Is there anyone out there who configures rack space with shared drives in a SAN or other highly sharable architecture?

What I'd like to have is the ability to move accounts from one dedicated server to another by simply moving the domain over, and not have to deal with disks at all.
Juggling accounts from one server to another to deal with load balancing would be simple in this model. If you had specific accounts that needed say a specially tuned mySQL database engine, you could cluster them on the same server.

If you had a bazillion little accounts that needed nothing special in the way of services (the free class of html-only users?), then these guys could easily be loaded on a single server with nothing extra on it at all. When these guys decided to grow in complexity, they could be moved to a new server by simply moving their IP address to the new hardware. They would have access to the new features without any pain at all.

The SAN drive arrays would allow you to add in more disk space, available to any account, without taking anyone offline, and of course be redundant and fault tolerant using RAID 5.

If you had a single SSL server for your domain, it could live at https://secure.mydomain.com and be used by all accounts, rather than locking users into a specific server today. Once I use the specific https image on my specific server today, I have to change URLs if I move to a new server. If I buy my own certificate, that's not a problem, but if I want to simply use the certificate from my hosting company, I'm locked into the specific server my account is on today.

Why should I actually ever care which server my account is on? Why not virtually host on any number of dedicated servers and float between them?

What's typically done today? It looks like a dedicated server with an internal hard drive or two; fill it up with accounts till it chokes, and then add another server. If you lose the hard drive, you lose all the accounts on that drive - shutting down the server and all the users till you swap in a new hard drive and restore from backup.

What are the other common options out there today?

-t

t,

You raise some interesting questions. I have a similar mindset and we have built our internal infrastructure in what we call a clustered hosting environment.

Hosting, in my opinion, can be separated into one of three types:

1) shared server - where one server holds numerous websites
2) dedicated/managed server - where one server is dedicated to one website
3) clustered hosting - where websites do not reside on servers but on redundant clustered appliances ( in our case we use Net Apps)

The problem with the first two methods is that websites will go down if the server goes down. In other words, there is a single point of failure no matter how much bandwidth, cpu speed, and electrical power is available. As well, a looping CGI on one website can kill a server and all of the websites that reside on it.

With the right architecture, it is possible to achieve near perfect reliability once you eliminate the reliance upon a single server architecture.

Your description of desired functionality is achievable with this type of system topology. :blush: