46

Telco Clouds: Modelling and Simulation Jakub Krzywda 1, William Tärneberg 2, P-O Östberg 1, Maria Kihl 2, Erik Elmroth 1 1 – Umeå University 2 – Lund

Embed Size (px)

Citation preview

Page 1: Telco Clouds: Modelling and Simulation Jakub Krzywda 1, William Tärneberg 2, P-O Östberg 1, Maria Kihl 2, Erik Elmroth 1 1 – Umeå University 2 – Lund
Page 2: Telco Clouds: Modelling and Simulation Jakub Krzywda 1, William Tärneberg 2, P-O Östberg 1, Maria Kihl 2, Erik Elmroth 1 1 – Umeå University 2 – Lund

Telco Clouds:Modelling and Simulation

Jakub Krzywda1, William Tärneberg2,P-O Östberg1, Maria Kihl2, Erik Elmroth1

1 – Umeå University2 – Lund University

Page 3: Telco Clouds: Modelling and Simulation Jakub Krzywda 1, William Tärneberg 2, P-O Östberg 1, Maria Kihl 2, Erik Elmroth 1 1 – Umeå University 2 – Lund

3

Multi-Model

Page 4: Telco Clouds: Modelling and Simulation Jakub Krzywda 1, William Tärneberg 2, P-O Östberg 1, Maria Kihl 2, Erik Elmroth 1 1 – Umeå University 2 – Lund

4

Telecommunication Cloud computing

… that combines:

Page 5: Telco Clouds: Modelling and Simulation Jakub Krzywda 1, William Tärneberg 2, P-O Östberg 1, Maria Kihl 2, Erik Elmroth 1 1 – Umeå University 2 – Lund

5

Mobility Infrastructure Setup

… to simulate the influence of:

Page 6: Telco Clouds: Modelling and Simulation Jakub Krzywda 1, William Tärneberg 2, P-O Östberg 1, Maria Kihl 2, Erik Elmroth 1 1 – Umeå University 2 – Lund

6

Performance Costs

… on:

Page 7: Telco Clouds: Modelling and Simulation Jakub Krzywda 1, William Tärneberg 2, P-O Östberg 1, Maria Kihl 2, Erik Elmroth 1 1 – Umeå University 2 – Lund

Computation Offloading

Mobile Devices• high requirements on Quality of Service• limited battery lifetime• less powerful resources

7

Page 8: Telco Clouds: Modelling and Simulation Jakub Krzywda 1, William Tärneberg 2, P-O Östberg 1, Maria Kihl 2, Erik Elmroth 1 1 – Umeå University 2 – Lund

8

Computation Offloading

Page 9: Telco Clouds: Modelling and Simulation Jakub Krzywda 1, William Tärneberg 2, P-O Östberg 1, Maria Kihl 2, Erik Elmroth 1 1 – Umeå University 2 – Lund

9

Other motivations

Page 10: Telco Clouds: Modelling and Simulation Jakub Krzywda 1, William Tärneberg 2, P-O Östberg 1, Maria Kihl 2, Erik Elmroth 1 1 – Umeå University 2 – Lund

10

Telecommunication Infrastructure

Cloud Computing Infrastructure

Current Situation

Page 11: Telco Clouds: Modelling and Simulation Jakub Krzywda 1, William Tärneberg 2, P-O Östberg 1, Maria Kihl 2, Erik Elmroth 1 1 – Umeå University 2 – Lund

11

Latency

Issues

Network Congestion

Hardware costs

Page 12: Telco Clouds: Modelling and Simulation Jakub Krzywda 1, William Tärneberg 2, P-O Östberg 1, Maria Kihl 2, Erik Elmroth 1 1 – Umeå University 2 – Lund

IntroducingTelco Cloud

12

Page 13: Telco Clouds: Modelling and Simulation Jakub Krzywda 1, William Tärneberg 2, P-O Östberg 1, Maria Kihl 2, Erik Elmroth 1 1 – Umeå University 2 – Lund

13

Current SituationMerging Telecommunication and Cloud Infrastructures

Page 14: Telco Clouds: Modelling and Simulation Jakub Krzywda 1, William Tärneberg 2, P-O Östberg 1, Maria Kihl 2, Erik Elmroth 1 1 – Umeå University 2 – Lund

14

Introducing ProximalData Centers

Telco Cloud

Page 15: Telco Clouds: Modelling and Simulation Jakub Krzywda 1, William Tärneberg 2, P-O Östberg 1, Maria Kihl 2, Erik Elmroth 1 1 – Umeå University 2 – Lund

15

Reduce Latency

Benefits

Prevent Network Congestion

Decrease costs

Page 16: Telco Clouds: Modelling and Simulation Jakub Krzywda 1, William Tärneberg 2, P-O Östberg 1, Maria Kihl 2, Erik Elmroth 1 1 – Umeå University 2 – Lund

• Applications• Mobility

• Quality of Service• Costs

• Topology• DC Capacity• Service placement

16

Telco Cloud Dynamics

Workload

ObjectivesSetup

Page 17: Telco Clouds: Modelling and Simulation Jakub Krzywda 1, William Tärneberg 2, P-O Östberg 1, Maria Kihl 2, Erik Elmroth 1 1 – Umeå University 2 – Lund

