We had a couple outages in our i-Dialogue hosting service today that resulted in about 7-10 minutes of total downtime for the week.
Our enterprise hosting provider guarantees 99.9% uptime, which as you can see from the table below, is 8.76 hours of downtime per year, or about 10 minutes per week.
| Guarantee % |
Allowable downtime per year |
| 99% |
87.6 hours (Ouch... thats 3.65 days!) |
| 99.9% |
8.76 hours (much better) |
| 99.99% |
52.5 minutes (not bad) |
| 99.999% |
5.2 minutes (The "5 nines" holy grail) |
| 99.9999% |
< 1 minute |
While it's tempting to make a lot of noise with our service provider about 7 minutes of downtime, it's actually still within the 10 minute per week SLA. I just wish it were 10 minutes of planned downtime at 3am in the morning 
We've definitely learned that 99.9% uptime on massively scaled, shared infrastructure is a constant battle. Us "software guys" do not envy the "hardware and network guys" at all.
So, we spent some time today researching how we could offer 4 and 5 nine uptime options to our customers who need near 100% reliability.
Isolation and redundancy are the keys to achieving 5 nines, which effectively means doubling the costs of the underlying infrastructure. The actual costs are not as exponential as I originally thought. The path from 3 nines to 5 nines is nearly linear based on our early research.
Keep an eye on our pricing page for more details in the future.