247 Hits in 6.8 sec

Sequence-to-sequence models for workload interference prediction on batch processing datacenters

David Buchaca, Joan Marcual, Josep LLuis Berral, David Carrera
2020 Future generations computer systems  
In this work we propose a methodology for modeling co-scheduling of jobs on data-centers, based on their behavior towards resources and execution time, using sequence-to-sequence models based on recurrent  ...  Efficient job placement on environments where resources are shared requires awareness on how jobs interfere during execution, to go far beyond ineffective resource overbooking techniques.  ...  In this work we present ResourceNet, a workload-toworkload forecasting methodology to predict the effects of application co-location interference, using sequence-to-sequence models based on Recurrent Neural  ... 
doi:10.1016/j.future.2020.03.058 fatcat:vw33tgjwdjfahfxfq7crf5fpqe


Mohan Raj Velayudhan Kumar, Shriram Raghunathan
2014 Journal of Computer Science  
Interference or contention on the limited shared resources among VMs leads to performance degradation and is referred to as performance interference.  ...  Experimental results on different scenarios with our energy efficiency and interference aware approach shows a reduction in energy consumption to the tune of 8 to 58% and 10× improvement in per request  ...  We follow a request batching approach [virtual batching] to answer the first question. The system waits for batching timeout then wakes up the CPU to process requests.  ... 
doi:10.3844/jcssp.2014.143.156 fatcat:livfktuhezajrid2nicrxqwqvq

Managing energy, performance and cost in large scale heterogeneous datacenters using migrations

Muhammad Zakarya, Lee Gillam
2019 Future generations computer systems  
We discuss these results for different combinations of VM allocation, migration policies and different benchmark workloads 1 . model resource and workload heterogeneities in the context of a cloud platform  ...  In this paper, we demonstrate how the performance of workloads across different CPU models leads to variability in energy efficiencies, and therefore costs.  ...  A specific CPU model may perform best for one kind of workload, but worst for another.  ... 
doi:10.1016/j.future.2018.10.044 fatcat:p5t3nwegmnc3xbbkqvvchy353m

Capelin: Data-Driven Capacity Procurement for Cloud Datacenters using Portfolios of Scenarios – Extended Technical Report [article]

Georgios Andreadis, Fabian Mastenbroek, Vincent van Beek, Alexandru Iosup
2021 arXiv   pre-print
Cloud datacenters provide a backbone to our digital society.  ...  Inaccurate capacity procurement for cloud datacenters can lead to significant performance degradation, denser targets for failure, and unsustainable energy consumption.  ...  We assume a common model [42, 60] of performance interference, with a score from 0 to 1 for a given set of collocated workloads, with 0 indicating full interference between VMs contending for the same  ... 
arXiv:2103.02060v1 fatcat:og66whvev5givdfh7upvg7qd3a

No DNN Left Behind: Improving Inference in the Cloud with Multi-Tenancy [article]

Amit Samanta and Suhas Shrinivasan and Antoine Kaufmann and Jonathan Mace
2019 arXiv   pre-print
Applications today rely on isolated ad-hoc deployments that force users to compromise on consistent latency, elasticity, or cost-efficiency, depending on workload characteristics.  ...  We argue that DNN inference is an ideal candidate for a multi-tenant system because of its narrow and well-defined interface and predictable resource requirements.  ...  Pre-and post-processing steps are often similar between DNNs, and can be batched, even across di erent model pipelines [29] .  ... 
arXiv:1901.06887v2 fatcat:k6bgjy7m3vcvrog5skrh55bzza

Understanding TCP incast throughput collapse in datacenter networks

Yanpei Chen, Rean Griffith, Junda Liu, Randy H. Katz, Anthony D. Joseph
2009 Proceedings of the 1st ACM workshop on Research on enterprise networking - WREN '09  
We propose an analytical model to account for the observed Incast symptoms, identify contributory factors, and explore the efficacy of solutions proposed by us and by others.  ...  This phenomenon has been observed by others in distributed storage, MapReduce and web-search workloads. In this paper we focus on understanding the dynamics of Incast.  ...  The authors would like to thank Lifan Zhang, Jon Kuroda and Keith Sklower for their help and support.  ... 
doi:10.1145/1592681.1592693 dblp:conf/sigcomm/ChenGLKJ09 fatcat:clriv4jsgrf3nnrlhndjfzxah4

Self-awareness of Cloud Applications [chapter]

Alex Iosup, Xiaoyun Zhu, Arif Merchant, Eva Kalyvianaki, Martina Maggio, Simon Spinner, Tarek Abdelzaher, Ole Mengshoel, Sara Bouchenak
2017 Self-Aware Computing Systems  
Last, we propose a roadmap for addressing open challenges in self-aware cloud and datacenter applications.  ...  A much shorter, revised version of this material will be available in print, as part of a Springer book on "Self-Aware Computing". The book is due to appear in 2017.  ...  Systems of Systems, by the Swedish Research Council (VR) for the projects "Cloud Control" and "Power and temperature control for large-scale computing infrastructures", and through the LCCC Linnaeus and  ... 
doi:10.1007/978-3-319-47474-8_20 fatcat:ckcxfmjmvvas5bhg7wp5qfmehy

The IX Operating System

Adam Belay, George Prekas, Mia Primorac, Ana Klimovic, Samuel Grossman, Christos Kozyrakis, Edouard Bugnion
2016 ACM Transactions on Computer Systems  
batches of packets to completion, and eliminating coherence traffic and multicore synchronization.  ...  The dataplane architecture builds upon a native, zero-copy API and optimizes for both bandwidth and latency by dedicating hardware threads and networking queues to dataplane instances, processing bounded  ...  ACKNOWLEDGMENTS The authors would like to thank David Mazières for his many insights into the system and his detailed feedback on the article.  ... 
doi:10.1145/2997641 fatcat:6nt2zson6jfjrieodatckxbfhi

