21
© Nokia 2016 1 <Change information classification in footer> Impact of Virtualization on Telecom Network Reliability Dr Olli Salmela, IEEE Senior Member IEEE - Emerging Technology Reliability Roundtable 2019

Impact of Virtualization on Telecom Network Reliabilitycqr.committees.comsoc.org/files/2019/09/06-Olli_Salmela_ETR2019_… · Servers HW-wise are alike. However, once virtualization

  • Upload
    others

  • View
    2

  • Download
    0

Embed Size (px)

Citation preview

Page 1: Impact of Virtualization on Telecom Network Reliabilitycqr.committees.comsoc.org/files/2019/09/06-Olli_Salmela_ETR2019_… · Servers HW-wise are alike. However, once virtualization

© Nokia 20161 <Change information classification in footer>

Impact of Virtualization on

Telecom Network Reliability

Dr Olli Salmela, IEEE Senior Member

IEEE - Emerging Technology Reliability Roundtable 2019

Page 2: Impact of Virtualization on Telecom Network Reliabilitycqr.committees.comsoc.org/files/2019/09/06-Olli_Salmela_ETR2019_… · Servers HW-wise are alike. However, once virtualization

© Nokia 20162

Networks main physical product is the radio base stationSome examples of real installations

Page 3: Impact of Virtualization on Telecom Network Reliabilitycqr.committees.comsoc.org/files/2019/09/06-Olli_Salmela_ETR2019_… · Servers HW-wise are alike. However, once virtualization

© Nokia 20163

Past Today Future Option

LNA

System Module

RF Module or RRH

System Module

Antenna

Feeder cable Optical link

ActiveAntenna

1 1/2 1/3Relative power consumption

We have significantly reduced the size and power consumption of base

station

Page 4: Impact of Virtualization on Telecom Network Reliabilitycqr.committees.comsoc.org/files/2019/09/06-Olli_Salmela_ETR2019_… · Servers HW-wise are alike. However, once virtualization

© Nokia 20164

And are taking the next step in miniaturization

Phone for scale

Page 5: Impact of Virtualization on Telecom Network Reliabilitycqr.committees.comsoc.org/files/2019/09/06-Olli_Salmela_ETR2019_… · Servers HW-wise are alike. However, once virtualization

© Nokia 20165

Base stations need to operate in harsh environments

53 C

Page 6: Impact of Virtualization on Telecom Network Reliabilitycqr.committees.comsoc.org/files/2019/09/06-Olli_Salmela_ETR2019_… · Servers HW-wise are alike. However, once virtualization

© Nokia 20166

Reliability of telecommunications equipment

Most network equipment has a lifetime expectancy of 10 to 20 years. Conversely, consumer products like mobile phones typically have lifetime requirements of 3 to 5 years.

Networks have “fine nines” service availability target

• In order to fulfill the requirements, one must guarantee that:- Components will survive for the required lifetime- Units do not fail under specified operating conditions- Components are available for the lifetime period

Page 7: Impact of Virtualization on Telecom Network Reliabilitycqr.committees.comsoc.org/files/2019/09/06-Olli_Salmela_ETR2019_… · Servers HW-wise are alike. However, once virtualization

© Nokia 20167

• System integration has played a significant role in improving telecom system performance and reliability and enabled miniaturization

– Integration means less interconnectionsand interfaces, which improves reliability

– Quality of HW components hassignificantly improved

• Needs for capacity, performance and reliability continue to increase

• A radical step is needed in miniaturization for small radio access points

Incremental Improvements Continue

Page 8: Impact of Virtualization on Telecom Network Reliabilitycqr.committees.comsoc.org/files/2019/09/06-Olli_Salmela_ETR2019_… · Servers HW-wise are alike. However, once virtualization

© Nokia 20168

New needs for reliability

<Change information classification in footer>

• Many of the services that rely on 5G include self-driving vehicles, health services, traffic

orchestration, energy power-grid management and other services directly responsible for

public safety

• New technologies implemented to meet these requirements are based on:

•Application-aware network traffic slicing

•Highly available and redundant network topologies

•Intelligent SDN routing

•Multiple cell towers providing spatial and frequency redundancy

•Network protocol extensions with near-instantaneous packet failover

•Advancements to carrier core packet processing systems

• MECs needed to meet the latency requirements.

