measuring availability in telecommunications networks mattias thulin, november 2004
Post on 15-Mar-2016
43 Views
Preview:
DESCRIPTION
TRANSCRIPT
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?
top related