In a recent post on his blog, Sun CEO Jonathan Schwartz coyly hinted at a rethinking of the corporate data center. “Now I understand that IT infrastructure has to be put somewhere,” he wrote. “But the whole concept of a datacenter is a bit of an anachronism. We certainly don’t put power generators in precious city center real estate, or put them on pristine raised flooring with luxuriant environmentals, or surround them with glass and dramatic lighting to host tours for customers … Surely it’s time we all started revisiting some basic assumptions.”
It wasn’t hard to see that Schwartz had something up his sleeve.
Today, in addition to announcing an expanded push into data center virtualization, Sun is revealing that a year from now it plans to begin selling readymade data centers in shipping containers at a starting price of a half million bucks a pop. Designed by supercomputing genius Danny Hillis, the data-center-in-a-box will, Schwartz told the New York Times’s John Markoff, “be attractive to customers that need to expand computing capacity quickly.”
The container, designed to hold up to 245 server computers, can be plopped anywhere that has water and electricity hookups. “Once plugged in,” reports Markoff, “it requires just five minutes to be ready to run applications.”
Welcome to trailer park computing.
The containerized data center is one more manifestation of the fundamental shift that is transforming corporate computing – the shift from the Second Age client-server model of fragmented, custom-built computing components to the Third Age model of standardized, utility-class infrastructure. As this shift plays out, the center of corporate computing will move from the personal computer upstream to the data center. And, inevitably, what happened to the PC – standardization and commoditization – will happen to the data center as well. What is Sun’s data-center-in-a-box but an early example of the data center as a standardized commodity, an off-the-shelf, turnkey black box? Indeed, the intitiative’s codename is Project Blackbox – and the prototype container that Sun is showing off is painted black.
The effort reflects Hillis’s belief that computing is fated to become a utility, writes Markoff:
Long an advocate of the concept of utility computing, analogous to the way electricity is currently delivered, Mr. Hillis said he realized that large companies were wasting significant time assembling their own systems from small building blocks. “It struck me that everyone is rolling their own in-house and doing manufacturing in-house,” he said. “We realized that this obviously is something that is shippable.”
In many ways, the containerized data center resembles the standardized electricity-generation system that Thomas Edison sold to factories at the end of the 19th century and the beginning of the 20th. Manufacturers bought a lot of those systems to replace their complex, custom-built hydraulic or steam systems for generating mechanical power. Edison’s off-the-shelf powerplant turned out to be a transitional product – though a very lucrative one. Once the distribution network – the electric grid – had matured, factories abandoned their private generating stations altogether, choosing to get their power for a monthly fee from utilities, the ultimate black boxes.
Something similar will happen – is happening – with computing, but how exactly computing assets end up being divided between companies and utilities remains to be seen. In the meantime, commodity data centers, in various physical and virtual forms, should prove increasingly popular to companies looking to radically simplify their computing infrastructure and reduce the single biggest cost of corporate computing today: labor.
UPDATE: Dan Farber covers the launch of the Blackbox, while Jonathan Schwartz makes Sun’s marketing pitch and Greg Papadopoulos puts the machine into the context of data-center evolution. Blackfriars calls it “the ultimate computing commoditization play,” saying it “changes the economics” of data center construction. Techdirt is skeptical about the size of the market: “The sweet spot of companies for whom this will be ideal seems small. Its impact on Sun’s business won’t be as significant as what it represents, the continuing commoditization of corporate infrastructure.” Sun’s Tim Bray writes, “I have no idea how big the market is. But I’m glad we built it, because it is just totally drop-dead fucking cool.” (Question for Miss Manners: Is it kosher for a corporate blogger to use the f-word?)