Page 9: Impact of Virtualization on Telecom Network Reliabilitycqr.committees.comsoc.org/files/2019/09/06-Olli_Salmela_ETR2019_… · Servers HW-wise are alike. However, once virtualization

9 © Nokia Solutions and Networks 2014

• Virtualization. Servers HW-wise are alike. However, once virtualization is implemented,

servers have a fixed functionality related to a certain, related underlying NE.

• Cloudification. The NE SW can be run on any of the servers. Servers are not anymore

dedicated to a certain NE functionality.

• NFV, Network Function Virtualization

• IaaS, Infrastructure as a Service

• PaaS, Platform as a Service

• SaaS, SW as a Service

• SDN, SW Defined Networking

• From physical to service availability

Telco Cloud, Basic Concepts

<Change information classification in footer>

Page 10: Impact of Virtualization on Telecom Network Reliabilitycqr.committees.comsoc.org/files/2019/09/06-Olli_Salmela_ETR2019_… · Servers HW-wise are alike. However, once virtualization

© Nokia 201610

Switches RacksStorageServers

DC ServicesAcceleratorsMobile DC/CDCHyperscale

Confidential

Nokia AirFrame Data Center Solution | Building Blocks

Page 11: Impact of Virtualization on Telecom Network Reliabilitycqr.committees.comsoc.org/files/2019/09/06-Olli_Salmela_ETR2019_… · Servers HW-wise are alike. However, once virtualization

11 © Nokia Solutions and Networks 2014

• More processing capacity

- Risk of congestion is reduced

- Off-loading capabilities further reduce the risk of congestion

• More data storage available

- New analytic capabilities for NFVs

- Capability to provide big data analytics as a service

• Virtualized network elements have an inherent built-in redundancy

- Network Element functionality not anymore fixed to a certain HW

- SW Defined Networking Controller re-routes data plane if needed

- Cost still sets limits whether redundant VNFs can be utilized

Impact of virtualization on reliability

<Change information classification in footer>

Page 12: Impact of Virtualization on Telecom Network Reliabilitycqr.committees.comsoc.org/files/2019/09/06-Olli_Salmela_ETR2019_… · Servers HW-wise are alike. However, once virtualization

12 © Nokia Solutions and Networks 2014

• Object-oriented design, modularity, re-use

• Decoupling of components

• Concise interfaces

• Pieces of Open Source SW (OSS) and 3rd party SW may be used

• Test automation is the key to manage complexity

• SW testing will adapt some new approaches, like Simian army

• Unintrusive fault detection

• Recovery, reconfiguration and restart and proper system closure

• The frequency of SW updates is likely to be much higher

Telco Cloud SW

Page 13: Impact of Virtualization on Telecom Network Reliabilitycqr.committees.comsoc.org/files/2019/09/06-Olli_Salmela_ETR2019_… · Servers HW-wise are alike. However, once virtualization

16/08/201913 © Nokia 2014 - File Name - Version - Creator - DocID

Confidential

• Time-domain approach is the most popular.

- Based on curve fitting of observed failure data. Tries to fit the data to some statistical models.

- Used both in modeling and prediction. SW Reliability Growth Models (SRGM).

• Data domain approach uses ”run” as a unit of exposure as opposed to time-domain approach that uses a continuous time as a unit.

• Error seeding and tagging approach is the act of inserting errors into SW in order to estimate the total number of inherent errors in the SW.

- ”Simian army” approach

SW Maturity Prediction Techniques

Page 14: Impact of Virtualization on Telecom Network Reliabilitycqr.committees.comsoc.org/files/2019/09/06-Olli_Salmela_ETR2019_… · Servers HW-wise are alike. However, once virtualization

14 © Nokia Solutions and Networks 2014

• Although adapting ”IT standard” equipment, the telco-specific requirements still need

to be fulfilled.

• Some telco-specific requirements:

1. High availability

2. Small latency

3. Ultra-high performance, especially in user plane

4. Handling configuration management complexity

• ETSI NFV work group has produced documents on reliability (documents which use the

acronym NFV-REL, standing for “NFV Reliability and Availability”), security (using the

acronym NFV-SEC) and NFV evolution and its ecosystem (documents using the NFV-

EVE, standing for “NFV Evolution and Ecosystem.

