With huge cloud players building their own stripped-down servers on a massive scale, midsize data centers could reap benefits from the altered server landscape.
This week, Wired magazine published an article entitled, "How Amazon Followed Google Into the World of Secret Servers." In the article, the author described the experience of a person who had the opportunity to see Google's servers in a data center in the early days of the company. I, too, had such an experience back in the dot-com era. My employer shared a co-location facility with a number of large .coms at the time, including Google, Yahoo, and eToys. The Google cages were, indeed, impressive in some ways and surprising in others. It was clear that the company was working with custom hardware, though. What I saw was this: Each 42U rack had 84 servers in it. Each was half-length and they were mounted both front and back of the rack. As you can imagine, that's a lot of weight. So, at the bottom of the rack, the engineers placed 2x4's across the bottom to support the servers.
Little did I know at the time that Google was in the early stages of a strategy that has lasted a decade and that would help to usher in a new era of computing. Today's huge providers -- Google, Facebook, Amazon, etc. are no longer buying servers from the likes of Dell and HP, choosing instead to build their own or spec their own through Original Design Manufacturers, which custom build servers to exactly the specifications desired. This allows these large providers to buy only the components they need and nothing more. After all, saving $25 for an unnecessary USB port when you're buying 40,000 servers can add up to big money.
Although most enterprises will never achieve scale the likes of the players mentioned before, companies like Dell and HP are creating gear that can help bring cloud-scale-like capability to the enterprise. Dell creates a line of cloud scale servers dubbed Dell PowerEdge C servers. These servers are built to scale. For example, the Dell C6220 is actually four complete server nodes in a 2U space.
This line of servers has its origins in what used to be called "DCS" units at Dell. DCS-Data Center Servers-were huge quantity orders of custom made servers for specific unnamed clients. Originally, these servers never made it to the general market, but it was common to find them on eBay once the original client's lease ended. This was a foray into providing a more commoditized server at a lower cost than a general purpose server.
Now, the C series servers have become an item that you can order from Dell, so they've come out of the woodwork a bit. Again, although these products are not yet in the mainstream, as CIOs consider server replacement in their organizations, it might be worth considering a higher density server such as the C6220. It provides great scalability at a better price than buying four individual units. It's a small step on the road to becoming a bit more like Google or Facebook.
Personally, I believe that, as time goes on, the further commoditization of hardware is simply inevitable. These kinds of systems will become the norm, even in the midmarket, as software platforms continue to mature. Already, servers based on software (vSphere) outnumber those based on hardware, so commoditizing and reducing the cost of hardware makes good economic sense.
We're also seeing a similar software-based commoditization scenario playing out with companies such as Nutanix, Pivot3, and Simplivity. These companies use inexpensive hardware and drive it with powerful software to create robust, scalable solutions.
I don't see most enterprises going the route of Google and Facebook and designing their own lines of servers, but I do see commoditization through hardware such as Dell's C-series servers, Nutanix, Simplivity, and Pivot3 as having a long-term disruptive influence on the overall server market. I see these solutions as having been inspired by the commodization movement that was kicked off by companies such as Google. In that small way, I think the mainstream is seeing benefits from the steps that Google, and other big players, have taken to reduce their own infrastructure costs.