Peng Wang;Yu Liu;Ziqi Liu;Zhelong Zhao;Ke Liu;Ke Zhou;Zhihai Huang","doi":"10.1109/TCC.2024.3420454","DOIUrl":"10.1109/TCC.2024.3420454","url":null,"abstract":"As dependence on Content Delivery Networks (CDNs) increases, there is a growing need for innovative solutions to optimize cache performance amid increasing traffic and complicated cache-sharing workloads. Allocating exclusive resources to applications in CDNs boosts the overall cache hit ratio (OHR), enhancing efficiency. However, the traditional method of creating the miss ratio curve (MRC) is unsuitable for CDNs due to the diverse sizes of items and the vast number of applications, leading to high computational overhead and performance inconsistency. To tackle this issue, we propose a \u0000<u>l</u>\u0000ightweight and \u0000<u>a</u>\u0000daptive cache \u0000<u>p</u>\u0000artitioning scheme called \u0000<inline-formula><tex-math>$varepsilon$</tex-math></inline-formula>\u0000-LAP. This scheme uses a corresponding shadow cache for each partition and sorts them based on the average hit numbers on the granularity unit in the shadow caches. During partition resizing, \u0000<inline-formula><tex-math>$varepsilon$</tex-math></inline-formula>\u0000-LAP transfers storage capacity, measured in units of granularity, from the \u0000<inline-formula><tex-math>$(N-k+1)$</tex-math></inline-formula>\u0000-th (\u0000<inline-formula><tex-math>$kleq frac{N}{2}$</tex-math></inline-formula>\u0000) partition to the \u0000<inline-formula><tex-math>$k$</tex-math></inline-formula>\u0000-th partition. A learning threshold parameter, i.e., \u0000<inline-formula><tex-math>$varepsilon$</tex-math></inline-formula>\u0000, is also introduced to prudently determine when to resize partitions, improving caching efficiency. This can eliminate about 96.8% of unnecessary partition resizing without compromising performance. \u0000<inline-formula><tex-math>$varepsilon$</tex-math></inline-formula>\u0000-LAP, when deployed in \u0000<i>PicCloud</i>\u0000 at \u0000<i>Tencent</i>\u0000, improved OHR by 9.34% and reduced the average user access latency by 12.5 ms. Experimental results show that \u0000<inline-formula><tex-math>$varepsilon$</tex-math></inline-formula>\u0000-LAP outperforms other cache partitioning schemes in terms of both OHR and access latency, and it effectively adapts to workload variations.","PeriodicalId":13202,"journal":{"name":"IEEE Transactions on Cloud Computing","volume":null,"pages":null},"PeriodicalIF":5.3,"publicationDate":"2024-06-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141503565","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2024-06-18DOI: 10.1109/tcc.2024.3415165
Houming Qiu, Kun Zhu, Dusit Niyato
{"title":"Secure and Flexible Coded Distributed Matrix Multiplication Based on Edge Computing for Industrial Metaverse","authors":"Houming Qiu, Kun Zhu, Dusit Niyato","doi":"10.1109/tcc.2024.3415165","DOIUrl":"https://doi.org/10.1109/tcc.2024.3415165","url":null,"abstract":"","PeriodicalId":13202,"journal":{"name":"IEEE Transactions on Cloud Computing","volume":null,"pages":null},"PeriodicalIF":6.5,"publicationDate":"2024-06-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141937583","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2024-06-14DOI: 10.1109/tcc.2024.3414440
Fahao Chen, Peng Li, Celimuge Wu, Song Guo
{"title":"Non-Clairvoyant Scheduling of Distributed Machine Learning with Inter-job and Intra-job Parallelism on Heterogeneous GPUs","authors":"Fahao Chen, Peng Li, Celimuge Wu, Song Guo","doi":"10.1109/tcc.2024.3414440","DOIUrl":"https://doi.org/10.1109/tcc.2024.3414440","url":null,"abstract":"","PeriodicalId":13202,"journal":{"name":"IEEE Transactions on Cloud Computing","volume":null,"pages":null},"PeriodicalIF":6.5,"publicationDate":"2024-06-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141937581","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2024-06-11DOI: 10.1109/tcc.2024.3410390
Zuodong Jin, Dan Tao, Peng Qi, Ruipeng Gao
{"title":"An Adaptive Cloud Resource Quota Scheme Based on Dynamic Portraits and Task-Resource Matching","authors":"Zuodong Jin, Dan Tao, Peng Qi, Ruipeng Gao","doi":"10.1109/tcc.2024.3410390","DOIUrl":"https://doi.org/10.1109/tcc.2024.3410390","url":null,"abstract":"","PeriodicalId":13202,"journal":{"name":"IEEE Transactions on Cloud Computing","volume":null,"pages":null},"PeriodicalIF":6.5,"publicationDate":"2024-06-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141937582","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2024-06-06DOI: 10.1109/tcc.2024.3410377
Ting Yang, Yuxing Hou, Shaotang Cai, Jie Yu, Haibo Pen
