Web gurus, how would you go about doing this?

LostAngel said:
You could do a solution like apple where they have servers in every major city area and serve pages based on the requests location. But it's more of an IT solution rather than a web development businesses solution. You'd need to contact a good IT outsourcing company and see what they can do for you.
Click to expand...
They do this through Akamai. As does Google, CNN, etc.
 
It's for work actually. We have two components to our site... the e-business center, and the website. The website is the portal to the ebusiness center. Right now the ebusiness center is on our own servers, we have three of them, but boss wants to separate our website from the ebusiness center, so that if one of the boxes goes down, as long as the website is up, we can update our links and redirect them to the other boxes.

I suggested a cheap fix might be to host the site with a couple of hosting companies, and set up a failover of some sort. It may not be 100%, but it'll be damn close.

I'm supposed to be the expert on this, but I have no experience with failovers, as I've never had to deal with mission critical sites.
 
That sounds idiotic. Why not just setup a loadbalancer on your side that will AUTOMATICALLY redirect people to a working server using a VIP or something. Jeez. No need to make it more complicated than it needs to be.
 
I need advice on an issue like this. However in my case, it is slightly different. I have a web app that runs on 1 server. I have a backup server that has the same app, but it isn't running. Since this web app is mission criticle (it runs something every 30 seconds) if the main server fails for any reason, I need the 2nd server to automatically start the web app.

I have been looking at open source service monitoring tools like Big Brother http://www.bb4.org/ and Nagios - http://www.nagios.org/ , but not sure if this is possible or not.
 
fb said:
99% Uptime Guarantee means that the site could be down for a total of 87.6 hours in a year.

99.9% Uptime Guarantee means that the site could be down for a total of 8.76 hours in a year.

Either of those guarantees are pretty respectable and should be enough to please a client.
Click to expand...

And they're usually down in the middle of the night. (unless you're in china.)
 
Monkey_b said:
What if the loadbalancer goes out? That's a single point of failure that would take everything down.
Click to expand...
You get a load balancer that has redundant power supplies, multiple motherboards, etc. A load balancer is meant to be a never-down solution.

We use load balancers for each one of our clusters. The worst that has ever happened is that it doesn't notice one server crashes (due to Nagios being dumb) and we manually log in to remove the IP address from the options until someone can get to the datacenter to fix the server.
 
What if the internet connection goes down for the load balancer?

Basically, we can't be geographically sensitive. If an Earthquake hits L.A. and all power goes out for hours, people still need to get to our site.

The 100% up-time isn't as important as being geographically insensitive. We need the site in multiple geographic locations so if a natural disaster strikes, the site is still up.
 
Monkey_b said:
What if the internet connection goes down for the load balancer?

Basically, we can't be geographically sensitive. If an Earthquake hits L.A. and all power goes out for hours, people still need to get to our site.
Click to expand...


Dual port nic's on separate networks, load balanced over multiple upstreams to the internet.

Because you are geographically sensitive you really need to use someone like http://www.akamai.com/index_flash.html
 
Monkey... The same could be said for ANY solution you decide upon. You HAVE to have some sort of expected outage. There is no perfect solution. Your solution has holes in the fact that it's expensive, the failover could not work, could be a long delay, could be routing issues between each side of the country, could be a peering issue between providers, your 3 servers could all fail at once for your ebusiness platform... I could go on.
 
What about 1 hosting company on the east coast on one network, and 1 on the east on a seperate network? That would provide damn good up-time, and be geographically insensitive, wouldn't it?

We're not talking about some intensive web-app, this is a simple website, it can be hosted on your average $20 host. I mean I look at serverseed, and the past year they've barely been down, but if they were down and I had another serverseed on the other side of the U.S. ready to kick in, I would be good to go.
 
Last edited:
Colosus said:
Monkey... The same could be said for ANY solution you decide upon. You HAVE to have some sort of expected outage. There is no perfect solution. Your solution has holes in the fact that it's expensive, the failover could not work, could be a long delay, could be routing issues between each side of the country, could be a peering issue between providers, your 3 servers could all fail at once for your ebusiness platform... I could go on.
Click to expand...

agreed

and furthermore at this point it is no longer your problem (as the web guru). That's what networking guys are for.
 
I would have put a "except for circumstances beyond our control such as power outages and major natural disasters (& etc.)" line somewharzs in that contract
 
Monkey_b said:
What about 1 hosting company on the east coast on one network, and 1 on the east on a seperate network? That would provide damn good up-time, and be geographically insensitive, wouldn't it?

It's just a website, it's not some intensive web app that we're talking about, it can be hosted on your average $20 host.
Click to expand...

You mean east and west right? Well, anyways yes it would...in a perfect world. The most you would ever find is 5 9's uptime or 99.999% garaunteed uptime. Even the companies that offer this have a hard time pulling it off and end up paying you in SLA misses.

It's just the way it is.
 
99% uptime would work, the issue is we don't want to risk prolonged downtime.

If hardware dies, or the network goes out, or a fire burns the building down, or a hurricane strikes, the site should be able to kick in from another place across the country, on it's own server, and it's own connection.

If the site is down a couple hours a month, that's not going to ruin the business, but if it's down for days in the middle of our busy season, we're screwed.
 
Last edited:
Then what you really need is a backup plan. Incremental backups can be done and even if all your servers explode, if you're any good, you can have everything back on at the next DNS TTL refresh. But that's not what you originally asked for. We're trying to give you VIABLE solutions.
 
You must log in or register to reply here.
Back
Top