FragSwitch: Hosting Tutorial: Uptime (Part Two) - FragSwitch

Continuing our Uptime Tutorial from last week, we will be covering methods and equipment used to prevent downtime and a little explanation as to what they do.

Electrical Downtime
In most First World countries, electricity is constant and mostly uninterrupted, however, there are occasions when the power is interrupted (black outs) or limited (grey outs). If you’ve been through a black out, you’ll know thateverything electrical goes off, if your server happens to be in a place with a black out and it shuts down your website will not be accessible. On these occasions, equipment is in place to limit the effect these have on servers and networks (the internet being one of the latter), however, not all data centres have the equipment, so it’s worth checking with your provider when you sign up.

Uninterrupted Power Supplies
A UPS is effectively a large battery, this are connected in between the server and the power supply, if the power supply suddenly stops supplying power, the battery will kick in and the server will begin to run off of that instead. The problem is that batteries only last so long, normally a couple of minutes at most when under the load of several servers.

Diesel Generators
As a UPS will only last so long, there has to be a more long term solution available to provide electricity. This is normally in the form of a diesel generator or two. In most situations, these are configured to start automatically once an electricity failure is detected and take the load from the Uninterrupted Power Supplies. Providers will normally have contracts with fuel suppliers to ensure they don’t run out of fuel before the power supply is restored.

Network Downtime
Downtime caused by Internet Service Providers can be a little bit tricky to solve as it requires a third party to fix the problem in most cases. Downtime like this could be caused by almost anything, including a power failure for their own equipment, but it could be that someone has planted a shovel between that highly expensive fibre optic cable of theirs (there was a case in the UK of a cable being dug up and stolen). Another cause could be that your provider’s router fails and stops connecting servers to the wider internet.

In both of these cases redundancy is the only way downtime can be prevented. Data centres will tend to have multiple physical connectionsto the internet, which should hopefully prevent downtime if someone happens to put a shovel through one of the cables. Providers will be responsible for their routers and this means to prevent downtime from occurring, they need to have clusters of routers prepared to take over in case one goes down.

In Conclusion
While these practices can ensure a server doesn’t go down, there will always becases of when there is a huge failure and the server does go down, normally for hours on end, the best solution if your site has to be up is to have several failover solutions. This is whereyour site is copied to several places around the world and if one location goes down, another is used to serve your site, however, this is costly.

If you have any comments, questions or suggestions, please feel free to contact us.