{"title":"Multi-Data Center Tie-line Power Smoothing Method Based on Demand Response","authors":"Ting Yang, Yuxing Hou, Shaotang Cai, Jie Yu, Haibo Pen","doi":"10.1109/tcc.2024.3410377","DOIUrl":"https://doi.org/10.1109/tcc.2024.3410377","url":null,"abstract":"","PeriodicalId":13202,"journal":{"name":"IEEE Transactions on Cloud Computing","volume":null,"pages":null},"PeriodicalIF":6.5,"publicationDate":"2024-06-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141937598","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2024-06-04DOI: 10.1109/TCC.2024.3408905
Sabrina De Capitani di Vimercati;Dario Facchinetti;Sara Foresti;Gianluca Oldani;Stefano Paraboschi;Matthew Rossi;Pierangela Samarati
We address the problem of indexing encrypted data outsourced to an external cloud server to support server-side execution of multi-attribute queries. Our approach partitions the dataset in groups with the same number of tuples, and associates all tuples in a group with the same combination of index values, so to guarantee protection against static inferences. Our indexing approach does not require any modifications to the server-side software stack, and requires limited storage at the client for query support. The experimental evaluation considers, for the storage of the encrypted and indexed dataset, both a relational database (PostgreSQL) and a key-value database (Redis). We carried out extensive experiments evaluating client-storage requirements and query performance. The experimental results confirm the efficiency of our solution. The proposal is supported by an open source implementation.
{"title":"Multi-Dimensional Flat Indexing for Encrypted Data","authors":"Sabrina De Capitani di Vimercati;Dario Facchinetti;Sara Foresti;Gianluca Oldani;Stefano Paraboschi;Matthew Rossi;Pierangela Samarati","doi":"10.1109/TCC.2024.3408905","DOIUrl":"10.1109/TCC.2024.3408905","url":null,"abstract":"We address the problem of indexing encrypted data outsourced to an external cloud server to support server-side execution of multi-attribute queries. Our approach partitions the dataset in groups with the same number of tuples, and associates all tuples in a group with the same combination of index values, so to guarantee protection against static inferences. Our indexing approach does not require any modifications to the server-side software stack, and requires limited storage at the client for query support. The experimental evaluation considers, for the storage of the encrypted and indexed dataset, both a relational database (PostgreSQL) and a key-value database (Redis). We carried out extensive experiments evaluating client-storage requirements and query performance. The experimental results confirm the efficiency of our solution. The proposal is supported by an open source implementation.","PeriodicalId":13202,"journal":{"name":"IEEE Transactions on Cloud Computing","volume":null,"pages":null},"PeriodicalIF":5.3,"publicationDate":"2024-06-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=10547318","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141937577","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Cloud-assisted computation empowers resource-constrained clients to efficiently tackle computationally intensive tasks by outsourcing them to resource-rich cloud servers. In the current era of Big Data, the widespread need to solve large-scale modular linear systems of equations ( $mathcal {LMLSE}$ ) of the form $mathbf {A}mathbf {x}equiv mathbf {b};{rm mod};{q}$ poses a significant challenge, particularly for lightweight devices. This paper delves into the secure outsourcing of $mathcal {LMLSE}$ under a malicious single-server model and, to the best of our knowledge, introduces the inaugural protocol tailored to this specific context. The cornerstone of our protocol lies in the innovation of a novel matrix encryption method based on sparse unimodular matrix transformations. This novel technique bestows our protocol with several key advantages. First and foremost, it ensures robust privacy for all computation inputs, encompassing $mathbf {A},mathbf {b}, q$ , and the output $mathbf {x}$ , as validated by thorough theoretical analysis. Second, the protocol delivers optimal verifiability, enabling clients to detect cloud server misbehavior with an unparalleled probability of 1. Furthermore, it boasts high efficiency, requiring only a single interaction between the client and the cloud server, significantly reducing local-client time costs. For an $m$ -by- $n$ matrix $mathbf {A}$ , a given parameter $lambda =omega (log q)$ , and $rho =2.371552$ , the time complexity is diminished from $O(max lbrace m n^{rho -1}, m^{rho -2} n^{2}rbrace cdot (log q)^{2})$ to $O((mn+m^{2})lambda log q+mn(log q)^{2})$ . The comprehensive results of our experimental performance evaluations substantiate the protocol's practical efficiency and effectiveness.
