A Methodology for Online Consolidation of Tasks through More Accurate Resource Estimations

Iglesias, Jesús Omana; Murphy, Liam; Cauwer, Milan De; Mehta, Deepak; O’Sullivan, Barry

doi:10.1109/ucc.2014.17

Cited by 12 publications

(6 citation statements)

References 18 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Using the Gaussian distribution is a remarkably simple approachwe achieve satisfying QoS without relying on machine learning [17] or artificial neural networks [4]. We claim that this proves how important high-frequency data is for allocation algorithms.…”

Section: Discussionmentioning

confidence: 78%

“…Our method estimates tasks' mean and standard deviation from tasks' observed instantaneous usage. [17] combines scheduling and two dimensional bin packing to optimize tasks' cumulative waiting time and machines' utilization. The method uses machine learning to predict tasks' peak CPU and memory consumption based on observing first 24 hours of task's resource usage.…”

Section: Related Workmentioning

confidence: 99%

See 1 more Smart Citation

SLO-aware colocation of data center tasks based on instantaneous processor requirements

Janus

Rządca

2017

Proceedings of the 2017 Symposium on Cloud Computing

View full text Add to dashboard Cite

In a cloud data center, a single physical machine simultaneously executes dozens of highly heterogeneous tasks. Such colocation results in more efficient utilization of machines, but, when tasks' requirements exceed available resources, some of the tasks might be throttled down or preempted. We analyze version 2.1 of the Google cluster trace that shows short-term (1 second) task CPU usage. Contrary to the assumptions taken by many theoretical studies, we demonstrate that the empirical distributions do not follow any single distribution. However, high percentiles of the total processor usage (summed over at least 10 tasks) can be reasonably estimated by the Gaussian distribution. We use this result for a probabilistic fit test, called the Gaussian Percentile Approximation (GPA), for standard bin-packing algorithms. To check whether a new task will fit into a machine, GPA checks whether the resulting distribution's percentile corresponding to the requested service level objective, SLO is still below the machine's capacity. In our simulation experiments, GPA resulted in colocations exceeding the machines' capacity with a frequency similar to the requested SLO. CCS CONCEPTS• Computer systems organization → Cloud computing; • Software and its engineering → Process management; • Computing methodologies → Planning and scheduling; KEYWORDS scheduling, resource management, stochastic bin packing ACM Reference Format:

show abstract

Section: Discussionmentioning

confidence: 78%

Section: Related Workmentioning

confidence: 99%

SLO-aware colocation of data center tasks based on instantaneous processor requirements

Janus

Rządca

2017

Proceedings of the 2017 Symposium on Cloud Computing

View full text Add to dashboard Cite

show abstract

“…Traces (either synthetic or the exact events) can be used for validation of various workload management algorithms. The Google trace has been used recently in [17] to evaluate consolidation strategies, in [4,5] to validate overcommitting (overbooking), in [38] to perform provisioning for heterogeneous systems and in [12] to investigate checkpointing algorithms. Again, data analysis is performed individually by the research groups and no specific tool was published.…”

Section: Related Workmentioning

confidence: 99%

A Big Data analyzer for large trace logs

et al. 2015

View full text Add to dashboard Cite

Current generation of Internet-based services are typically hosted on large data centers that take the form of warehouse-size structures housing tens of thousands of servers. Continued availability of a modern data center is the result of a complex orchestration among many internal and external actors including computing hardware, multiple layers of intricate software, networking and storage devices, electrical power and cooling plants. During the course of their operation, many of these components produce large amounts of data in the form of event and error logs that are essential not only for identifying and resolving problems but also for improving data center efficiency and management. Most of these activities would benefit significantly from data analytics techniques to exploit hidden statistical patterns and correlations that may be present in the data. The sheer volume of data to be analyzed makes uncovering these correlations and patterns a challenging task. This paper presents Big Data analyzer (BiDAl), a prototype Java tool for log-data analysis that incorporates several Big Data technologies in order to simplify the task of extracting information from data traces produced by large clusters and server farms. BiDAl provides the user with several analysis languages (SQL, R and Hadoop MapReduce) and storage backends (HDFS and SQLite) that can be freely mixed and matched so that a custom tool for a specific task can be easily constructed. BiDAl has a modular architecture so that it can be extended with other backends and analysis languages in the future. In this paper we present the design of BiDAl and describe our experience using it to analyze publicly-available traces from Google data clusters, with the goal of building a realistic model of a complex data center

show abstract

“…A different type of usage is for validation of various workload management algorithms. Examples are [21] where the trace is used to evaluate consolidation strategies, [22], [23] where over-committing (overbooking) is validated, [24] who take heterogeneity into account to perform provisioning or [25] investigating checkpointing algorithms.…”

Section: Related Workmentioning

confidence: 99%

Towards Data-Driven Autonomics in Data Centers

Sîrbu

Babaoğlu

2015

2015 International Conference on Cloud and Autonomic Computing

View full text Add to dashboard Cite

Continued reliance on human operators for managing\ud data centers is a major impediment for them from\ud ever reaching extreme dimensions. Large computer systems\ud in general, and data centers in particular, will ultimately be\ud managed using predictive computational and executable models\ud obtained through data-science tools, and at that point, the\ud intervention of humans will be limited to setting high-level\ud goals and policies rather than performing low-level operations.\ud Data-driven autonomics, where management and control are\ud based on holistic predictive models that are built and updated\ud using generated data, opens one possible path towards limiting\ud the role of operators in data centers. In this paper, we present\ud a data-science study of a public Google dataset collected in a\ud 12K-node cluster with the goal of building and evaluating a\ud predictive model for node failures. We use BigQuery, the big\ud data SQL platform from the Google Cloud suite, to process\ud massive amounts of data and generate a rich feature set\ud characterizing machine state over time. We describe how an\ud ensemble classifier can be built out of many Random Forest\ud classifiers each trained on these features, to predict if machines\ud will fail in a future 24-hour window. Our evaluation reveals\ud that if we limit false positive rates to 5%, we can achieve true\ud positive rates between 27% and 88% with precision varying\ud between 50% and 72%. We discuss the practicality of including\ud our predictive model as the central component of a data-driven\ud autonomic manager and operating it on-line with live data\ud streams (rather than off-line on data logs). All of the scripts\ud used for BigQuery and classification analyses are publicly\ud available from the authors’ website

show abstract

A Methodology for Online Consolidation of Tasks through More Accurate Resource Estimations

Cited by 12 publications

References 18 publications

SLO-aware colocation of data center tasks based on instantaneous processor requirements

SLO-aware colocation of data center tasks based on instantaneous processor requirements

A Big Data analyzer for large trace logs

Towards Data-Driven Autonomics in Data Centers

Contact Info

Product

Resources

About