Take the 2-minute tour ×
Server Fault is a question and answer site for professional system and network administrators. It's 100% free, no registration required.

I have a website and a while ago, the web server of the company hosting my website was down for about a day. I consulted the company for a solution on how i can stop this from happening in future and they suggested to have a second machine and which will be connected to my current website/web server by a "load balancer" (at an additional huge cost!!!). The second machine will be replicate of the first one and so if i goes down, the other will always be running.

---- Explanation ----- My hosting company suggested that it will be a good idea to have a second machine running at the same time and both the machines will be connected by a load balancer which reduces the rist of a downtime. The second machine will be a mirror of the first and any changes to first must be replicated in the second.

I don't mind spending money if it really saves my website from going down. I want to know is it worth having this "load balancer" for my purpose?

My website is a 24/7 service. I cannot afford an outage of 24 hours/1 hour. I don't mind using this "load balancer" as far as it is really worth. I am not sure if its just a marketing trick of my hosting company or really a "best" solution.

Thanks for help.

Regards

share|improve this question

migrated from stackoverflow.com Nov 14 '11 at 1:11

This question came from our site for professional and enthusiast programmers.

1  
Is your webserver mission critical? Do you lose money when your site is down? You need to balance the answers to those questions vs. the expense of setting up a load balancer (or some other option) - we can'd do that for you. Really, the question you should be asking is "how can I maximize my uptime and minimize my expenses?". –  Jason Nov 14 '11 at 0:57
    
To expand on what Jason said, it's about weighing up the potential for loss (both in direct monetary loss and loss of confidence in your service) against the additional cost of having an improved availability. You need to decide what constitutes good value for money in this sense - e.g. if you expect your current level of outage to cost you $1000/yr in potential custom, you should look to be spending at most $800/yr on a solution that cuts the outage level by 80%. Obviously this is only a rough indication, but that's the kind of thought process you need to undertake. –  Polynomial Nov 14 '11 at 1:14
    
Thanks a lot for replying i have update my question. Regards –  user427969 Nov 14 '11 at 1:38
    
Signatures are discouraged on all posts –  random Nov 24 '11 at 1:19
add comment

3 Answers

Depends what you require your service level to be. If you need to be up 24/7, then go for a high availability solution.

Your host's current suggestion sounds a little questionable. As you've described it, if the primary server goes down, the load balancing service dies too and there's no way to offload traffic to the backup. When load balancing you usually opt for a DNS based solution that has multiple A records, each pointing to one of your servers. If one goes down, clients will try the next in the list. This means that there is no single point of failure that is within your control.

As to whether it's worth paying a lot extra for, it depends on a lot of things.

Here's my suggestions:

  1. Evaluate your availability requirements. What's your budget? What's your minimum uptime requirement? Do you need all of your service to be high availability, or just some of it?

  2. Research replication technologies for the types of service you are using. Most SQL database servers offer replication as a feature. It may be cheaper and more reliable to set up your own load balancing and replication across two separate hosting platforms, e.g. one on your existing provider and one on a cloud VPS provider such as Amazon EC2.

  3. Look into hosting services that provide a Service-Level Agreement (SLA). A "three nines" SLA means that they are required to be up 99.9% of the time, and must pay you a penalty fee if they fail to meet it. Four nines is 99.99%, five nines is 99.999%, etc. It may end up being cheaper to go for a decent SLA rather than explicitly setting up replication, though both may be a good solution if you need very high availability.

  4. Ask your provider(s) about backup facilities and the average fix response time for different hardware failures (e.g. dead disk, fried PSU, etc). Shop around to find the best figures for the right price.

  5. Consider looking at providers that use data centers that operate on more than one fiber network, e.g. Cogent and Level3. This means that if one connection goes down, you're very likely to be still up via the other connection.

It's all about tailoring a solution that's right for your service. If you find that your budget isn't sufficient to keep availability at the level you need, then it may be time to rethink the feasibility of your project.

share|improve this answer
    
Thanks a lot. I have updated my question to what my hosting told me about load balancer. Budget is not a problem as far as my customers are happy. I just want to know if this is a really good idea if i want to avoid outage in future or this is just a marketing/sales tactic. Regards –  user427969 Nov 14 '11 at 1:49
    
Sounds a little fishy. Ask them what happens if the load balancer goes down. What you're looking for is a system that has no single point of failure (SPOF), i.e. "if any one single network connection, load balancer, server, backup server, database server, HTTP daemon or any other software or hardware component fails, the system continues to be available". You may want to express this to them and find out what solutions are best for you. I'd also advise you to contact other providers to see what solutions they have. –  Polynomial Nov 14 '11 at 6:55
    
You may also want to invest in an SLA. I'd guess four nines (99.99%) should be enough for you, which is max of 53 minutes of downtime per year. You'll also want to check what contractual tems it is under. A lot of providers say 99.99% SLA, then go on to say something like "an outage is defined to be any 10 minute period when the service is out of service", which means they can get away with having your service down 90% of the time (9 mins down, 1 min up, 9 mins down, 1 min up, etc...) and still not breach their SLA. Also check how much money they give you if they fail to meet the SLA. –  Polynomial Nov 14 '11 at 6:57
add comment

In that case, no. A load balancer is designed to distribute the load of clients between two or more servers in order to keep a fast service to everyone at busy periods.

What you need in this case is some sort of failover system, of which a decent way is another server at a different location. This can then be used by changing the DNS records for your domain when there is a problem. Of course, then you need to replicate to both machines regularly and without losing any data.

Really though, this is the fault of your host, and it's pretty pathetic that they're suggesting something with a huge cost to protect against something which was their problem. Also, if the load balancer dies, your whole infrastructure will go with it, so you'd need to create some sort of failover for your load balancer too.

share|improve this answer
add comment

I like your question. It's simple and has happened to everyone once in awhile. However your hosting company's solution has a flaw.

They can be down because of network issues at their end. Which means having a load blancer and X number of machines is still fail because if the company network is down they are down. Also load balancer is also a single machine.

If you want to really throw money there is a much cheaper and more effective solution. Its called Amazon AWS. the load balancer are cheap, servers are cheap and the possibility ot AWS going down (is there) but far lesser than ABC and XYZ bros hosting INC.

You will need expert help to do this right so this is where the money part is. But in my experience, life has been much easier and going smoother for me even if some people disagree with AWS as a choice.

share|improve this answer
    
Hi Thanks for replying. I asked this same question to my hosting company and they told me that they have 4 switches from different vendor now and so there are very minor chance that their server will go down again; previously they had 2 switches from same vendor and it got bug so both went down at the same time. –  user427969 Nov 23 '11 at 21:56
    
yes I understand. However I am speaking of connectivity to their data center. For example you may have 4 switches at home but only one internet access subscription and one point. I am hoping they don't. If I really want to throw money on a dedicated load balancer I would really start looking at AWS. It's easier to manage if you don't have expert admins at hand. MySQL failover is also provided by AWS and is only a click away. –  Vangel Nov 24 '11 at 16:14
add comment

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.