This is kind of an important issue.

I'm working on a new project, which is a very exclusive online magazine. (I will not give out the URL just yet).

Anyways, I have partnered with one web hosting company, which I personally know and trust. But, as everybody else, there might be downtimes. It may take only a few minutes, but still not a good thing when you try to get national publicity and people, trying to access your site, find it unavailable. So, this company sponsors us for a credit/publicity - something like "powered by...." stuff.

I currently have 2 nameservers with them.

Is it technically possible?

Is it possible to find one more company that will give us two more nameservers, so that if we put them into our domain information - the site will automatically be redirected to those nameservers if the first one goes down temporarily? Same goes to e-mail servers (POP3 and SMTP servers should be the same i guess...).

So you see, basically, all the work will be down by the first host and the second will play the role of a back up. I'm trying to figure out if we can achieve 100% uptime.

Will this technically work?? :cartman:

Yes, if the both hosts are capable of "mirroring".

Say for example you do to a download site that has a certain script and it offers you different locations to download it from. Same concept here.

Oh, man.. Silly me, i forgot there there is only one domain name that should be somehow activated in both companies, right??...

or am i messing something up again?

I don't really understand the "mirroring" feature. Does it apply to accessing the site in general or just certain internal parts/files of it?

Well, it's pretty simple actually.

First, your registrar has to allow multiple nameservers.

Your entries look like this.

NS1.YOURHOST.COM
This resolves to your primary host IP XXX.XX.XXX.123
NS2.YOURHOST.COM
This resolves to primary host IP backup XXX.XX.XXX.456

Then you have another server on a different provider line. *This is very important because it doesn't really matter if it's on a different server or not because of the line goes down, you don't have redundancy*

NS3.YOURHOST.COM
This resolves to backup host primary IP XXX.XX.XXX.098
NS4.YOURHOST.COM
This resolves to backup host backup IP XXX.XX.XXX.765

Thus at your registrar you have four nameservers. The two primary support each other in case of hardware failure. The two backup support the primary in case of connection failure.

However, the four servers need to be able to communicate with each other and "mirror" your content on each respective hard drive.

Same simple principle as home networking and multi-writing, but just applying the concept to servers located 1000 miles away from each other. :D

We basically did the same thing with my old company, www.ecampus.com Dependant upon where the visitor was located and accessing the site from, they would be directed to the server closest to them to reduce ping times and also to balance server load.

Now if you're going to ask me how to do that... I'll have to bring in the charts and consulants and presentations and *drones on for hours*.

Thank you Coach very much!

I was guessing this is how it would be, but you've made it clear!! No need for consultants.. :D

Best,

Not a problem at all. I'm glad I could be of help to you. Normally I'm clear at mud. :)

Hi Coach and Artashes,

I am following your conversation with much interest. In fact, I have just brought this up yesterday to an alternate hosting company, and I am interested in developing this as a nifty feature for myself, and if it works, launch it as a product for one of my hosting companies. So, I would love to hear from either of you if there are working models out there, and how I could put together a protocol to get started on a trial.

Thanks for your input!

Peter

This is a somewhat cost-prohibitive setup that I am familiar with. The scale with which my previous company deployed this was in the millions of dollars, however it *can* be done on a much smaller scale.

The three things you should really concern yourself with are as follows.

1: Storage redundancy
2: Server redundancy
3: Clustering

The first two are pretty self explanatory. Clustering is simply distributing the workload among all the servers.

A quick search on google might best help you out with information on each of those.

Keep in mind that to actually do this you have to have the same setup on each box. A Redhat can't go along with a Win box and FreeBSD won't work with TurboLinux for example.