How does this work? Where is the data stored for the domains... on all the servers?
I'm not exactly sure how this would work

If you're referring to hardware based firewalls, yes, all the data would be stored on servers behind the load balancer. That piece of hardware would then try and distribute the requests to the least heavily loaded server.

In Load Balancing, as The Planet is offering it, you buy a packages of services (eg, http, pop3, smtp) and nodes (The Planet IP addresses). You combine your nodes into a pool, which you then attach to a Virtual Server. A Virtual Server specifies the service you are load balancing, the nodes that provide the service and a few other settings.

If one of the nodes in your pool goes offline the load balancer automatically stops sending requests to that node until it's functioning again, assuming the health check for the pool is set up. The health check makes sure the service is working every few seconds.

The nodes in your pool can be any system inside The Planet IP space (that you currently own). The nodes can be systems you currently have in production, or you can buy a node as you need it. Each node would have to have identical content to serve to users.

Additionally, we've built some basic Urchin reporting that uses Urchin's Load Balancing Module which puts (or should, anyway) together a better picture of your traffic than two separate profiles, one for each node. Obviously, the Urchin reports are only useful if your balancing HTTP.


Because of the way we're doing this, the load balancing is being done at what's referred to as the Application layer. This is a trade off for quite a bit of flexibility and functionality. The Planet doesn't have to limit you to machines on a certain part of the network, or the number of nodes, or the services you can balance.

The most noticeable effect of using the application layer is that all of the requests to the nodes appear to come from the load balancer. For HTTP this is easily fixed for the purposes of logging. The Load Balancer will send an 'X-Cluster-Client-Ip' header, which you can then put into a LogFormat directive in apache. For example:



(That's all supposed to be all on one line)

You then set the log format "nickname" to be Zeus in the virtualhost configuration on the nodes:



This will cause the logs on the local nodes to reflect the clients real IP. It seems to me that you could go further and use SetEnvIf to distinguish between hits from the load balancer and direct hits, but I haven't had time to come up with a recipe for that yet.

What's the pricing for all this?

$49 a month for their basic service, which is:



+$10 per month for each additional service


+$49 per month for each additional node

OK, Could you make the subject clearer with this example. I'm sure there are still many people who don't understand how SM's load balancing works.

Let's say:

I have Linux box with cPanel hosting a number of sites. I'll then buy another box, which is going to be identical in contents to the first one.

I want to load balance httpd, so I'll buy the service and the rest of it ($49?). Now:

1. How would the data between my two nodes synchronised?

2. Would I have to change the shared IP of my site on my first node?

Please add you questions ..

Thanks

That's your choice. I usually use rsync http://samba.anu.edu.au/rsync/ to get the data between two machines identical. Obviously if there's a database involved things get a bit more complicated.



I'm not entirely sure what you're asking, but you would have to change the DNS records for the site(s) that you're load balancing to point at the public ip we assign you from the load balancer farm.

Thanks James,

I bet many people think:

Just get a load balancer, add an extra node and you'll have a fault-tolerant server.

I believe over the next few weeks, will see more and more questions and threads about this new service.

Are you able to do load balancing for a mysql database too? In the way that there are two database servers and content is kept identically? E.g. mySQL cluster solution?

Or is only the simple form possible, e.g. 3 web servers and 1 database server?

Now that would be interesting.....

As far as I know the only feature of MySQL for parallel processing is data replication which is not mature enough yet.

I would be a little concerned with speed issues for writes -- if one server is being written to, the other has to be locked for that duration to prevent data mismatch. Is that right or is some more advanced mechanism being used?

I can imagine that one can imagine a scenario with one main DB server and one fallback server, where the fallback server turns into the main server once the older main server is somehow disabled.

Exactly, according to the planets boastful advertising, I did to

Redundancy

Total Balance not only provides redundancy within your server topology, but ensures full redundancy upstream as well. With two dedicated load balancers forwarding traffic to your server cluster you'll never have to worry about a single point of failure within the Total Balance system.


a lot of donkeycrap !

T

The pricing here is correct. If you are interested in adding this to your account please contact Sales.


We will get you setup.

Hi,
About pricing when I asked for an offer I got the following email which looks like 10 times the price you are talking in this forum.
Reply from sale was:
"Excellent, we provide the Alteon 180e hardware loadbalancer for $595.00/mo with a $595.00 setup fee in which our Technicians will install this and manage it for you. I have attached some helpful information on the Alteon 180e for you to review."

Look at the dates of these posts, all back in 2004. Any pricing quoted is no longer valid. They used to use a cheaper load balancer which apparently didn't work at all.