35
improving the Scalability of Data Center Networks with Traffic-aware Virtual Machine Placement B99705021 資資資 資資資 http://ppt.cc/41rH

improving the Scalability of Data Center Networks with Traffic-aware Virtual Machine Placement

  • Upload
    bran

  • View
    43

  • Download
    1

Embed Size (px)

DESCRIPTION

B99705021 資管三 李奕德 http://ppt.cc/41rH. improving the Scalability of Data Center Networks with Traffic-aware Virtual Machine Placement. Outline . Introduction Background Virtual machine placement Algorithm Algorithm evaluation Result Discussion and future work. introduction. - PowerPoint PPT Presentation

Citation preview

Page 1: improving the Scalability of  Data Center Networks  with Traffic-aware  Virtual Machine Placement

improving the Scalability of Data Center Networks with Traffic-aware Virtual Machine Placement

B99705021 資管三 李奕德http://ppt.cc/41rH

Page 2: improving the Scalability of  Data Center Networks  with Traffic-aware  Virtual Machine Placement

Outline

Introduction Background Virtual machine placement Algorithm Algorithm evaluation Result Discussion and future work

Page 3: improving the Scalability of  Data Center Networks  with Traffic-aware  Virtual Machine Placement

introduction

Scalability issue Aim to solve different problem

- Dcell, Bcube, PortLand, VL2…… No thinking of traffic issue - high traffic from end to end

Page 4: improving the Scalability of  Data Center Networks  with Traffic-aware  Virtual Machine Placement

introduction

three character of all traffic1. average pairwise traffic rate & end-to-end

cost has low correlation2. Uneven between VMs3. Stays almost the same Traffic-aware placement may be beneficial

Page 5: improving the Scalability of  Data Center Networks  with Traffic-aware  Virtual Machine Placement

introduction

Traffic-aware VM Placement Problem (TVMPP)

given: traffic matrix , cost matrix Goal: minimize cost Cost can be: Total switch used/Compute Time An algorithm that solve the NP-hard problem Architecture difference

Page 6: improving the Scalability of  Data Center Networks  with Traffic-aware  Virtual Machine Placement

NP- hard

NP: by nondeterministic algorithms in polynomial time

nondeterministic -Every “guess by hunch” is right

at least as hard as the hardest problems in NP

Page 7: improving the Scalability of  Data Center Networks  with Traffic-aware  Virtual Machine Placement

Outline

Introduction Background Virtual machine placement Algorithm Algorithm evaluation Result Discussion and future work

Page 8: improving the Scalability of  Data Center Networks  with Traffic-aware  Virtual Machine Placement

Background – traffic analysis

Data set I : IBM Global Services’ data warehouse About 17000 virtual machines Data set II: Server cluster About Hundreds of virtual machines round-trip latency measurement at 68 VM

Page 9: improving the Scalability of  Data Center Networks  with Traffic-aware  Virtual Machine Placement

Background- traffic analysis

Uneven between VMs

80% of VM’s traffic < 800kb/sec 4% of VM’s traffic > 8mb/sec

Page 10: improving the Scalability of  Data Center Networks  with Traffic-aware  Virtual Machine Placement

Background- traffic analysis

Stays almost the same

Page 11: improving the Scalability of  Data Center Networks  with Traffic-aware  Virtual Machine Placement

Background- traffic analysis

Low correlation between average pairwise traffic rate & end-to-end cost

Correlation : -0.32

Page 12: improving the Scalability of  Data Center Networks  with Traffic-aware  Virtual Machine Placement

Background - Achitecture

Old style

Page 13: improving the Scalability of  Data Center Networks  with Traffic-aware  Virtual Machine Placement

Background - Achitecture

VL2

Page 14: improving the Scalability of  Data Center Networks  with Traffic-aware  Virtual Machine Placement

Background - Achitecture

Portland

Bcube

Page 15: improving the Scalability of  Data Center Networks  with Traffic-aware  Virtual Machine Placement

Outline

Introduction Background Virtual machine placement Algorithm Algorithm evaluation Result Discussion and future work

Page 16: improving the Scalability of  Data Center Networks  with Traffic-aware  Virtual Machine Placement

Virtual machine placement- cost function

n VM to assign n slot for VM static and single-path routing Cost and traffic matrix from historical data

