99.999 or Five Nines Uptime
99.999% Uptime is difficult to achieve and may not be worth the cost.
From the table below, your website would be unreachable for only 5 minutes a year if you are to be called 5 nines or 99.999% available.
| Uptime (%) | Downtime |
|---|---|
| 90% | 876 hours (36.5 days) |
| 95% | 438 hours (18.25 days) |
| 99% | 87.6 hours (3.65 days) |
| 99.9% | 8.76 hours |
| 99.99% | 52.56 minutes |
| 99.999% | 5.256 minutes |
| 99.9999% | 31.536 seconds |
It's very difficult to be 99.999% available.
If you have a hard disk error do you think you could fix it within 5 minutes each year ? probably not. You would need a second system that can step in at times of failure, this backup system would need to copy anything that could possibly fail. You would also need to think about how quickly you can switch from your failed system to your backup.It's expensive and the money may be better spent elsewhere.
You can use a company or service to configure a high availability system that would meet the needs for 99.999% uptime. But, they are not cheap, high availability systems are complex and require expensive technical skills. Some sites would benefit from expenditure in areas such as usability, performance and design first. Although if your site does go down, an error page that offers to inform the customer when the site is back up would be a nice addition to many sites.Your customers may not mind lower downtime
Do your customers really need your service to be 99.999% available ?. For most sites I suspect the answer is no, especially when the costs of supplying such availability are taken into account.
Posted on Wednesday, Dec 14, 2005
4 Comments
Should one the servers fail, the others share the load.
Just recently, with the advent of Electronic Health records, hospitals do need high availability for the historical data to compare to the current data. The current data can and probably always will be available manually.
Most computer systems in hospitals are there for record keeping and are not as critical as one would think. Some areas, like Radiology, are changing that by collecting data electronically.
Typically we have to do one of these cold boots / year. A warm boot, where the OS exits and then immediately reloads is possible, and that takes under 90 second on this hardware ( which is nothing special, typical HP pizza box server. )
5-9's is not hard to achieve when you use reliable code.
Certainly 4-9's is easily achievable for most systems for which a recovery is not required / often necessary. And with GW, at least, there are 3rd party products which make a hot standby server easy to create. But we don't use those.... there is no requirement for 5-nines for these systems. Clustering also allows for sub-3 minute failover of services on shared storage.
%99.999 is easy to achieve if you don't have to reboot the box every other tuesday to patch components you don't use, but are present by fiat.
The same works in favor of Linux, where you can strip down the OS to the kernel and the applications. Don't need a GUI? No problem. Not using print services? No problem. With OSX and Windows, so much of the OS is the fenestrations, you end up havinbg to apply patches for things that you don't use to float your services. Andf less code means less code to go wonky, be attacked, need updating, ......
Virtuallization is another factor. ANY OS will boot 2-3 x faster on a virtual platform because there is no POST, SCSI probing, ...... so even Windows will boot in 30 seconds. And that is a great help to super-tuesday reboots - which can really eat into the uptime. So for most Windows servers you can get close to 99.99 easily, and Linux / NetWare you can achieve higher uptime values by managing the obligatory patch related reboots.
Leave a Comment