Value Based Scheduling for Oversubscribed Power-Constrained Homogeneous HPC Systems

Kumbhare, Nirmal; Tunc, Cihan; Machovec, Dylan; Akoglu, Ali; Hariri, Salim; Siegel, Howard Jay

doi:10.1109/iccac.2017.19

Cited by 10 publications

(6 citation statements)

References 29 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…We are currently addressing challenges of VDCs management on simpler environments, on cloud resource management heuristics, (e.g., [8,9,6,7]), big data analysis, and data mining for performance prediction. To simulate, evaluate, analyze, and compare different heuristics, we will build simulators for simpler environments and combine open-source simulators for different levels of the JITA-4DS hierarchy.…”

Section: Discussionmentioning

confidence: 99%

“…To assess disaggregated DC's, we study how to model and validate their performance in large-scale settings. We rely on our novel model-driven resource management heuristics [8,6,7] based on metrics that measure a service's value for achieving a balance between competing goals (e.g., completion time and energy consumption). Our focus is on defining new system performance measures that combine objectives, such as execution time and energy use, that dynamically change during the day.…”

Section: Towards Just In Time Virtual Data Centres For Data Science W...mentioning

confidence: 99%

See 1 more Smart Citation

Graph Analytics Workflows Enactment on Just in Time Data Centres Position Paper

Akoglu

Zechinelli-Martini

Kheddouci

et al. 2022

Service-Oriented Computing – ICSOC 2021 Workshops

Self Cite

View full text Add to dashboard Cite

This paper discusses our vision about multirole-capable decisionmaking systems across a broad range of Data Science (DS) workflows working on graphs through disaggregated data centres. Our vision is that an alternative is possible, to work on a disaggregated solution for the provision of computational services under the notion of a disaggregated data centre. We define this alternative as a virtual entity that dynamically provides resources crosscutting the layers of edge, fog and data centre according to the workloads submitted by the workflows and their Service Level Objectives. This work is funded by the project GALILEAN, LIRIS intergroup collaboration https://galilean-project.github.io.

show abstract

Section: Discussionmentioning

confidence: 99%

Section: Towards Just In Time Virtual Data Centres For Data Science W...mentioning

confidence: 99%

Graph Analytics Workflows Enactment on Just in Time Data Centres Position Paper

Akoglu

Zechinelli-Martini

Kheddouci

et al. 2022

Service-Oriented Computing – ICSOC 2021 Workshops

Self Cite

View full text Add to dashboard Cite

show abstract

“…JITA-4DS encourages a novel resource management methodology that is based on the time-dependent Value of Service (VoS) metric [12] to guide the assignment of resources to each VDC and achieve a balance between goals that usually compete with each other (e.g., completion time and energy consumption). VoS allows considering the relative importance of the competing goals, the submission time of the task (e.g., peak vs non-peak period), or the task's nature as a function of task completion time.…”

Section: Value Of Service Based Scheduling and Resource Managementmentioning

confidence: 99%

“…Each task was associated with a task type, which has estimated execution time and energy consumption characteristics (through historical information or experiments) for a given number of assigned cores and assigned memory. To predict each application type's execution time and energy consumption, we use statistical and data mining techniques [12,10,11], which represent the execution time and energy consumption as a function of the VDC resources. As an example, one of the heuristics was Maximum Value-per-Total Resources (Maximum VPTR).…”

Section: Value Of Service Based Scheduling and Resource Managementmentioning

confidence: 99%

Putting Data Science Pipelines on the Edge

Akoglu¹,

Vargas-Solar²

2021

Preprint

Self Cite

View full text Add to dashboard Cite

This paper proposes a composable "Just in Time Architecture" for Data Science (DS) Pipelines named JITA-4DS and associated resource management techniques for configuring disaggregated data centers (DCs). DCs under our approach are composable based on vertical integration of the application, middleware/operating system, and hardware layers customized dynamically to meet application Service Level Objectives (SLO -application-aware management). Thereby, pipelines utilize a set of flexible building blocks that can be dynamically and automatically assembled and re-assembled to meet the dynamic changes in the workload's SLOs. To assess disaggregated DC's, we study how to model and validate their performance in large-scale settings.

show abstract

“…JITA-4DS encourages a novel resource management methodology that is based on the time dependent value of service (VoS) metric to guide the assignment of resources to each VDC that maximizes the overall systemwide VoS metric. To predict the execution time and energy consumption of each application type, we use statistical and data mining techniques [20][21][22][23], which represent the execution time and energy consumption as a function of the VDC resources. A complete study of these aspects for JITA-4DS have been described in [12].…”

Section: Jita-4ds: Just In Time Edge Based Data Science Pipelines Exe...mentioning

confidence: 99%

JITA4DS: Disaggregated execution of Data Science Pipelines between the Edge and the Data Centre

Vargas-Solar¹,

Akoglu²,

Hassan³

2021

Preprint

Self Cite

View full text Add to dashboard Cite

This paper targets the execution of data science (DS) pipelines supported by data processing, transmission and sharing across several resources executing greedy processes. Current data science pipelines environments provide various infrastructure services with computing resources such as general-purpose processors (GPP), Graphics Processing Units (GPUs), Field Programmable Gate Arrays (FPGAs) and Tensor Processing Unit (TPU) coupled with platform and software services to design, run and maintain DS pipelines. These one-fits-all solutions impose the complete externalization of data pipeline tasks. However, some tasks can be executed in the edge, and the backend can provide just in time resources to ensure ad-hoc and elastic execution environments.This paper introduces an innovative composable "Just in Time Architecture" for configuring DCs for Data Science Pipelines (JITA-4DS) and associated resource management techniques. JITA-4DS is a cross-layer management system that is aware of both the application characteristics and the underlying infrastructures to break the barriers between applications, middleware/operating system, and hardware layers. Vertical integration of these layers is needed for building a customizable Virtual Data Center (VDC) to meet the dynamically changing data science pipelines' requirements such as

show abstract

Value Based Scheduling for Oversubscribed Power-Constrained Homogeneous HPC Systems

Cited by 10 publications

References 29 publications

Graph Analytics Workflows Enactment on Just in Time Data Centres Position Paper

Graph Analytics Workflows Enactment on Just in Time Data Centres Position Paper

Putting Data Science Pipelines on the Edge

JITA4DS: Disaggregated execution of Data Science Pipelines between the Edge and the Data Centre

Contact Info

Product

Resources

About