Energy proportionality and workload consolidation for latency-critical applications

George Prekas, Mia Primorac, Adam Belay, Christos Kozyrakis, Edouard Bugnion
2015 Proceedings of the Sixth ACM Symposium on Cloud Computing - SoCC '15  
Such applications represent a growing subset of datacenter workloads and are typically deployed on dedicated servers, which is the simplest way to ensure low tail latency across all loads.  ...  We present the OS mechanisms and dynamic control needed to adjust core allocation and voltage/frequency settings based on the measured delays for latency-critical workloads.  ...  which relies on interrupts, is much more subject to load than individual static configurations of IX, which relies on a polling model.  ... 
doi:10.1145/2806777.2806848 dblp:conf/cloud/PrekasPBKB15 fatcat:4eh25vkcxvftzd5lkdmpjcd7a4

Intelligent Resource Scheduling at Scale: A Machine Learning Perspective

Renyu Yang, Xue Ouyang, Yaofeng Chen, Paul Townend, Jie Xu
2018 2018 IEEE Symposium on Service-Oriented System Engineering (SOSE)  
In this paper, we describe and discuss how ML can be used to autonomously exploit and understand both workloads and environments, and to learn how to efficiently deal with scheduling problems such as consolidating  ...  The exhibited heterogeneity of workload and server characteristics in Cloud-scale or Internetscale environments has raised unprecedented new challenges for cluster scheduling.  ...  RNN can use their internal memory to process arbitrary sequences of inputs.  ... 
doi:10.1109/sose.2018.00025 dblp:conf/sose/YangOCTX18 fatcat:yxyi5lwxmjdtdpqbrvblbqan5u

A Taxonomy and Future Directions for Sustainable Cloud Computing: 360 Degree View [article]

Sukhpal Singh Gill, Rajkumar Buyya
2018 arXiv   pre-print
A conceptual model for sustainable cloud computing has been proposed along with discussion on future research directions.  ...  The taxonomy is used to investigate the existing techniques for sustainability that need careful attention and investigation as proposed by several academic and industry groups.  ...  ACKNOWLEDGEMENTS We thank Patricia Arroba, Minxian Xu and Shashikant Ilager for their useful suggestions.  ... 
arXiv:1712.02899v2 fatcat:t26xxbgiijesneqgzi2mqz4gta

Hybrid Resource Management for HPC and Data Intensive Workloads

Abel Souza, Mohamad Rezaei, Erwin Laure, Johan Tordsson
2019 2019 19th IEEE/ACM International Symposium on Cluster, Cloud and Grid Computing (CCGRID)  
High Performance Computing (HPC) and Data Intensive (DI) workloads have been executed on separate clusters using different tools for resource and application management.  ...  The architecture is easily extensible to current and new types of distributed applications, allowing efficient combination of hybrid workloads on HPC resources with increased job throughput and higher  ...  In addition, this work was partially supported by the Swedish Research Council (VR) for the projects "Cloud Control" and by the Brazilian National Council for Scientific and Technological Development (  ... 
doi:10.1109/ccgrid.2019.00054 dblp:conf/ccgrid/SouzaRLT19 fatcat:vgw6xru7fnhszpysonfkucoohi

Prediction-Based Power Oversubscription in Cloud Platforms [article]

Alok Kumbhare, Reza Azimi, Ioannis Manousakis, Anand Bonde, Felipe Frujeri, Nithish Mahalingam, Pulkit Misra, Seyyed Ahmad Javadi, Bianca Schroeder, Marcus Fontoura, Ricardo Bianchini
2020 arXiv   pre-print
Datacenter designers rely on conservative estimates of IT equipment power draw to provision resources. This leaves resources underutilized and requires more datacenters to be built.  ...  In this paper, we argue that providers can use predictions of workload performance criticality and virtual machine (VM) resource utilization to increase oversubscription.  ...  For example, observing whether a VM exchanges messages does not work because many non-user-facing workloads communicate externally (e.g., to bring data in for batch processing).  ... 
arXiv:2010.15388v1 fatcat:vofbvotojnf2jnol6jxbnoreuq

D5.1: Accelerator Deployment Models

Eleni Kanellou, Nikolaos Chrysos, Angelos Bilas, Christoforos Kachris
2017 Zenodo  
VineSim can be used to explore how different deployments respond to different kinds of workloads, thus allowing one to determine how to best compose a datacenter based on particular workload, performance  ...  We have developed VineSim, a software simulator of a datacenter, based on the aforementioned theoretical modeling.  ...  Apart from analytical models, popular tools for such a purpose are software datacenter simulators, since they provide a cost-effective but also flexible way of doing any necessary predictions and estimations  ... 
doi:10.5281/zenodo.898170 fatcat:l6gfsswmbbg7niupwnix7tfmjm

Deep reinforcement learning for multi-objective placement of virtual machines in cloud datacenters

Luca Caviglione, Mauro Gaggero, Massimo Paolucci, Roberto Ronco
2020 Soft Computing - A Fusion of Foundations, Methodologies and Applications  
Placement strategies are found by using a deep reinforcement learning framework to select the best placement heuristic for each virtual machine composing the workload.  ...  Unfortunately, high packing factors could lead to performance and security issues, e.g., virtual machines can compete for hardware resources or collude to leak data.  ...  Acknowledgements We thank Netalia ( for the traces used as workload in the simulations.  ... 
doi:10.1007/s00500-020-05462-x fatcat:wibofjdyzfbwjgcnidv43mcelm
« Previous Showing results 1 — 15 out of 247 results