I am trying to calculate availability for a proposed system. The system is to be available for six days a week between 0600-2100 -- business hours. However, in the business hours, it is to be available within 30 minutes. If the system goes down, what would the availability be -- 1-30/17*6, 1-30/24*6 or 1-30/24*7?
I'm afraid that availability calculations don't really work the way you describe. You cannot calculate availability based on a recovery requirement, nor can you calculate it before the fact. Availability calculations are traditionally a historic look back at how available a system was over a given period of time. The formula that is most commonly used is:
MTTF A = ----------------- MTTF + MTTR
Where "A" stands for the availability, presented as a percentage. "MTTF" is mean time to failure. That's the time that the system is actually operational and "MTTR" is the time it takes to repair, restore, or recover the system. Normally you only consider the period of time when the system was required to be operational. So, if your system is down 30 minutes during a full seven days, 168 hour week, you would divide 167.5 (the uptime) by 168 (the total available time) and get 99.7%. If the requirements were that the system be up only five days a week, eight hours a day, that same 30 minute outage would bring your availability number down to 39.5/40, or 98.75%. In your question, you are not telling me how long the system was up. What you are asking is for a way to calculate availability based on a recovery time objective. If, in your example, the system were to recover in 30 minutes once in a week, your availability figure would be the 99.7% we discussed above. If instead, the system went down for 30 minutes every hour but never for longer than 30 minutes, you would still meet the Service Level Agreement you describe, but you'd only achieve 50% availability. When you do us historical data to calculate uptime percentages, you should base them on the operational hours. So, I would use 90 hour weeks (6 * 15 hours) in my calculations. And. I would count any outage, scheduled or not, that occurred during those periods. If I misunderstood your question, or if you have other questions, don't hesitate to ask. Evan L. Marcus Editor's note: Do you agree with this expert's response? If you have more to share, post it in one of our discussion forums.
Dig Deeper on Storage management and analytics
Related Q&A from Evan Marcus
Storage expert Evan Marcus compares software and hardware RAID and outlines the benefits and drawbacks of each. Continue Reading
Can I install an operating system (OS) and put data in different partitions of RAID-5, or is RAID-1 (hardware-based RAID) a better solution? Is it ok... Continue Reading
Storage expert Evan Marcus offers an answer to the question: "Is hot-hot replication possible with EMC's SRDF?" Continue Reading