Sun puts “Sparc” into Web 2.0

Company aims its Redshift initiative at expanding network loads

Moore’s law and the commoditization of server boxes had most of us believing that the days of big iron were over. For a while at least, it looked as if Intel and Windows Server would take over the heart of the datacenter.

Perhaps even Sun Microsystems believed this would happen. How else to explain its adoption of x86 chips to the detriment of its high-performance Sparc product line?

The argument posited that it would be insane to spend $50K, for starters, on a Sun Solaris box when an Intel cluster at a third the price would do. Actually, in 1996, a high-end Sparc “mainframe” could cost more than $1 million.

Although many analysts — and nearly all vendors with Windows hardware and software to sell — endorsed the idea wholeheartedly, I don’t think IT ever really bought into it.

If the poet William Blake said, “You never know what is enough until you know what is more than enough,” then IT departments never got there. You can never have enough performance, as IT knows. And now, with the advent of Web 2.0, SaaS, widgets, YouTube, and streaming video over your cell phone, that truth is self-evident.

Enter Redshift. Sun’s answer to the newest demands on the network, one that could in fact indicate an actual shift taking place in the IT industry.

The name of Sun’s initiative comes from an astronomical phenomenon — the shifting of light toward the red end of the spectrum due to the expansion of the universe. Get it? Sun, universe, expanding — as in scaling to meet the needs of an expanding network. You could say that with Redshift, Sun wants to be the Sparc behind the Web 2.0 network infrastructure as it evolves.

Of course, you can still deploy x86 rack-mount and blade servers, even with the Sun logo. But what Sun seems to be saying with Redshift is that fast, cheap, and easy no longer scale well enough. Wintel will just have to wait a bit longer before it becomes the heartbeat of the datacenter.

To meet these expanding network needs, Sun’s latest products are built around Solaris ZFS, a 128-bit file system that yields almost unlimited data capacity, says Peder Ulander, vice president of marketing for Web 2.0 at Sun.

At the risk of becoming a commercial for Sun, here’s how I see Sun reading the market as revealed by its Redshift initiative.

Project Blackbox is literally a virtual datacenter in a box. It houses eight server racks in a 160-square-foot shipping container. At 38 units per rack, it has the capacity for more than 700 CPUs, 2,000 cores, or 8,000 compute threads. And the entire network system architecture and management network are included inside.

Sounds to me as if Sun is trying to transform the datacenter into a commodity product. Yet commoditization usually occurs when a high-demand market is saturated with enough vendors that they end up competing almost entirely on price.

Obviously, datacenters in shipping containers aren’t quite there yet, but what Sun seems to be saying with Blackbox is that, as everything moves onto the network, owning and running a datacenter will not remain in the hands of a few suppliers. It can’t, simply because too many companies will find themselves in need of those levels of performance.

For example, SaaS solutions require a powerful infrastructure to deliver a user experience equivalent to that found on the desktop. And if more than 50 percent of all new software startups are delivering products via the Web, then the demand placed on datacenters will increase tenfold.

Code-named Thumper, aka Sun Fire x4500, is similar to Blackbox only it is a hybrid data server/storage solution in a box with 24TB in 4U of rack space. It combines the functions of 2 dual core AMD Operton processors, network fabric and switch, and SATA storage in a single integrated system.

Code-named Streamstar, aka Sun Streaming System, provides broadband in a box and is designed for the delivery of high-definition broadband TV, IPTV, and steaming video to the home.

Project Darkstar is a box optimized for the gaming industry.

Network.com is a Sun subscription-based service that delivers pay-per-use CPU cycles.

We have all seen companies selling products “in a box.” Not many have succeeded. Some were too far ahead of their time; others failed because some technologies just change too rapidly to be put into a box.

Nevertheless, I think Sun is reading the market correctly, and while I’m not 100 percent convinced that Project Darkstar and Streamstar will be roaring successes, Project Blackbox has a lot going for it, as does Thumper.

Whatever the outcome, it will be interesting to watch the market and see how products and partnerships develop to meet the new demands on the network.

Perhaps IT will get what it has always wanted: performance to the max.