GReM: Dynamic SSD resource allocation in virtualized storage systems with heterogeneous IO workloads

Zhengyu Yang, Jianzhe Tai, Janki Bhimani, Jiayin Wang, Ningfang Mi, Bo Sheng

Research output: Chapter in Book/Report/Conference proceedingConference contribution

26 Citations (Scopus)

Abstract

In a shared virtualized storage system that runs VMs with heterogeneous IO demands, it becomes a problem for the hypervisor to cost-effectively partition and allocate SSD resources among multiple VMs. There are two straightforward approaches to solving this problem: equally assigning SSDs to each VM or managing SSD resources in a fair competition mode. Unfortunately, neither of these approaches can fully utilize the benefits of SSD resources, particularly when the workloads frequently change and bursty IOs occur from time to time. In this paper, we design a Global SSD Resource Management solution - GReM, which aims to fully utilize SSD resources as a second-level cache under the consideration of performance isolation. In particular, GReM takes dynamic IO demands of all VMs into consideration to split the entire SSD space into a long-term zone and a short-term zone, and cost-effectively updates the content of SSDs in these two zones. GReM is able to adaptively adjust the reservation for each VM inside the long-term zone based on their IO changes. GReM can further dynamically partition SSDs between the long- and short-term zones during runtime by leveraging the feedbacks from both cache performance and bursty workloads. Experimental results show that GReM can capture the cross-VM IO changes to make correct decisions on resource allocation, and thus obtain high IO hit ratio and low IO management costs, compared with both traditional and state-of-the-art caching algorithms.

Original languageEnglish
Title of host publication2016 IEEE 35th International Performance Computing and Communications Conference, IPCCC 2016
PublisherInstitute of Electrical and Electronics Engineers Inc.
ISBN (Electronic)9781509052523
DOIs
StatePublished - 17 Jan 2017
Event35th IEEE International Performance Computing and Communications Conference, IPCCC 2016 - Las Vegas, United States
Duration: 9 Dec 201611 Dec 2016

Publication series

Name2016 IEEE 35th International Performance Computing and Communications Conference, IPCCC 2016

Other

Other35th IEEE International Performance Computing and Communications Conference, IPCCC 2016
CountryUnited States
CityLas Vegas
Period9/12/1611/12/16

Fingerprint

Resource allocation
Costs
Feedback

Keywords

  • Bursty Detection
  • Caching Algorithms
  • I/O Workload Characterization
  • Resource Allocation
  • Solid State Drives
  • Virtualized Storage Systems

Cite this

Yang, Z., Tai, J., Bhimani, J., Wang, J., Mi, N., & Sheng, B. (2017). GReM: Dynamic SSD resource allocation in virtualized storage systems with heterogeneous IO workloads. In 2016 IEEE 35th International Performance Computing and Communications Conference, IPCCC 2016 [7820658] (2016 IEEE 35th International Performance Computing and Communications Conference, IPCCC 2016). Institute of Electrical and Electronics Engineers Inc.. https://doi.org/10.1109/PCCC.2016.7820658
Yang, Zhengyu ; Tai, Jianzhe ; Bhimani, Janki ; Wang, Jiayin ; Mi, Ningfang ; Sheng, Bo. / GReM : Dynamic SSD resource allocation in virtualized storage systems with heterogeneous IO workloads. 2016 IEEE 35th International Performance Computing and Communications Conference, IPCCC 2016. Institute of Electrical and Electronics Engineers Inc., 2017. (2016 IEEE 35th International Performance Computing and Communications Conference, IPCCC 2016).
@inproceedings{9e5f4f08a86345d793e2b7eb585c6e14,
title = "GReM: Dynamic SSD resource allocation in virtualized storage systems with heterogeneous IO workloads",
abstract = "In a shared virtualized storage system that runs VMs with heterogeneous IO demands, it becomes a problem for the hypervisor to cost-effectively partition and allocate SSD resources among multiple VMs. There are two straightforward approaches to solving this problem: equally assigning SSDs to each VM or managing SSD resources in a fair competition mode. Unfortunately, neither of these approaches can fully utilize the benefits of SSD resources, particularly when the workloads frequently change and bursty IOs occur from time to time. In this paper, we design a Global SSD Resource Management solution - GReM, which aims to fully utilize SSD resources as a second-level cache under the consideration of performance isolation. In particular, GReM takes dynamic IO demands of all VMs into consideration to split the entire SSD space into a long-term zone and a short-term zone, and cost-effectively updates the content of SSDs in these two zones. GReM is able to adaptively adjust the reservation for each VM inside the long-term zone based on their IO changes. GReM can further dynamically partition SSDs between the long- and short-term zones during runtime by leveraging the feedbacks from both cache performance and bursty workloads. Experimental results show that GReM can capture the cross-VM IO changes to make correct decisions on resource allocation, and thus obtain high IO hit ratio and low IO management costs, compared with both traditional and state-of-the-art caching algorithms.",
keywords = "Bursty Detection, Caching Algorithms, I/O Workload Characterization, Resource Allocation, Solid State Drives, Virtualized Storage Systems",
author = "Zhengyu Yang and Jianzhe Tai and Janki Bhimani and Jiayin Wang and Ningfang Mi and Bo Sheng",
year = "2017",
month = "1",
day = "17",
doi = "10.1109/PCCC.2016.7820658",
language = "English",
series = "2016 IEEE 35th International Performance Computing and Communications Conference, IPCCC 2016",
publisher = "Institute of Electrical and Electronics Engineers Inc.",
booktitle = "2016 IEEE 35th International Performance Computing and Communications Conference, IPCCC 2016",