{"title":"How to Securely and Efficiently Solve the Large-Scale Modular System of Linear Equations on the Cloud","authors":"Chengliang Tian;Jia Yu;Panpan Meng;Guoyan Zhang;Weizhong Tian;Yan Zhang","doi":"10.1109/TCC.2024.3408240","DOIUrl":"10.1109/TCC.2024.3408240","url":null,"abstract":"Cloud-assisted computation empowers resource-constrained clients to efficiently tackle computationally intensive tasks by outsourcing them to resource-rich cloud servers. In the current era of Big Data, the widespread need to solve large-scale modular linear systems of equations (\u0000<inline-formula><tex-math>$mathcal {LMLSE}$</tex-math></inline-formula>\u0000) of the form \u0000<inline-formula><tex-math>$mathbf {A}mathbf {x}equiv mathbf {b};{rm mod};{q}$</tex-math></inline-formula>\u0000 poses a significant challenge, particularly for lightweight devices. This paper delves into the secure outsourcing of \u0000<inline-formula><tex-math>$mathcal {LMLSE}$</tex-math></inline-formula>\u0000 under a malicious single-server model and, to the best of our knowledge, introduces the inaugural protocol tailored to this specific context. The cornerstone of our protocol lies in the innovation of a novel matrix encryption method based on sparse unimodular matrix transformations. This novel technique bestows our protocol with several key advantages. First and foremost, it ensures robust privacy for all computation inputs, encompassing \u0000<inline-formula><tex-math>$mathbf {A},mathbf {b}, q$</tex-math></inline-formula>\u0000, and the output \u0000<inline-formula><tex-math>$mathbf {x}$</tex-math></inline-formula>\u0000, as validated by thorough theoretical analysis. Second, the protocol delivers optimal verifiability, enabling clients to detect cloud server misbehavior with an unparalleled probability of 1. Furthermore, it boasts high efficiency, requiring only a single interaction between the client and the cloud server, significantly reducing local-client time costs. For an \u0000<inline-formula><tex-math>$m$</tex-math></inline-formula>\u0000-by-\u0000<inline-formula><tex-math>$n$</tex-math></inline-formula>\u0000 matrix \u0000<inline-formula><tex-math>$mathbf {A}$</tex-math></inline-formula>\u0000, a given parameter \u0000<inline-formula><tex-math>$lambda =omega (log q)$</tex-math></inline-formula>\u0000, and \u0000<inline-formula><tex-math>$rho =2.371552$</tex-math></inline-formula>\u0000, the time complexity is diminished from \u0000<inline-formula><tex-math>$O(max lbrace m n^{rho -1}, m^{rho -2} n^{2}rbrace cdot (log q)^{2})$</tex-math></inline-formula>\u0000 to \u0000<inline-formula><tex-math>$O((mn+m^{2})lambda log q+mn(log q)^{2})$</tex-math></inline-formula>\u0000. The comprehensive results of our experimental performance evaluations substantiate the protocol's practical efficiency and effectiveness.","PeriodicalId":13202,"journal":{"name":"IEEE Transactions on Cloud Computing","volume":null,"pages":null},"PeriodicalIF":5.3,"publicationDate":"2024-06-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141937601","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Public goods projects, such as open-source technology, are essential for the blockchain ecosystem's growth. However, funding these projects effectively remains a critical issue within the ecosystem. Currently, the funding protocols for blockchain public goods lack professionalism and fail to learn from past experiences. To address this challenge, our research introduces a human oracle protocol involving public goods projects, experts, and funders. In our approach, funders contribute investments to a funding pool, while experts offer investment advice based on their expertise in public goods projects. The oracle's decisions on funding support are influenced by the reputations of the experts. Experts earn or lose reputation based on how well their project implementations align with their advice, with successful investments leading to higher reputations. Our oracle is designed to adapt to changing circumstances, such as experts exiting or entering the decision-making process. We also introduce a regret bound to gauge the oracle's effectiveness. Theoretically, we establish an upper regret bound for both static and dynamic models and demonstrate its closeness to an asymptotically equal lower bound. Empirically, we implement our protocol on a test chain and show that our oracle's investment decisions closely mirror optimal investments in hindsight.
