Platform and Co-Runner Affinities for Many-Task Applications in Distributed Computing Platforms

Kim, Byungsu; Hwang, Eunji; Yoo, Tae-kyung; Kim, Jik-Soo; Hwang, Soonwook; Choi, Young-ri

doi:10.1109/ccgrid.2015.129

Cited by 11 publications

(35 citation statements)

References 16 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The operator of the data center should define types according to observed performance dependencies. A type could correspond to a specific application (as in [20]); but it could also be more general, gathering, for example, all webservers under a single type, and all databases under another one. Here we assume that the task's type is known to the resource manager either from the analysis of previous submissions, or from users' declarations (this assumption corresponds to the clairvoyance assumption in classic scheduling).…”

Section: Side-effects Of Colocating Tasks: a Modelmentioning

confidence: 99%

“…Tasks colocated on a machine compete for shared hardware. Despite significant advances in both OS-level fairness and VM hypervisors, virtualization is not transparent: multiple studies show [19], [20], [21], [29], [36] that the performance of colocated tasks drops. Suspected reasons include difficulties in sharing the CPU cache or the memory bandwidth.…”

Section: Introductionmentioning

confidence: 99%

“…Each type additionally defines a performance function mapping this vector of loads to a type-relevant performance metric. As datacenters execute multiple instances of tasks, such a function can be inferred by a monitoring module [20], [29], [36] which will match task's reported performance (such as the 95th percentile response time) with observed or reported loads.…”

Section: Introductionmentioning

confidence: 99%

“…We use a linear performance function: on each machine, the influence that a type t has on the performance of a task of type t is a product of the load of type t on this machine and a coefficient α t ,t . The coefficient α t ,t describes how compatible t load is with a task of type t (our coefficients are similar to interference/affinity metrics proposed in [20], [29]). Low values (0 ≤ α t ,t < 1) correspond to compatible types (e.g.…”

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

Optimizing Egalitarian Performance when Colocating Tasks with Types for Cloud Data Center Resource Management

Pascual

Rządca

2019

IEEE Trans. Parallel Distrib. Syst.

View full text Add to dashboard Cite

In data centers, up to dozens of tasks are colocated on a single physical machine. Machines are used more efficiently, but the performance of the tasks deteriorates, as the colocated tasks compete for shared resources. Since the tasks are heterogeneous, the resulting performance dependencies are complex. In our previous work [26], [27] we proposed a new combinatorial optimization model that uses two parameters of a task -its size and its type -to characterize how a task influences the performance of other tasks allocated to the same machine.In this paper, we study the egalitarian optimization goal: the aim is to optimize the performance of the worst-off task. This problem generalizes the classic makespan minimization on multiple processors (P||C max ). We prove that polynomially-solvable variants of P||C max are NP-hard for this generalization, and that the problem is hard to approximate when the number of types is not constant. For a constant number of types, we propose a PTAS, a fast approximation algorithm, and a series of heuristics. We simulate the algorithms on instances derived from a trace of one of Google clusters. Compared with baseline algorithms solving P||C max , our proposed algorithms aware of the types of the jobs lead to significantly better tasks' performance. The notion of type enables us to extend standard combinatorial optimization methods to handle degradation of performance caused by colocation. Types add a layer of additional complexity. However, our results -approximation algorithms and good average-case performance -show that types can be handled efficiently.

show abstract

Section: Side-effects Of Colocating Tasks: a Modelmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Optimizing Egalitarian Performance when Colocating Tasks with Types for Cloud Data Center Resource Management

Pascual

Rządca

2019

IEEE Trans. Parallel Distrib. Syst.

View full text Add to dashboard Cite

show abstract

“…We use a linear performance function: on each machine, the influence a type t has on type t performance is a product of the load of type t and a coefficient α t ,t . The coefficient α t ,t describes how compatible t load is with t performance (the coefficient is similar to interference/affinity metrics proposed in [13,20]). Low values (0 ≤ α t ,t < 1) correspond with compatible types (e.g.…”

Section: Introductionmentioning

confidence: 99%

Optimizing Egalitarian Performance in the Side-Effects Model of Colocation for Data Center Resource Management

Pascual

Rządca

2017

Lecture Notes in Computer Science

View full text Add to dashboard Cite

Abstract. In data centers, up to dozens of tasks are colocated on a single physical machine. Machines are used more efficiently, but tasks' performance deteriorates, as colocated tasks compete for shared resources. As tasks are heterogeneous, the resulting performance dependencies are complex. In our previous work [18] we proposed a new combinatorial optimization model that uses two parameters of a task -its size and its type -to characterize how a task influences the performance of other tasks allocated to the same machine. In this paper, we study the egalitarian optimization goal: maximizing the worstoff performance. This problem generalizes the classic makespan minimization on multiple processors (P||C max ). We prove that polynomially-solvable variants of P||C max are NP-hard and hard to approximate when the number of types is not constant. For a constant number of types, we propose a PTAS, a fast approximation algorithm, and a series of heuristics. We simulate the algorithms on instances derived from a trace of one of Google clusters. Algorithms aware of jobs' types lead to better performance compared to algorithms solving P||C max . The notion of type enables us to model degeneration of performance caused by colocation using standard combinatorial optimization methods. Types add a layer of additional complexity. However, our results -approximation algorithms and good average-case performance -show that types can be handled efficiently.

show abstract

Hierarchical Recursive Resource Sharing for Containerized Applications

Kim

Lee

Han

et al. 2018

Service-Oriented Computing

View full text Add to dashboard Cite

Platform and Co-Runner Affinities for Many-Task Applications in Distributed Computing Platforms

Cited by 11 publications

References 16 publications

Optimizing Egalitarian Performance when Colocating Tasks with Types for Cloud Data Center Resource Management

Optimizing Egalitarian Performance when Colocating Tasks with Types for Cloud Data Center Resource Management

Optimizing Egalitarian Performance in the Side-Effects Model of Colocation for Data Center Resource Management

Hierarchical Recursive Resource Sharing for Containerized Applications

Contact Info

Product

Resources

About