System availability

System availability

System Availability

• The percentage of time that the system can perform its intended function

System Availability = 1 – System Downtime


Downtime per year

Availability	Nines	Downtime
90%	1	36.5	days/year
99%	2	3.65	DAys/year
99.9%	3	8.78	hours/year
99.99%	4	52	minutes/year
99.999%	5	5	minutes/year

System Downtime

Many events causes system downtime:

• HW fault
• Software fault
• Vandalism
• Extreme conditions (fire, flooding etc)
• Power outage
• IP network failure
• Planned system maintenance

System Downtime = ∑ P * S * MTTR

P	= Probability of event taken place
S	= Severity of event
	= Percentage of service affected by fault
MTTR	= Mean Time To Repair
	= mean time to detect fault + mean time to fix fault

HW failure

MTBF

• Probability of HW faults calculated using MTBF figures
• MTBF ≠ System Availability

MTBF calculations

• Emperical method
• MIL-HDBK-217
• Telcordia

Emperical methods

• Based on statistics from the field

MIL-HDBK-217 and Telcordia

• All component entered in database with set environmental condition
• Provides usually lower MTBF figure than emperical methods
  • Does include real usage conditions
  • Use worst case environmental conditions

*More components gives higher MTBF
*MTBF and single points of failure

Other failures

Software fault

• Automatic watch dog functions
• Automatic recovery functions
• Maturity of system
• Structured software design and test

Vandalism and Extreme conditions

• Robustness to vandalism and extreme conditions
• IP and IK class
• IP security functions to hinder denial of service attacks (DOS)

Power outage

• UPS and redundant power supplier

IP network failure

• Network service level
• Redundant and switchover functions

Planned system maintenance

• Expansion, add users etc
• Ability to do maintenance without service interruptions

Redundancy

Redundancy is about parallelism and removing single point of failures

Redundancy usually gives lower MTBF figures

• Require more components

Redundancy usually provides significant higher service availability

• A single failure shall have minimum or no impact on service availability