Multi-Model

17

Page 18: Telco Clouds: Modelling and Simulation Jakub Krzywda 1, William Tärneberg 2, P-O Östberg 1, Maria Kihl 2, Erik Elmroth 1 1 – Umeå University 2 – Lund

Multi-Model Parameters

• Objectives– Performance Overhead

(e.g., migration cost)

18

Page 19: Telco Clouds: Modelling and Simulation Jakub Krzywda 1, William Tärneberg 2, P-O Östberg 1, Maria Kihl 2, Erik Elmroth 1 1 – Umeå University 2 – Lund

Multi-Model Parameters

• Objectives• Workload

– User’s State

19

Request

Response State1010

Stateful Applications

User’s State Size

Total Size ofUser’s Requests

Page 20: Telco Clouds: Modelling and Simulation Jakub Krzywda 1, William Tärneberg 2, P-O Östberg 1, Maria Kihl 2, Erik Elmroth 1 1 – Umeå University 2 – Lund

• Objectives• Workload

– User’s State– Mobility

Multi-Model Parameters

20

User’s State has to follow User’s movements User

changes PDC

Sta

te

Page 21: Telco Clouds: Modelling and Simulation Jakub Krzywda 1, William Tärneberg 2, P-O Östberg 1, Maria Kihl 2, Erik Elmroth 1 1 – Umeå University 2 – Lund

• Objectives• Workload

– User’s State– Mobility

• Setup– Data Center Catchment

Multi-Model Parameters

21

:

Ratio between

RBS and DC

Page 22: Telco Clouds: Modelling and Simulation Jakub Krzywda 1, William Tärneberg 2, P-O Östberg 1, Maria Kihl 2, Erik Elmroth 1 1 – Umeå University 2 – Lund

• Objectives• Workload

– User’s State– Mobility

• Setup– Data Center Catchment– Virtual Machine Life-Cycle

Multi-Model Parameters

22

Page 23: Telco Clouds: Modelling and Simulation Jakub Krzywda 1, William Tärneberg 2, P-O Östberg 1, Maria Kihl 2, Erik Elmroth 1 1 – Umeå University 2 – Lund

Virtual Machine Life-Cycle

23

Page 24: Telco Clouds: Modelling and Simulation Jakub Krzywda 1, William Tärneberg 2, P-O Östberg 1, Maria Kihl 2, Erik Elmroth 1 1 – Umeå University 2 – Lund

Simulation Showcase

24

Page 25: Telco Clouds: Modelling and Simulation Jakub Krzywda 1, William Tärneberg 2, P-O Östberg 1, Maria Kihl 2, Erik Elmroth 1 1 – Umeå University 2 – Lund

Simulation Showcase Setup

• 16 Radio Base Stations (4x4 layout)• Catchment of Data Centers varies

25

Page 26: Telco Clouds: Modelling and Simulation Jakub Krzywda 1, William Tärneberg 2, P-O Östberg 1, Maria Kihl 2, Erik Elmroth 1 1 – Umeå University 2 – Lund

Results

26

Static Mobile

Small Catchment Few Devices Inactivity

Bigger Catchment Longer Stay More to Migrate

Page 27: Telco Clouds: Modelling and Simulation Jakub Krzywda 1, William Tärneberg 2, P-O Östberg 1, Maria Kihl 2, Erik Elmroth 1 1 – Umeå University 2 – Lund

Summary

27

Page 28: Telco Clouds: Modelling and Simulation Jakub Krzywda 1, William Tärneberg 2, P-O Östberg 1, Maria Kihl 2, Erik Elmroth 1 1 – Umeå University 2 – Lund

28

Telco Cloud Dynamics

Workload

ObjectivesSetup

Page 29: Telco Clouds: Modelling and Simulation Jakub Krzywda 1, William Tärneberg 2, P-O Östberg 1, Maria Kihl 2, Erik Elmroth 1 1 – Umeå University 2 – Lund

29

Multi-Model

Page 30: Telco Clouds: Modelling and Simulation Jakub Krzywda 1, William Tärneberg 2, P-O Östberg 1, Maria Kihl 2, Erik Elmroth 1 1 – Umeå University 2 – Lund

Who can benefit?

