When are servers not servers?
In his Dr Dobb’s overview, Andrew Binstock talks about the prevalence of low-cost, low-powers and suggest in the title of the piece that they have begun their steady ascent over more traditional servers. His concluding statement, in fact, suggests that they will replace the “pizza box” servers we have come to know and love.
Ironically, to me, the notion of a “server” still conjures up images of row upon row of full-tower machines, whirring away. In fact, I have one of those under my work desk at home, doing… nothing. Right now I have it more or less permanently switched off.
Andrew and I have disagreed on things before, but on this score, he’s right: the machines we commonly call “servers” are, step by step, slowly but surely, becoming smaller, quieter, lighter, better power-friendly, and all the other things we have traditionally associated with the client side of the client/server equation. It’s not new: I have a couple of friends who, in order to do “cloud” or “cluster” presentations, carry around with them a small private cloud. One of them carries around (as in, with them to conferences and such) about a half-dozen laptops, the other, a custom-made rack of Mac Minis, a router, and other accoutrements. Yes, if you attend TechEd, you probably know exactly whom I mean.
But this raises some interesting questions. If servers are becoming smaller and lighter and are still fast enough to be considered servers, what does this have to say about infrastructure? Andrew touches on it briefly,
This model of low-cost, low-power devices is the way of the future. What I am describing here is not terribly different than building your own personal cloud from inexpensive machines. If you had chosen to keep the $300, you could have gotten this much from Rackspace's cloud server: 512MB RAM and 20GB HDD running Linux. That's not close to as much horsepower as my machine delivers However, it gives you two advantages: You have no additional ongoing costs (power consumption, parts replacement), and because it's off site, you have an instant off-site backup of your code base. Other companies, such as IntoVPS.com, give you about twice Rackspace's resources for the same price. Eventually, the pricing of cloud options will drop to close to the low-power, on-site devices, I expect. (Source: http://drdobbs.com/tools/232500406?cid=DDJ_nl_mdev_2012-01-25_h&elq=5c23117c5cff4d06820726bd0294693a)
… but putting the discussion of “on-premise” vs “cloud” off to one side for a moment, it raises a more interesting question: if servers are small enough to carry around with us, are they still servers? Historically, the server has always been the machine in the data center, but if we have tools that allow servers to synchronize data between them easily (such as we see going on in tools like Dropbox or Evernote), and the servers are small and portable enough to fit in our pockets, then are they still servers?
Think about this for a moment: the servers that Andrew describes (“a 1.8GHz dual-core Intel Atom chip, 2GB RAM, 250 GB SATA, HDMI, 6 ea. USB, Wifi, and GbE” and “a dual-core 1GHz ARM-based Tegra chip from Nvidia, had robust Nvidia graphics (HDMI), 1GB RAM, a 32GB SSD or a large capacity HDD, and all the USB and other ports you could possibly want”) are hardly the heavy-metal monsters we used to think about when discussing “servers”, and yet still serve the purpose. If we don’t need the server for its processing power, and if we don’t need it for its central location (as a rendezvous point for clients to discover each other and/or centralize data), then what purpose does the server serve?
Maybe it’s time to take a really hard look again into those peer-to-peer ideas from about a half-decade ago.