Measuring Availability in Telecommunications NetworksMattias Thulin, November 2004

Disposition

1. Introduction2. Method3. Network Availability4. SDH Network description5. ITU-T standard G.8266. Analysis7. Implementation8. Result9. Conclusions

2

Song Networks

Nordic network provider Optical fiber network covering

Northern Europe

Products: IP-VPN Internet connections Telephone services Hosting Carrier services

Introduction

Market demand for network quality

Important to measure network availabilityMaintain service-level agreementsAttract new customers Indicator of network quality for internal maintenance

Methods of measuring and defining Network Availability vary between operators

Purpose

How is Network Availability defined? How can it be measured? Why should it be measured? What standards exist? Are there any recommended values for availability parameters? How can availability measurements be applied to Song

Networks SDH transmission network?

Delimitations

1) General study on Network Availability2) Develop a method for availability measurement

Nortel SDH equipment Four rings 44 links

Oriented towards network-operation

Method

Literature study Network study Monitoring system Preside Interviews Standards Design model for availability measurement and presentation

Network Availability - definition

The ability of a functional unit to be in a state to perform a required function under given conditions at a given instant of time or over a given time interval, assuming that the required external resources are provided.

ISO 2382-14, 1997

Network Availability – The “five-nines”

Percentage value of uptime for a given time period

“Five-nines” 99,999%

Viewed as desired uptime in network core-level

Availability Downtime per year

99,9999% 32s

99,999% 5min 15s

99,99% 52min 36s

99,9% 8h 46min

99% 3 days 15h 40min

Theoretic Availability

99,99% 99,99% 99,99%

9996,09999,0 4 Total availability =

Summing availabilitySumming availabilitySerial units

Parallel units

A B

A

B

Total availability = A * B

Total availability = A + B - A * B

Reactive Availability

Data from trouble-tickets

Good for measuring customer-experienced availability

Easy to identify what equipment failed and what solved the error

Can lack information of short interruptions and outside of office hours

Customer- vs. Network-management oriented

Important to know for whom or for what purpose are we measuring

Customer oriented Includes all layersCalculate downtime when the customer connection is not

working.

Network-management orientedWhat links have lower availability?Considered as downtime although the traffic is rerouted

SDH Network Description

SDH – Synchronous Digital Hierarchy

Based on American standard SONET

Normally build in ring structure

Error correction and retransmission is done by overlaying protocols

Song Networks’ SDH network

Baltic ringBaltic ring

Nordic ringNordic ring Europe ringEurope ringSweden ringSweden ring

Song Networks’ SDH network

Sweden ring: 9 Network ElementsNordic ring: 7 Network ElementsEuropean ring: 7 Network ElementsBaltic ring: 3 Network Elements

6050 6052

6054 6056

G17 G17

G17G17

G18 G18

G18G18

G11G12

Surveillance and statistics

OPC

NENE NE

NE

Preside

NENE

Preside

Global performance Alarm lists Query performance

statistics

Preside log files

Comma-delimited text files (CSV) One file per Network Element 96 15-min counts (past 24 hours) 8 24-hour counts (past week)

1200,Ottawa,OC48,Term,DS3,G7,2,Line,Rx,Ne,SES,03/07/99,03/07/99,16:00,0,0,0,2,3,8,12,6,0,0,1,0,0,0,0,1,0,0,0,2,0,0,0,1,0,0,1,0,0,0,0,0,0,0,0,0,0,0,0,0

ITU-T Standard G.826

Errored Block (EB)1 bit-error/block

Errored Second (ES)1 sec with 1 EB

Severly Errored Second(SES)

1 sec with 30% EB

Bit-error

Parameters:

Unavailable Second (UAS)

T1313780-98

10 s 10 s<10 s

Time

Unavailability detected Availability detected

Unavailable period Available period

Severely Errored Second

Errored Second (non-SES)

Error-free Second

A period of unavailable time begins at the onset of ten consecutive SES events. These ten seconds are considered to be part of unavailable time. A new period of available time begins at the onset of ten consecutive non-SES events. These ten seconds are considered to be part of available time.(ITU-T G.826, 2002)

Analysis

Define availability

Develop model for calculating average availability

Define database structure for saving availability statistics

Specify format for availability reports

Analysis

Follow ITU-T Standard G.826 Apply to all active links in the network Calculate average availability per link, per ring and total network Present first five significant figures First calculate average UAS, convert to percentage in last step to avoid

rounding error

100*meMeasuredTi

UASmeMeasuredTityAvailabili

Analysis

+ +

3

Availability for a ring is the average UAS for all the links in the ring

Implementation

DatabaseLog Files Parser Analyze Report

Implementation - parser program

Programmed in Java for platform independence Parse all log-files in directory for:

NELinkDayUAS count

Insert into MySQL database table

Implementation – report generating

Web interface for easy access Input parameters: start and end

date PHP-script query database for

UAS values and calculate average availability Per link Per ring Total network

Report can be saved to PDF format (PHP-script)

Implementation – Graphic reports

Crystal Reports

Start-date and end-date are entered and the program queries database and produces graphic reports

Can be exported to PDF file

Result

Between 2004-07-12 and 2004-09-19

SDH Ring Availability in %

Sweden DX 99.983

Nordic DX 99.997

Europe DX 99.462

Baltic DX 99.814

TOTAL 99.81

NE Link Availability in %

6052 G11 99.989

6052 G12 99.989

6052 G17 99.944

6052 G18 99.958

6053 G17 99.856

6053 G18 99.959

6054 G17 100

6054 G18 99.95

6056 G11 99.989

6056 G12 100

6056 G17 99.951

6056 G18 100

6060 G11 100

6060 G12 100

6060 G17 100

6060 G18 99.993

6058 G17 99.993

6058 G18 100

6044 G11 100

6044 G12 99.989

6044 G17 100

6044 G18 100

6045 G17 100

6045 G18 100

6046 G17 100

6046 G18 100

ResultBetween 2004-07-12 and 2004-09-19

ResultBetween 2004-07-12 and 2004-09-19

Conclusions

Background study can be used for planning future measurements

Positive feedback from network operations management for the weekly reports

Need more statistic in the database to observe general trends By studying trends Song Networks can cut maintenance

spending and better forecast future cost by directing resources to maintain a high network quality

Future work: Measure backbone availability from a customer point of view

using relational databasesHow do errors in the backbone affect distribution layer?

Questions?Questions?