Quiet Neighborhoods: Key to Protect Job Performance Predictability

Jokanović, Ana; Sancho, José Carlos; Rodríguez, Germán; Lucero, Alejandro; Minkenberg, Cyriel; Labarta, Jesús

doi:10.1109/ipdps.2015.87

Cited by 37 publications

(12 citation statements)

References 17 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…However, adaptive routing does not necessarily reduce contention significantly [1]. Researchers have also studied job placement strategies that minimize contention [3], [23], [24], but these approaches typically reduce machine utilization.…”

Section: Related Workmentioning

confidence: 99%

Mitigating Inter-Job Interference via Process-Level Quality-of-Service

Savoie

Lowenthal

Supinski

et al. 2019

2019 IEEE International Conference on Cluster Computing (CLUSTER)

View full text Add to dashboard Cite

Section: Related Workmentioning

confidence: 99%

Mitigating Inter-Job Interference via Process-Level Quality-of-Service

Savoie

Lowenthal

Supinski

et al. 2019

2019 IEEE International Conference on Cluster Computing (CLUSTER)

View full text Add to dashboard Cite

“…Bhatele et al [21] investigated the performance variability in Cray machines and found out that the interference of multiple jobs that share the same network links is the primary factor for high performance variability. Jokanovic et al [22] introduce the concept of quiet neighborhoods, a job allocation technique based on the job sizes. This technique helps control the fragmentation in the HPC systems and reduces the number of jobs sharing the network, with the aim of minimizing the interference.…”

Section: Related Workmentioning

confidence: 99%

On the Root Causes of Cross-Application I/O Interference in HPC Storage Systems

Yildiz

Dorier

Ibrahim

et al. 2016

2016 IEEE International Parallel and Distributed Processing Symposium (IPDPS)

View full text Add to dashboard Cite

As we move toward the exascale era, performance variability in HPC systems remains a challenge. I/O interference, a major cause of this variability, is becoming more important every day with the growing number of concurrent applications that share larger machines. Earlier research efforts on mitigating I/O interference focus on a single potential cause of interference (e.g., the network). Yet the root causes of I/O interference can be diverse. In this work, we conduct an extensive experimental campaign to explore the various root causes of I/O interference in HPC storage systems. We use microbenchmarks on the Grid'5000 testbed to evaluate how the applications' access pattern, the network components, the file system's configuration, and the backend storage devices influence I/O interference. Our studies reveal that in many situations interference is a result of bad flow control in the I/O path, rather than being caused by some single bottleneck in one of its components. We further show that interferencefree behavior is not necessarily a sign of optimal performance. To the best of our knowledge, our work provides the first deep insight into the role of each of the potential root causes of interference and their interplay. Our findings can help developers and platform owners improve I/O performance and motivate further research addressing the problem across all components of the I/O stack.

show abstract

“…By definition, clouds provide a shared resource model where multiple tenants are served from the same data center infrastructure. The sharing of resources may result in unpredictable application performance [40,79,[99][100][101]. The performance unpredictability in a multi-tenant cloud computing system typically arises from server virtualization and network sharing.…”

Section: Research Questionsmentioning

confidence: 99%