{"title":"Decentralized Funding of Public Goods in Blockchain System: Leveraging Expert Advice","authors":"Jichen Li;Yukun Cheng;Wenhan Huang;Mengqian Zhang;Jiarui Fan;Xiaotie Deng;Jan Xie;Jie Zhang","doi":"10.1109/TCC.2024.3394973","DOIUrl":"10.1109/TCC.2024.3394973","url":null,"abstract":"Public goods projects, such as open-source technology, are essential for the blockchain ecosystem's growth. However, funding these projects effectively remains a critical issue within the ecosystem. Currently, the funding protocols for blockchain public goods lack professionalism and fail to learn from past experiences. To address this challenge, our research introduces a human oracle protocol involving public goods projects, experts, and funders. In our approach, funders contribute investments to a funding pool, while experts offer investment advice based on their expertise in public goods projects. The oracle's decisions on funding support are influenced by the reputations of the experts. Experts earn or lose reputation based on how well their project implementations align with their advice, with successful investments leading to higher reputations. Our oracle is designed to adapt to changing circumstances, such as experts exiting or entering the decision-making process. We also introduce a regret bound to gauge the oracle's effectiveness. Theoretically, we establish an upper regret bound for both static and dynamic models and demonstrate its closeness to an asymptotically equal lower bound. Empirically, we implement our protocol on a test chain and show that our oracle's investment decisions closely mirror optimal investments in hindsight.","PeriodicalId":13202,"journal":{"name":"IEEE Transactions on Cloud Computing","volume":null,"pages":null},"PeriodicalIF":6.5,"publicationDate":"2024-04-30","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"140836678","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2024-04-26DOI: 10.1109/TCC.2024.3393895
Dumitrel Loghin
Almost all major cloud providers offer virtual machines running on servers with 64-bit ARM CPUs. For example, Amazon Web Services (AWS) designed custom ARM-based CPUs named Graviton2 and Graviton3. Other cloud providers, such as Microsoft Azure and Google Cloud Platform (GCP), employ servers with Ampere Altra CPUs. In this context, we conduct a comprehensive experimental study covering in-memory key-value stores, relational databases, enterprise blockchains, and Machine Learning inference. We cover all the available types of ARM cloud processors, including Graviton2 (AWS), Graviton3 (AWS), Ampere Altra (Azure and GCP), Yitian 710 (Alibaba Cloud), and Kunpeng 920 (Huawei Cloud). Our analysis shows that Yitian and Graviton3 are serious competitors for servers with Intel Xeon CPUs, achieving similar or better results with in-memory workloads. However, the performance of OLAP, ML inference, and blockchain on ARM-based servers is below that of Xeon. The reasons are mainly threefold 1) un-optimized software, 2) lower clock frequency, and 3) lower performance at core level. Surprisingly, ARM servers spend 2X more time in Linux kernel system calls compared to Xeon servers. Nonetheless, ARM-based servers show great potential. Given their lower cloud computing price, ARM servers could be the ideal choice when the performance is not critical.