Page 17: improving the Scalability of  Data Center Networks  with Traffic-aware  Virtual Machine Placement

Virtual machine placement- cost function

is equivalent of finding

Dummy VM is assigned when no. slot > no. VM

ini

inji

jiij geCDCost

,...,1,...,1,

TTTT

XgeXXCDXtr

min

Page 18: improving the Scalability of  Data Center Networks  with Traffic-aware  Virtual Machine Placement

Virtual machine placement- complexity

Quadratic Assignment Problem (NP-hard) Impossible to find optimality when size > 15 TVMPP is a special case of QAP reduction from Balanced Minimum K-cut

Problem (BMKP) BMKP: extended problem from the Minimum

Bisection Problem (MBP) BMKP & MBP are NP-hard

Page 19: improving the Scalability of  Data Center Networks  with Traffic-aware  Virtual Machine Placement

Outline

Introduction Background Virtual machine placement Algorithm Algorithm evaluation Result Discussion and future work

Page 20: improving the Scalability of  Data Center Networks  with Traffic-aware  Virtual Machine Placement

Algorithm

approximation algorithm Cluster-and-Cut Divide VM into VM cluster Divide slot into slot cluster Put VM cluster into slot cluster A smaller problem Feasible when size is sufficient small

Page 21: improving the Scalability of  Data Center Networks  with Traffic-aware  Virtual Machine Placement

Algorithm – pseudo code

Page 22: improving the Scalability of  Data Center Networks  with Traffic-aware  Virtual Machine Placement

Algorithm – pseudo code

Page 23: improving the Scalability of  Data Center Networks  with Traffic-aware  Virtual Machine Placement

Algorithm - complexity

Complexity determine by SlotClustering and VMMinKcut

Slotclustering: O(nk) VMMinKcut: O(n4) Total complexity = O(n4)

Page 24: improving the Scalability of  Data Center Networks  with Traffic-aware  Virtual Machine Placement

Outline

Introduction Background Virtual machine placement Algorithm Algorithm evaluation Result Discussion and future work

Page 25: improving the Scalability of  Data Center Networks  with Traffic-aware  Virtual Machine Placement

Algorithm evaluation- cluster and cut

Cluster and cut VS. other benchmark algorithms

Local Optimal Pairwise Interchange (LOPI) Simulated Annealing (SA)

hybrid traffic model Gravity model compute the GLB for each settings

Page 26: improving the Scalability of  Data Center Networks  with Traffic-aware  Virtual Machine Placement

Algorithm evaluation - result

Page 27: improving the Scalability of  Data Center Networks  with Traffic-aware  Virtual Machine Placement

Outline

Introduction Background Virtual machine placement Algorithm Algorithm evaluation Result Discussion and future work

Page 28: improving the Scalability of  Data Center Networks  with Traffic-aware  Virtual Machine Placement

Result

Cost matrix

Compare with random assign

Page 29: improving the Scalability of  Data Center Networks  with Traffic-aware  Virtual Machine Placement

Result

Traffic is assumed to be in normal distribution Variance is change to show difference

Different architecture & variance affect result

Page 30: improving the Scalability of  Data Center Networks  with Traffic-aware  Virtual Machine Placement

Result

View as VM cluster GLB prediction

Page 31: improving the Scalability of  Data Center Networks  with Traffic-aware  Virtual Machine Placement

Result

GLB prediction VS. optimal solution

Page 32: improving the Scalability of  Data Center Networks  with Traffic-aware  Virtual Machine Placement

conclusion

Thing that brings better performance: - bigger variance - smaller cluster (less VM in a group) - Architecture difference (generally) Bcube > tree > fat-tree > VL2 Good scenario: multiple service in a data

center Bad scenario: single service / map-reduce

Page 33: improving the Scalability of  Data Center Networks  with Traffic-aware  Virtual Machine Placement

Outline

Introduction Background Virtual machine placement Algorithm Algorithm evaluation Result Discussion and future work

Page 34: improving the Scalability of  Data Center Networks  with Traffic-aware  Virtual Machine Placement

Discussion and future

Dynamic VM placement Other VM placement with different goal

Page 35: improving the Scalability of  Data Center Networks  with Traffic-aware  Virtual Machine Placement

Q&AThank you for your attention