Efficient Workload Consolidation for Composable/Disaggregated Data Centers Considering Migration Cost

Chao Guo; Moshe Zukerman

doi:10.1109/ICTON59386.2023.10207292

Efficient Workload Consolidation for Composable/Disaggregated Data Centers Considering Migration Cost

Chao Guo^*
, Moshe Zukerman

^*Corresponding author for this work

Department of Electrical Engineering

Research output: Chapters, Conference Papers, Creative and Literary Works › RGC 32 - Refereed conference paper (with host publication) › peer-review

2 Citations (Scopus)

60 Downloads (CityUHK Scholars)

Abstract

Resource disaggregation significantly improves resource flexibility, efficiency, and availability in Data Centers (DCs) by decoupling different resource modules in servers and interconnecting them with high-speed networking technologies, e.g., optical networking, leading to Composable/Disaggregated DCs (CDCs or DDCs). CDC also brings benefits to workload migration, a technique widely used to consolidate workloads to fewer nodes to save energy. CDC may reduce migration costs as a workload can be partly migrated instead of entirely as in a server-based DC (SDC). We consider the problem of workload consolidation in a CDC with objectives to minimize both the number of active nodes and the number of migrated elements. An integer linear programming formulation that uses a weighted sum approach is provided to address this two-objective problem. Results show that higher resource utilization and lower migration costs can simultaneously be achieved in a CDC as compared to an SDC.

©2023 IEEE

Original language	English
Title of host publication	2023 23rd International Conference on Transparent Optical Networks (ICTON)
Publisher	IEEE
ISBN (Electronic)	979-8-3503-0303-2
DOIs	https://doi.org/10.1109/ICTON59386.2023.10207292
Publication status	Published - 2023
Event	23^rd International Conference on Transparent Optical Networks - Central Library of University Politehnica Bucharest, Bucharest, Romania Duration: 2 Jul 2023 → 6 Jul 2023 https://icton2023.upb.ro/

Conference

Conference	23^rd International Conference on Transparent Optical Networks
Place	Romania
City	Bucharest
Period	2/07/23 → 6/07/23
Internet address	https://icton2023.upb.ro/

Research Keywords

Composable/disaggregated infrastructure
data center
ILP
consolidation
migration

Publisher's Copyright Statement

COPYRIGHT TERMS OF DEPOSITED POSTPRINT FILE: © 2023 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses, in any current or future media, including reprinting/republishing this material for advertising or promotional purposes, creating new collective works, for resale or redistribution to servers or lists, or reuse of any copyrighted component of this work in other works. Guo, C., & Zukerman, M. (2023). Efficient Workload Consolidation for Composable/Disaggregated Data Centers Considering Migration Cost. In 2023 23rd International Conference on Transparent Optical Networks (ICTON) Institute of Electrical and Electronics Engineers, Inc. https://doi.org/10.1109/ICTON59386.2023.10207292

Access Output

10.1109/ICTON59386.2023.10207292

161355164.pdfPost-print, 369 KB

Other Access Options

Check CityUHK Library

Permanent Link

Right click to copy link

Cite this

@inproceedings{009e355f2a00428bad72fea226283c34,

title = "Efficient Workload Consolidation for Composable/Disaggregated Data Centers Considering Migration Cost",

abstract = "Resource disaggregation significantly improves resource flexibility, efficiency, and availability in Data Centers (DCs) by decoupling different resource modules in servers and interconnecting them with high-speed networking technologies, e.g., optical networking, leading to Composable/Disaggregated DCs (CDCs or DDCs). CDC also brings benefits to workload migration, a technique widely used to consolidate workloads to fewer nodes to save energy. CDC may reduce migration costs as a workload can be partly migrated instead of entirely as in a server-based DC (SDC). We consider the problem of workload consolidation in a CDC with objectives to minimize both the number of active nodes and the number of migrated elements. An integer linear programming formulation that uses a weighted sum approach is provided to address this two-objective problem. Results show that higher resource utilization and lower migration costs can simultaneously be achieved in a CDC as compared to an SDC. {\textcopyright}2023 IEEE",

keywords = "Composable/disaggregated infrastructure, data center, ILP, consolidation, migration",

author = "Chao Guo and Moshe Zukerman",

year = "2023",

doi = "10.1109/ICTON59386.2023.10207292",

language = "English",

booktitle = "2023 23rd International Conference on Transparent Optical Networks (ICTON)",

publisher = "IEEE",

address = "United States",

note = "23<sup>rd</sup> International Conference on Transparent Optical Networks ; Conference date: 02-07-2023 Through 06-07-2023",

url = "https://icton2023.upb.ro/",

}

Efficient Workload Consolidation for Composable/Disaggregated Data Centers Considering Migration Cost. / Guo, Chao ; Zukerman, Moshe.
2023 23rd International Conference on Transparent Optical Networks (ICTON). IEEE, 2023.

Research output: Chapters, Conference Papers, Creative and Literary Works › RGC 32 - Refereed conference paper (with host publication) › peer-review

TY - GEN

T1 - Efficient Workload Consolidation for Composable/Disaggregated Data Centers Considering Migration Cost

AU - Guo, Chao

AU - Zukerman, Moshe

PY - 2023

Y1 - 2023

N2 - Resource disaggregation significantly improves resource flexibility, efficiency, and availability in Data Centers (DCs) by decoupling different resource modules in servers and interconnecting them with high-speed networking technologies, e.g., optical networking, leading to Composable/Disaggregated DCs (CDCs or DDCs). CDC also brings benefits to workload migration, a technique widely used to consolidate workloads to fewer nodes to save energy. CDC may reduce migration costs as a workload can be partly migrated instead of entirely as in a server-based DC (SDC). We consider the problem of workload consolidation in a CDC with objectives to minimize both the number of active nodes and the number of migrated elements. An integer linear programming formulation that uses a weighted sum approach is provided to address this two-objective problem. Results show that higher resource utilization and lower migration costs can simultaneously be achieved in a CDC as compared to an SDC. ©2023 IEEE

AB - Resource disaggregation significantly improves resource flexibility, efficiency, and availability in Data Centers (DCs) by decoupling different resource modules in servers and interconnecting them with high-speed networking technologies, e.g., optical networking, leading to Composable/Disaggregated DCs (CDCs or DDCs). CDC also brings benefits to workload migration, a technique widely used to consolidate workloads to fewer nodes to save energy. CDC may reduce migration costs as a workload can be partly migrated instead of entirely as in a server-based DC (SDC). We consider the problem of workload consolidation in a CDC with objectives to minimize both the number of active nodes and the number of migrated elements. An integer linear programming formulation that uses a weighted sum approach is provided to address this two-objective problem. Results show that higher resource utilization and lower migration costs can simultaneously be achieved in a CDC as compared to an SDC. ©2023 IEEE

KW - Composable/disaggregated infrastructure

KW - data center

KW - ILP

KW - consolidation

KW - migration

U2 - 10.1109/ICTON59386.2023.10207292

DO - 10.1109/ICTON59386.2023.10207292

M3 - RGC 32 - Refereed conference paper (with host publication)

BT - 2023 23rd International Conference on Transparent Optical Networks (ICTON)

PB - IEEE

T2 - 23<sup>rd</sup> International Conference on Transparent Optical Networks

Y2 - 2 July 2023 through 6 July 2023

ER -

Efficient Workload Consolidation for Composable/Disaggregated Data Centers Considering Migration Cost

Abstract

Conference

Research Keywords

Publisher's Copyright Statement

Access Output

Other Access Options

Permanent Link

Fingerprint

Cite this