For example, assume an IT Service having four components with below availability percentages.
In the above case Availability is calculated as:
Availability = Host * Network * Server * Workstation
Calculation = 0.98 * 0.98 * 0.975 * 0.96 = 0.8989
Total Service Availability = 89.89%.
My question is: How do we calculate availability for entire IT Infrastructure (with many IT Services) or for the entire IT Service organisation? Can I use the above formula and arrive at total availability?
For example, for five IT Services assuming availability percentages as below.
Joined: Jan 12, 2007 Posts: 48 Location: Warsaw, Poland
Posted: Wed Apr 18, 2007 6:40 pm Post subject:
What is the purpose of your calculation? Normally you should calculate the availability for every service separately as every service is pointed separately in SLA.
If you want to combine the overall availability you would probably need to evaluate importance of services and add it to calculation.
For example - if the importance factor is of range 1 to 10 you can evaluate the services like this:
As for the total Service Availability, you can calculate it this way only if the elements depend on each other in serial configuration. If you have different architecture, then you may use different calculation (and get better results ). _________________ Krzysztof (Chris) Baczkiewicz
IT Standards Support
In my mind, availability is measrued per service, taking into accounts all the components of the service. Mixing services together does make sense to me.
I am a little bit skeptical about your calculation: it may give a distorsed vision of availability.
Ex: yesterday there was quite a major power outage in the datacenter that led everything to fail. Everything needed to be restarted. The network was back in 30 minutes, however some applications did not become fully available before 1 hour.
If you calculate network availability and application availability separately and factor the 2 you get a wrong picture. Let's say the normal operation window is 18 hours a day (on which the availability is measured).
network availability for that day = 100% - 30/18*60 = 97,2%
application X availability = 100 - 60/18*60= 94,4 %
If we take your formula availability = 97,2% * 94,4% = 91,8 % .
However, the correct availability ratio is 94,4 % , because network and application were interrupted at the same time , so you take the longest interruption.
The example is to ilustrate that calculating availability of components and trying calculations to evaluate the availability of the service does generally not work.
What you need to put in place is a mechanism to track down directly the downtime of the service as a whole, and calculate the availability from there.
For example you put a of PC somewhere in the far end of your network , with a piece of software that emulates users activities (up to business transactions) and tracks down service information like response time per transaction, delays, availability (based on predefined-time-out). The availability ratio will then be: 100% - total of unavailability duration (whatever the reason) / planned operation window.
N.B. you can/should still measure the availability by component, to know where improvement are most needed , but that will hardly provide you real availability measurements as users see it.
hope thiscan heps
JP _________________ JP Gilles
As long as 1 of the 2 was up, the availability for the traffic link was 100%.
You have to look at the SLA to get the operating hours for each of the services
If the systems are to be used 0800-1800 only and there is no unauthorized downtime/outage during that period, then the systems are available 100%
If you have a SPOF for the server service, then you cant have any down time or if the SLA allows scheduedl approved downtime, then that needs to be factored into availability numbers
The workstation addition puzzles me. IF the service is available; but the workstation is turned off, is the service available ? Yes it is
Pedantic but that is how you should look at it. The W/S availability should be a separate one unless the w/s is a required component of the service _________________ John Hardesty
ITSM Manager's Certificate (Red Badge)
Change Management is POWER & CONTROL. /....evil laughter
The Fault Tree analysis is mentioned in SD, that's correct.
At the end of the day, the key point to remember is that availability is in the eye of the beholder, ... the user of the service. What ITIL says is that availability needs to be measured from the customer's perspective. That is the reason it explains how to roll up system availability data into a system availability. It is a very important point to limit the common disconnect that exists between IT's service quality metrics and the customer's perception of that quality.
But as it was mentioned above, if a service only needs to be available from 8AM til 5PM, there is no need to measure its availability outside of those operating hours, unless that service participates in delivering another service which has a different window of availability. _________________ BR,
Technology Consulting | Service Excellence
Red Badge Certified
You cannot post new topics in this forum You cannot reply to topics in this forum You cannot edit your posts in this forum You cannot delete your posts in this forum You cannot vote in polls in this forum