几乎所有主要的云计算提供商都提供在配备 64 位 ARM CPU 的服务器上运行的虚拟机。例如，亚马逊网络服务（AWS）设计了基于 ARM 的定制 CPU，命名为 Graviton2 和 Graviton3。其他云提供商，如微软 Azure 和谷歌云平台（GCP），则采用了配备 Ampere Altra CPU 的服务器。在此背景下，我们进行了一项全面的实验研究，涵盖内存键值存储、关系数据库、企业区块链和机器学习推理。我们涵盖了所有可用的 ARM 云处理器类型，包括 Graviton2（AWS）、Graviton3（AWS）、Ampere Altra（Azure 和 GCP）、倚天 710（阿里巴巴云）和鲲鹏 920（华为云）。我们的分析表明，倚天和 Graviton3 是英特尔至强 CPU 服务器的有力竞争者，在内存工作负载方面取得了相似或更好的结果。然而，在基于 ARM 的服务器上，OLAP、ML 推理和区块链的性能却低于至强。原因主要有三个方面：1）软件未优化；2）时钟频率较低；3）内核级性能较低。令人惊讶的是，与 Xeon 服务器相比，ARM 服务器在 Linux 内核系统调用上花费的时间多出 2 倍。不过，基于 ARM 的服务器显示出巨大的潜力。鉴于其较低的云计算价格，ARM 服务器可能是性能要求不高时的理想选择。
{"title":"Are ARM Cloud Servers Ready for Database Workloads? an Experimental Study","authors":"Dumitrel Loghin","doi":"10.1109/TCC.2024.3393895","DOIUrl":"10.1109/TCC.2024.3393895","url":null,"abstract":"Almost all major cloud providers offer virtual machines running on servers with 64-bit ARM CPUs. For example, Amazon Web Services (AWS) designed custom ARM-based CPUs named Graviton2 and Graviton3. Other cloud providers, such as Microsoft Azure and Google Cloud Platform (GCP), employ servers with Ampere Altra CPUs. In this context, we conduct a comprehensive experimental study covering in-memory key-value stores, relational databases, enterprise blockchains, and Machine Learning inference. We cover all the available types of ARM cloud processors, including Graviton2 (AWS), Graviton3 (AWS), Ampere Altra (Azure and GCP), Yitian 710 (Alibaba Cloud), and Kunpeng 920 (Huawei Cloud). Our analysis shows that Yitian and Graviton3 are serious competitors for servers with Intel Xeon CPUs, achieving similar or better results with in-memory workloads. However, the performance of OLAP, ML inference, and blockchain on ARM-based servers is below that of Xeon. The reasons are mainly threefold 1) un-optimized software, 2) lower clock frequency, and 3) lower performance at core level. Surprisingly, ARM servers spend 2X more time in Linux kernel system calls compared to Xeon servers. Nonetheless, ARM-based servers show great potential. Given their lower cloud computing price, ARM servers could be the ideal choice when the performance is not critical.","PeriodicalId":13202,"journal":{"name":"IEEE Transactions on Cloud Computing","volume":null,"pages":null},"PeriodicalIF":5.3,"publicationDate":"2024-04-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"140799413","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Workload characterization and subsequent prediction are significant steps in maintaining the elasticity and scalability of resources in Cloud Data Centers. Due to the high variance in cloud workloads, designing a prediction algorithm that models the variations in the workload is a non-trivial task. If the workload predictor is unable to handle the dynamism in the workloads, then the result of the predictor may lead to over-provisioning or under-provisioning of cloud resources. To address this problem, we have created a Super Markov Prediction Model (SMPM) whose behaviour changes as per the change in the workload patterns. As the time progresses, based on the workload pattern SMPM uses different sequence models to predict the future workload. To evaluate the proposed model, we have experimented with Alibaba trace 2018, Google Cluster Trace (GCT), Alibaba trace 2020 and TPC-W workload trace. We have compared SMPM's prediction results with existing state-of-the-art prediction models and empirically verified that the proposed prediction model achieves a better accuracy as quantified using Root Mean Square Error (RMSE) and Mean Absolute Error (MAE).
{"title":"Design and Evaluation of a Hierarchical Characterization and Adaptive Prediction Model for Cloud Workloads","authors":"Karthick Seshadri;Korrapati Sindhu;S. Nagesh Bhattu;Chidambaran Kollengode","doi":"10.1109/TCC.2024.3393114","DOIUrl":"10.1109/TCC.2024.3393114","url":null,"abstract":"Workload characterization and subsequent prediction are significant steps in maintaining the elasticity and scalability of resources in Cloud Data Centers. Due to the high variance in cloud workloads, designing a prediction algorithm that models the variations in the workload is a non-trivial task. If the workload predictor is unable to handle the dynamism in the workloads, then the result of the predictor may lead to over-provisioning or under-provisioning of cloud resources. To address this problem, we have created a Super Markov Prediction Model (SMPM) whose behaviour changes as per the change in the workload patterns. As the time progresses, based on the workload pattern SMPM uses different sequence models to predict the future workload. To evaluate the proposed model, we have experimented with Alibaba trace 2018, Google Cluster Trace (GCT), Alibaba trace 2020 and TPC-W workload trace. We have compared SMPM's prediction results with existing state-of-the-art prediction models and empirically verified that the proposed prediction model achieves a better accuracy as quantified using Root Mean Square Error (RMSE) and Mean Absolute Error (MAE).","PeriodicalId":13202,"journal":{"name":"IEEE Transactions on Cloud Computing","volume":null,"pages":null},"PeriodicalIF":6.5,"publicationDate":"2024-04-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"140806316","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：