Telco-Specific Requirements in IT Cloud Environment

<Change information classification in footer>

Page 15: Impact of Virtualization on Telecom Network Reliabilitycqr.committees.comsoc.org/files/2019/09/06-Olli_Salmela_ETR2019_… · Servers HW-wise are alike. However, once virtualization

15 © Nokia Solutions and Networks 2014

ETSI Service Availability Classification

<Change information classification in footer>

Page 16: Impact of Virtualization on Telecom Network Reliabilitycqr.committees.comsoc.org/files/2019/09/06-Olli_Salmela_ETR2019_… · Servers HW-wise are alike. However, once virtualization

16/08/201916 © Nokia 2014 - File Name - Version - Creator - DocID

Confidential

• Fault trees

- Impact of SW failures on a system level can be estimated

- Comparison of different designs

• Reliability block diagrams

- Widely used when estimating HW reliability.

- Modified version applicable also in case of SW reliability.

• Design Failure Mode and Effect Analysis (DFMEA)

Traditional SW Reliability Modeling Methods

Page 17: Impact of Virtualization on Telecom Network Reliabilitycqr.committees.comsoc.org/files/2019/09/06-Olli_Salmela_ETR2019_… · Servers HW-wise are alike. However, once virtualization

17 © Nokia Solutions and Networks 2014

• Functionality no more permanently linked to a certain HW

• Prediction methodologies likely to change from HW related, like RBD, to state-

space based and/or behavioral modeling due to virtualization

Change of Reliability Prediction Methodologies

<Change information classification in footer>

Reliability Block Diagram (RBD) Markov chain analysis

Page 18: Impact of Virtualization on Telecom Network Reliabilitycqr.committees.comsoc.org/files/2019/09/06-Olli_Salmela_ETR2019_… · Servers HW-wise are alike. However, once virtualization

16/08/201918 © Nokia 2014 - File Name - Version - Creator - DocID

Confidential

From Block Diagrams to More Advanced Methodologies

Page 19: Impact of Virtualization on Telecom Network Reliabilitycqr.committees.comsoc.org/files/2019/09/06-Olli_Salmela_ETR2019_… · Servers HW-wise are alike. However, once virtualization

© Nokia 201619

• A lot of symptom data can be stored and analyzed. – Predictive care approach enables early indication of latent or forthcoming failures.

More over, it even gives a possibility to fully prevent a failure from taking place.

– Possibilities to collect and analyze locally (edge analytics) and globally (collaborative data analysis)

• Other data, like geographical, environmental, and business data

can also be included in the analysis.– Possibility to analyze the impact of various factors on reliability, find patterns and to suggest improvements.

– Process modelling

• The big data analytics capability can also be used to analyze other equipment and to improveservices.

– Large business potential to supply data analytics as a service. Operators are already using Big Data to improve their ownbusiness.

– For example, Machine-to-Machine (M2M) communication can benefit from big data analytics.

• Visualization techniques include e.g. Self-Organizing Maps (SOM) a.k.a. Kohonen maps.- Requires understanding of NE behavior especially in the learning phase

Big Data Analytics for Reliability Improvements

Page 20: Impact of Virtualization on Telecom Network Reliabilitycqr.committees.comsoc.org/files/2019/09/06-Olli_Salmela_ETR2019_… · Servers HW-wise are alike. However, once virtualization

20 © Nokia Solutions and Networks 2014

• Both 5G and virtualization are setting some new challenges to reliability engineers

working in the telecom industry.

- Mission-critical use cases increase

- Higher frequencies and higher power levels

- Focus from HW to SW

- From simple reliability block diagrams to deep understanding of functionality

- OSS and 3rd party SW

• The good news

- More controlled use environment

- Possibility to increase redundance w/o any significant HW cost

- IoT creates a huge amount of data that can help in improving reliability

- Data analytics help to prevent and analyze failures

Conclusions

<Change information classification in footer>

Page 21: Impact of Virtualization on Telecom Network Reliabilitycqr.committees.comsoc.org/files/2019/09/06-Olli_Salmela_ETR2019_… · Servers HW-wise are alike. However, once virtualization