}

Yang, Z, Tai, J, Bhimani, J, Wang, J, Mi, N & Sheng, B 2017, GReM: Dynamic SSD resource allocation in virtualized storage systems with heterogeneous IO workloads. in 2016 IEEE 35th International Performance Computing and Communications Conference, IPCCC 2016., 7820658, 2016 IEEE 35th International Performance Computing and Communications Conference, IPCCC 2016, Institute of Electrical and Electronics Engineers Inc., 35th IEEE International Performance Computing and Communications Conference, IPCCC 2016, Las Vegas, United States, 9/12/16. https://doi.org/10.1109/PCCC.2016.7820658

GReM : Dynamic SSD resource allocation in virtualized storage systems with heterogeneous IO workloads. / Yang, Zhengyu; Tai, Jianzhe; Bhimani, Janki; Wang, Jiayin; Mi, Ningfang; Sheng, Bo.

2016 IEEE 35th International Performance Computing and Communications Conference, IPCCC 2016. Institute of Electrical and Electronics Engineers Inc., 2017. 7820658 (2016 IEEE 35th International Performance Computing and Communications Conference, IPCCC 2016).

Research output: Chapter in Book/Report/Conference proceedingConference contribution

TY - GEN

T1 - GReM

T2 - Dynamic SSD resource allocation in virtualized storage systems with heterogeneous IO workloads

AU - Yang, Zhengyu

AU - Tai, Jianzhe

AU - Bhimani, Janki

AU - Wang, Jiayin

AU - Mi, Ningfang

AU - Sheng, Bo

PY - 2017/1/17

Y1 - 2017/1/17

N2 - In a shared virtualized storage system that runs VMs with heterogeneous IO demands, it becomes a problem for the hypervisor to cost-effectively partition and allocate SSD resources among multiple VMs. There are two straightforward approaches to solving this problem: equally assigning SSDs to each VM or managing SSD resources in a fair competition mode. Unfortunately, neither of these approaches can fully utilize the benefits of SSD resources, particularly when the workloads frequently change and bursty IOs occur from time to time. In this paper, we design a Global SSD Resource Management solution - GReM, which aims to fully utilize SSD resources as a second-level cache under the consideration of performance isolation. In particular, GReM takes dynamic IO demands of all VMs into consideration to split the entire SSD space into a long-term zone and a short-term zone, and cost-effectively updates the content of SSDs in these two zones. GReM is able to adaptively adjust the reservation for each VM inside the long-term zone based on their IO changes. GReM can further dynamically partition SSDs between the long- and short-term zones during runtime by leveraging the feedbacks from both cache performance and bursty workloads. Experimental results show that GReM can capture the cross-VM IO changes to make correct decisions on resource allocation, and thus obtain high IO hit ratio and low IO management costs, compared with both traditional and state-of-the-art caching algorithms.

AB - In a shared virtualized storage system that runs VMs with heterogeneous IO demands, it becomes a problem for the hypervisor to cost-effectively partition and allocate SSD resources among multiple VMs. There are two straightforward approaches to solving this problem: equally assigning SSDs to each VM or managing SSD resources in a fair competition mode. Unfortunately, neither of these approaches can fully utilize the benefits of SSD resources, particularly when the workloads frequently change and bursty IOs occur from time to time. In this paper, we design a Global SSD Resource Management solution - GReM, which aims to fully utilize SSD resources as a second-level cache under the consideration of performance isolation. In particular, GReM takes dynamic IO demands of all VMs into consideration to split the entire SSD space into a long-term zone and a short-term zone, and cost-effectively updates the content of SSDs in these two zones. GReM is able to adaptively adjust the reservation for each VM inside the long-term zone based on their IO changes. GReM can further dynamically partition SSDs between the long- and short-term zones during runtime by leveraging the feedbacks from both cache performance and bursty workloads. Experimental results show that GReM can capture the cross-VM IO changes to make correct decisions on resource allocation, and thus obtain high IO hit ratio and low IO management costs, compared with both traditional and state-of-the-art caching algorithms.

KW - Bursty Detection

KW - Caching Algorithms

KW - I/O Workload Characterization

KW - Resource Allocation

KW - Solid State Drives

KW - Virtualized Storage Systems

UR - http://www.scopus.com/inward/record.url?scp=85013374591&partnerID=8YFLogxK

U2 - 10.1109/PCCC.2016.7820658

DO - 10.1109/PCCC.2016.7820658

M3 - Conference contribution

AN - SCOPUS:85013374591

T3 - 2016 IEEE 35th International Performance Computing and Communications Conference, IPCCC 2016

BT - 2016 IEEE 35th International Performance Computing and Communications Conference, IPCCC 2016

PB - Institute of Electrical and Electronics Engineers Inc.

ER -

Yang Z, Tai J, Bhimani J, Wang J, Mi N, Sheng B. GReM: Dynamic SSD resource allocation in virtualized storage systems with heterogeneous IO workloads. In 2016 IEEE 35th International Performance Computing and Communications Conference, IPCCC 2016. Institute of Electrical and Electronics Engineers Inc. 2017. 7820658. (2016 IEEE 35th International Performance Computing and Communications Conference, IPCCC 2016). https://doi.org/10.1109/PCCC.2016.7820658