• Telecommunication Operators and Equipment Developers– Model existing infrastructure and plan changes

• Researchers– Testing algorithms for resource management

• Mobile Applications Developers– Testing behavior of apps in Telco Cloud

environment

30

Page 31: Telco Clouds: Modelling and Simulation Jakub Krzywda 1, William Tärneberg 2, P-O Östberg 1, Maria Kihl 2, Erik Elmroth 1 1 – Umeå University 2 – Lund
Page 32: Telco Clouds: Modelling and Simulation Jakub Krzywda 1, William Tärneberg 2, P-O Östberg 1, Maria Kihl 2, Erik Elmroth 1 1 – Umeå University 2 – Lund

Limitations

• Homogeneous requests (at application level)– All requests contribute to User’s State

• Simplified mobile access network model– Tangent, non-overlapping, square cells– No physical layer, channel provisioning and cell

load balancing

32

Page 33: Telco Clouds: Modelling and Simulation Jakub Krzywda 1, William Tärneberg 2, P-O Östberg 1, Maria Kihl 2, Erik Elmroth 1 1 – Umeå University 2 – Lund

Simulation Showcase Setup

• 16 Radio Base Stations (4x4 layout)• Catchment of Data Centers varies• Constant capacity of Data Centers• #VMs scales with the catchment• workload is balanced between VMs• 480 Mobile Devices

33

Page 34: Telco Clouds: Modelling and Simulation Jakub Krzywda 1, William Tärneberg 2, P-O Östberg 1, Maria Kihl 2, Erik Elmroth 1 1 – Umeå University 2 – Lund

Simulation requirements

• Scale (number of)– Mobile Devices – hundreds/thousands– Radio Base Stations – tens– Data Centers – several

• Length and granularity of simulation (time) – Minutes/hours for mobility patterns– Milliseconds for latency

34

Page 35: Telco Clouds: Modelling and Simulation Jakub Krzywda 1, William Tärneberg 2, P-O Östberg 1, Maria Kihl 2, Erik Elmroth 1 1 – Umeå University 2 – Lund

Why simulation?

• No existing Telco Cloud infrastructure

• Extremely high costs of building infrastructure for testing purposes

• Huge number of possible setups

35

Page 36: Telco Clouds: Modelling and Simulation Jakub Krzywda 1, William Tärneberg 2, P-O Östberg 1, Maria Kihl 2, Erik Elmroth 1 1 – Umeå University 2 – Lund

When to Offload?

36

Kumar, K.; Yung-Hsiang Lu, "Cloud Computing for Mobile Users: Can Offloading Computation Save Energy?," Computer , vol.43, no.4, pp.51,56, April 2010

Page 37: Telco Clouds: Modelling and Simulation Jakub Krzywda 1, William Tärneberg 2, P-O Östberg 1, Maria Kihl 2, Erik Elmroth 1 1 – Umeå University 2 – Lund

Multi-Model

37

Page 38: Telco Clouds: Modelling and Simulation Jakub Krzywda 1, William Tärneberg 2, P-O Östberg 1, Maria Kihl 2, Erik Elmroth 1 1 – Umeå University 2 – Lund

Multi-Model Parameters

• Workload– Request Generation

1010

38

Page 39: Telco Clouds: Modelling and Simulation Jakub Krzywda 1, William Tärneberg 2, P-O Östberg 1, Maria Kihl 2, Erik Elmroth 1 1 – Umeå University 2 – Lund

Multi-Model Parameters

1010

• Workload– Request Generation– Resource

Requirements

39

Page 40: Telco Clouds: Modelling and Simulation Jakub Krzywda 1, William Tärneberg 2, P-O Östberg 1, Maria Kihl 2, Erik Elmroth 1 1 – Umeå University 2 – Lund

• Workload– Request Generation– Resource Requirements– Mobility

Multi-Model Parameters

40

Page 41: Telco Clouds: Modelling and Simulation Jakub Krzywda 1, William Tärneberg 2, P-O Östberg 1, Maria Kihl 2, Erik Elmroth 1 1 – Umeå University 2 – Lund

Multi-Model Parameters

• Workload• Setup

– Mobile Network

41

Page 42: Telco Clouds: Modelling and Simulation Jakub Krzywda 1, William Tärneberg 2, P-O Östberg 1, Maria Kihl 2, Erik Elmroth 1 1 – Umeå University 2 – Lund

Multi-Model Parameters

• Workload• Setup

– Mobile Network– Internet

42

Page 43: Telco Clouds: Modelling and Simulation Jakub Krzywda 1, William Tärneberg 2, P-O Östberg 1, Maria Kihl 2, Erik Elmroth 1 1 – Umeå University 2 – Lund

• Workload• Setup

– Mobile Network– Internet– Data Center

Multi-Model Parameters

43

Page 44: Telco Clouds: Modelling and Simulation Jakub Krzywda 1, William Tärneberg 2, P-O Östberg 1, Maria Kihl 2, Erik Elmroth 1 1 – Umeå University 2 – Lund

• Workload• Setup

– Mobile Network– Internet– Data Center– Service Placement

Multi-Model Parameters

?

?

44

Page 45: Telco Clouds: Modelling and Simulation Jakub Krzywda 1, William Tärneberg 2, P-O Östberg 1, Maria Kihl 2, Erik Elmroth 1 1 – Umeå University 2 – Lund

Multi-Model Parameters

• Workload• Setup• Objectives

– Quality of Service

45

Page 46: Telco Clouds: Modelling and Simulation Jakub Krzywda 1, William Tärneberg 2, P-O Östberg 1, Maria Kihl 2, Erik Elmroth 1 1 – Umeå University 2 – Lund

Multi-Model Parameters

• Workload• Setup• Objectives

– Quality of Service– Costs

46