An interesting dimension of this story is that Sun seems to be tracking Google’s practices. Thus Cringely wrote last year: / “[I]n one of Google’s underground parking garages in Mountain View. There, in a secret area off-limits even to regular GoogleFolk, is a shipping container. But it isn’t just any shipping container. This shipping container is a prototype data center. Google hired a pair of very bright industrial designers to figure out how to cram the greatest number of CPUs, the most storage, memory and power support into a 20- or 40-foot box.” /
One of Sun’s new servers (the one they called Thumper) looks like an evolution of a Google-style combination of simple servers and storage. /
Solaris is now open source, and thus a possible inheritor if Richard Stallman screws up Linux in the course of the GPLv3 revision. / It looks like Sun is positioning itself to be Google’s hardware supplier, or perhaps the hardware supplier to Google’s competitors. Logically, this makes sense. A computer manufacturer should be able to make computers that are better and cheaper than a search engine company — if not, most of what we know about industrial specialization and economies of scale and focus must be wrong.
Given that cooling has to be one of the main concerns, why did they paint it black? Wouldn’t a highly reflective silver been a better choice?
All the modern computer manufacturers have outsourced their manufacturing. Google buys in sufficient quantities to just go start to the contract manufacturers and negotiate a bulk purchase.
They also take the expedient step of having the extraneous materials removed (which typically mean something to the buyer… the look and feel and the surface to carry the vendors “logo”).
Shipping these delicate “naked” boards in proper packing and assembling the racks on site makes perfect sense from a QA perspective. Google also assumes a high rate of board failures and they just build that assumption into their operational
practices… they have built a custom “OS” (through Open Source) that routes around these defects.
Sun’s let us ship you a container with a million dollars of equipment inside sounds like marketing in search of a problem most companies don’t have.
Their IT budgets are hemorraging with payroll and outsourced consulting costs… not the cost of depreciable hardware. If the container helps with that problem then they might sell a few.
It is pretty good marketing, for a change, from Sun.
First of all, I believe the protocol for corporate bloggers and the F-bomb is that as long as it is used sparingly and for ornamental purposes only, then it is completely fucking kosher. However, I don’t think many would want to follow the “Jeff Nolan Guide to Corporate Blogging” so proceed with caution.
More to the point of the Sun datacenter doublewides, I thought Google was already doing this?
http://www.pbs.org/cringely/pulpit/pulpit20051117.html
so what’s the big deal?
Jeff,
The Google container has been rumored for a long time, since before Cringely wrote about it even, but to the best of my knowledge it remains unconfirmed. More important, even if it does exist – and my guess is that it does – it doesn’t do much good for anyone else. The Sun container makes the concept available to anyone with the necessary cash – that’s the big deal. I should say that the Sun container will make the concept available, assuming it actually, uh, ships.
Nick
Sun’s Tim Bray writes, “I have no idea how big the market is. But I’m glad we built it, because it is just totally drop-dead fucking cool.”
WTF?! Why is Sun building products when they’re not sure if there is a market? Is there any other way of flushing shareholder value down the toilet? Sun would be better off making Niagara servers cheaper – no one believes that TCO of a Niagara is lower than an Intel or AMD based server. And if Sun does indeed believe that TCO is lower, a better approach would be to provide a utility computing platform line EC2 running on Niagara.
Data-center in a box?! For gods’ sakes – this makes the Sunfire try-before-you-get-frustrated-and-return gimmick sound like a brilliant idea.
1. Turning a product into a commodity seems to be destiny of IT.
2. The cost of real estate is reasonable concern (especially after 11-Sep).
2. Totally agree with Gilbert Pilz that an silver colored would be cooler than black paint.
3. As for corporate cool and kosher question, consider the following entry from my inbox this morning:
The “CONTENT FILTER” filter has detected the following message using its “Incoming Policy” rule, was blocked because of Incoming Profanity Filter.
Please contact eMailReview via email to release this email if you believe it is Business related or in error.
Sender: CNET_Networks_#3.198098.3836353930353238.b@newsletters.online.com
Recipient:
Subject: Killer device drivers leave no OS safe; FBI asks ISPs to track users [TECH UPDATE]
Date and Time: Thu Oct 19 02:27:01 2006
Fuckin’ A, Dr. Carr. (But I did decide not to use that as a title of an ongoing post.)
Shivc would have a point if there hadn’t actually been any market research; but there has. I’m pretty sure the world needs a substantial number of these things. Now we’ll find out.
If IT spending is not increasing, which spending does?
Many CEO’s sarary has increased in the past a few years. Is that the only strategy for business to be competitive?
By far the best take on this is here.
Cheers.