There's some amusingly old hardware information archived in the Wayback Machine.

BTW, Linode uses dual-CPU servers. Currently they use quad-core CPUs, meaning there are 8 cores total, though your node only has access to 4.

Linode probably uses rackmounts rather than whitebox ATX simply because it's cheaper; Linode colocates all their hardware, so it would make much more sense to use small rackmount boxes rather than huge whiteboxes.

SuperMicro is affordable, but not low quality. They're pretty decent.

The super cheap dedicated servers are probably better value for money, but just remember that hardware can fail.

I know that Linodes aren't immune to the problems, but you've got less hassle when things go wrong. If there's hardware issues on a cheap dedicated server then you'll have to send a support ticket and pray that they answer it in a reasonable time, which sometimes doesn't include out of hours support. If your server goes down on Saturday you might not get a working server until Monday morning unless you pay extra for it.

Whilst I don't mind doing the sysadmin work on a physical server, it's one less headache to admin a virtual server and concentrate on the software side of things.

Out-of-band console access is worth its weight in gold (well, since it weighs nothing, I guess it's worth more). Good luck finding that on a $30 dedicated server!

I personally would have no problem with a company using standard off the shelf commodity hardware as long as they don't have excessive downtime and, when they do, they can restore everything quickly and accurately. I mean, really, what is wrong with using two desktop computers for web servers if they are configured and working for redundancy?

I believe there's a little company somewhere on the internet that uses commodity hardware - disk drives etc. I think they're called Google or something like that

It works for Google because of scale.. If a few percent of their 100k machines in a center are down, what's the harm? It's all a cloud, doesn't matter any. When you're talking about dedicated hardware though, it doesn't work so well - if I'm renting a server, I want higher availability than the spare machine I have plugged into my cable. Hence going with a high performance VPS like Linode.

Rent a cheap dedi from someone using commodity hardware, you're going to be regretting it the first time something goes wrong (and something always goes wrong!)

I know where you are coming from, but I disagree. The high availability is not a hardware problem. It is a configuration problem. The reason commodity hardware works for Google is not because of scale, it is because of configuration. Assuming $1,500 for a mid- to high-end desktop system, I could buy anywhere from two to seven desktops for the cost of a single server. That gives me plenty of systems to build high availability into the configuration.

The only thing high end server hardware provides you is more time between failures. That means, when your server is down your site is *still* down. Now, if you had configured the system for high availability then you could lose a server and still be available.

It's all configuration, not hardware.

Hardware is not completely absolved. Failures do come in twos, more than I ever would have thought before working around a large-scale deployment.

Choice of hardware is still quite important when planning a deployment. Even if you've planned for failure, replacing said failed components can cost a company its bottom line.

_________________
Disclaimer: I am no longer employed by Linode; opinions are my own alone.

Hardware fails no matter the configuration. I know that, but in a proper high availability configuration the probability that enough systems (2 or more) running the same services will fail within a relatively small amount of time is low to very low.

Failures occur no matter the hardware. That is why I plan for high availability whether I am using $1,000 desktops or $10,000 servers.

As far as the choice of hardware, I do agree, but probably not to the extent you think. In my previous message I estimated $1,500 for a mid- to high-end desktop system and not a $400 low-end desktop system. Buying high-end desktop parts is still cheaper than buying most low-end server parts.