A deep investigation into network performance in virtual machine based cloud environments

Shea, Ryan; Wang, Feng; Wang, Haiyang; Liu, Jiangchuan

doi:10.1109/infocom.2014.6848061

Cited by 55 publications

(27 citation statements)

References 13 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The first category [9,35,37,40,42] concentrates on studying the impact of virtualization on the network and application performance in a cloud environment. These works show that sharing of hardware resources can have a negative effect on latency, throughput, and bandwidth of applications running on these virtual machines.…”

Section: Related Workmentioning

confidence: 99%

The nearest replica can be farther than you think

Bogdanov

Peón-Quirós

Maguire

et al. 2015

Proceedings of the Sixth ACM Symposium on Cloud Computing

View full text Add to dashboard Cite

Modern distributed systems are geo-distributed for reasons of increased performance, reliability, and survivability. At the heart of many such systems, e.g., the widely used Cassandra and MongoDB data stores, is an algorithm for choosing a closest set of replicas to service a client request. Suboptimal replica choices due to dynamically changing network conditions result in reduced performance as a result of increased response latency. We present GeoPerf, a tool that tries to automate the process of systematically testing the performance of replica selection algorithms for geodistributed storage systems. Our key idea is to combine symbolic execution and lightweight modeling to generate a set of inputs that can expose weaknesses in replica selection. As part of our evaluation, we analyzed network round trip times between geographically distributed Amazon EC2 regions, and showed a significant number of daily changes in nearest-K replica orders. We tested Cassandra and MongoDB using our tool, and found bugs in each of these systems. Finally, we use our collected Amazon EC2 latency traces to quantify the time lost due to these bugs. For example due to the bug in Cassandra, the median wasted time for 10% of all requests is above 50 ms.

show abstract

Section: Related Workmentioning

confidence: 99%

The nearest replica can be farther than you think

Bogdanov

Peón-Quirós

Maguire

et al. 2015

Proceedings of the Sixth ACM Symposium on Cloud Computing

View full text Add to dashboard Cite

show abstract

“…Many benchmarking studies are devoted to understanding the performance of EC2 VM instances [1], [16], their network [3], [6], [17] and applications deployed in them [18], [19]. These studies found that virtualization and multi-tenancy are the major causes for resource contention as multiple VMs are placed on a single host.…”

Section: Performance Interference In Paravirtualizationmentioning

confidence: 99%

“…CPU performance is affected by the time slices allocated to the VMs which are based on a weight (the CPU share for each VM and the hypervisor), a cap (the maximum usage), and the amount of pending tasks for the physical processors [23], [24], [25]. Disk and network I/O suffer from overhead caused by moving data between the VM, shared memory, and the physical devices [17], [25]. As shown in Figure 1, Xen uses two daemon processes, blkfront and blkback, to pass I/O requests between DomU and Dom0 via a shared memory page.…”

Section: A Virtualization Backgroundmentioning

confidence: 99%

“…Cherkasova et al [23] analyzed the impact of different Xen schedulers on application performance and discussed challenges in estimating application resource requirements in virtualized environments. Shea et al [17] studied the performance degradation and variation for TCP and UDP traffic, then provided a hypervisor reconfiguration solution to enhance the performance. Pu et al [25] presented experimental research on performance interference on CPU and network intensive workloads on the Xen hypervisor and reached the conclusion that identifying the impact of exchanged memory pages is important to the in-depth understanding of interference costs in Dom0.…”

Section: Related Workmentioning

confidence: 99%

See 1 more Smart Citation

CloudScope: Diagnosing and Managing Performance Interference in Multi-tenant Clouds

Chen

Rupprecht

Osman

et al. 2015

2015 IEEE 23rd International Symposium on Modeling, Analysis, and Simulation of Computer and Telecommunication Systems

View full text Add to dashboard Cite

Abstract-Virtual machine consolidation is attractive in cloud computing platforms for several reasons including reduced infrastructure costs, lower energy consumption and ease of management. However, the interference between co-resident workloads caused by virtualization can violate the service level objectives (SLOs) that the cloud platform guarantees. Existing solutions to minimize interference between virtual machines (VMs) are mostly based on comprehensive micro-benchmarks or online training which makes them computationally intensive.In this paper, we present CloudScope, a system for diagnosing interference for multi-tenant cloud systems in a lightweight way. CloudScope employs a discrete-time Markov Chain model for the online prediction of performance interference of co-resident VMs. It uses the results to optimally (re)assign VMs to physical machines and to optimize the hypervisor configuration, e.g. the CPU share it can use, for different workloads. We have implemented CloudScope on top of the Xen hypervisor and conducted experiments using a set of CPU, disk, and network intensive workloads and a real system (MapReduce). Our results show that CloudScope interference prediction achieves an average error of 9%. The interference-aware scheduler improves VM performance by up to 10% compared to the default scheduler. In addition, the hypervisor reconfiguration can improve network throughput by up to 30%.

show abstract

“…Nevertheless, recent studies have shown that RTT figures between VMs may vary widely in a cloud environment and this is due mainly to the virtualization overhead since the CPU has a dual role in both computation and networking [9]. Consequently, it is not true that network adjacent VMs in a cloud environment will necessarily be physically close to one another, as is generally the case in network topologies such as that of the Internet.…”

Section: A Latency Matrixmentioning

confidence: 99%

End-to-end informed VM selection in compute clouds

Teixeira

Bestavros

2015

2015 IEEE International Conference on Communication Workshop (ICCW)

View full text Add to dashboard Cite

Abstract-The selection of resources, particularly VMs, in current public IaaS clouds is usually done in a blind fashion, as cloud users do not have much information about resource consumption by co-tenant third-party tasks. In particular, communication patterns can play a significant part in cloud application performance and responsiveness, specially in the case of novel latencysensitive applications, increasingly common in today's clouds. Thus, herein we propose an end-to-end approach to the VM allocation problem using policies based uniquely on round-trip time measurements between VMs. Those become part of a userlevel 'Recommender Service' that receives VM allocation requests with certain network-related demands and matches them to a suitable subset of VMs available to the user within the cloud. We propose and implement end-to-end algorithms for VM selection that cover desirable profiles of communications between VMs in distributed applications in a cloud setting, such as profiles with prevailing pair-wise, hub-and-spokes, or clustered communication patterns between constituent VMs. We quantify the expected benefits from deploying our Recommender Service by comparing our informed VM allocation approaches to conventional, random allocation methods, based on real measurements of latencies between Amazon EC2 instances. We also show that our approach is completely independent from cloud architecture details, is adaptable to different types of applications and workloads, and is lightweight and transparent to cloud providers.

show abstract

A deep investigation into network performance in virtual machine based cloud environments

Cited by 55 publications

References 13 publications

The nearest replica can be farther than you think

The nearest replica can be farther than you think

CloudScope: Diagnosing and Managing Performance Interference in Multi-tenant Clouds

End-to-end informed VM selection in compute clouds

Contact Info

Product

Resources

About