No matter how much Sun Microsystems Inc doesn't like it, selling Linux servers based on Intel-compatible processors complicates the pat, streamlined, self-reflecting Sparc-Solaris marketing message which has been the company's hallmark since it got into the server business for real in the early 1990s and
stormed the enterprise in the mid-1990s,Timothy Prickett Morgan writes
Any time an executive utters the phrase "what you have to get through your head" while talking to journalists on an open mike, as Stephen DeWitt, the former CEO of Cobalt and now vice president and general manager of content delivery and edge computing at Sun, did yesterday during a conference call, you know that a company is frustrated that something it is doing is not being correctly understood out there in the world.
That's why DeWitt and Anil Gadre, vice president and general manager of Solaris development at Sun, held a conference with analysts and press yesterday in San Francisco. To clear the matter up, once and for all.
But as the executives themselves point out, the matter is far from clear.
The boundary lines between operating systems and the layers in the network where they run - edge, application or database tiers - are all being blurred, and now they are being blurred at Sun, too. It's not Sun's fault, exactly. It's the fault of all the Unix vendors, who could not come to terms to create a single Unix operating system in the late 1980s and early 1990s.
This left the door wide open for Bill Gates and his Windows server aspirations, and made the counterculture, open source reaction of the Linux community inevitable. If it hadn't been Linux, it would have been Workplace OS (remember that?), FreeBSD, BeOS, or something else filling in the gap. Sun is only partly to blame, you see.
That said, a certain amount of confusion is understandable when trying to understand this new Sun, a company that at first sneakily adopted Linux - how many times have we heard that the operating system in a server appliance doesn't matter, and never quite believed it - by buying Cobalt Networks a little more than a year ago. Then last month Sun said that it would create a line of general purpose Linux-X86 servers that would sell along side inexpensive Sparc-Solaris edge servers. Is this the beginning of the end of Sparc-Solaris?
Gadre scoffs at the idea that Sun's support of Linux is somehow about Sun eventually dropping Solaris for Linux in the long term. "We would be ridiculously stupid to squander the Solaris asset," he said in yesterday's briefing to clarify Sun's position on and adoption of Linux. So, to recap: Sun is not abandoning Solaris. Sun is not abandoning Solaris. Sun is not abandoning Solaris.
As for the skeptics who doubt that Sun can withstand the intense pressure to move to Intel chips over the long haul, DeWitt says forget that line of thinking, which he characterized as being something that pops up whenever journalists need something to write about. (There's a certain amount of truth to that, but that doesn't make the question any less valid.)
DeWitt made Sun's position on Sparc abundantly clear: "We have the second-biggest chip design team in the world - second only to Intel - and we are not wavering from our course. The Linux announcements we have made have nothing to do with the viability of Sparc and everything to do about expanding Sun into new markets."
Sun contends that its sales force is not confused when it comes to presenting Linux and Solaris solutions to customers, and that its sales force is not just intimately acquainted with how to sell both product lines, but that the Sun sales is what has, in part, convinced Sun that it needs a more general purpose Linux server offering to get into new accounts and to preserve sales of edge servers in accounts that, for whatever reason, want Linux solutions. "This is not something we thought up sitting around the campfire," explained DeWitt.
"This is not a hedge or a bet. This is something that Sun's customers are asking for, and the important thing is to embrace customer requirements."
DeWitt tried to outline what Sun was up to and why. "The edge of the data center is a highly evolving market," he said.
This evolution includes high-density 1U servers and will eventually include blade servers that offer better density than rack-mounted servers while being easier to configure and manage, thus lowering total cost of ownership.
The basic idea is that the services that will be deployed at the edge of data centers will have to be deployed using dirt cheap servers and infrastructure because, to be quite honest, most of these services aren't going to be worth much money individually.
In a post-dot-com world where money is not freely printed as stock, transactions cannot cost more to process than the value of the transactions themselves. If it costs $1 to process a transaction for a $10 item that costs $8 to market and $2 to create, then the company providing that product will go broke. Hence the desire for very inexpensive infrastructure to support businesses.
That's why Sun bought Cobalt to gain some quick expertise in building integrated, cheap edge servers that are really pseudo-application servers that are constructed on the nearly free Linux operating system and dirt cheap Intel-compatible chips and chipsets. Cobalt's real value has been the way it integrates Apache, Sendmail, open source databases, firewalls and other programs; a Cobalt server lets internet newbies figure get up and running without having to learn TCP/IP.
Selling 100,000 Cobalts a year is a great business, but Sun (we guess) wants to sell many more than that, and that is why it has decided to bite the bullet - or rather, the Penguin - and make more general purpose Cobalt Linux servers in uniprocessor and dual-processor models. We can expect to see such machines, says DeWitt, "very early in the second half" of this year. He wouldn't say if Sun would be using real Intel chips or clones from AMD or Transmeta or some other supplier, and he wouldn't say what chipsets the company was using in the machines as well.
The simple fact is that more and more companies want Linux because they don't want to have to worry about what Sun, IBM Corp, Hewlett Packard Co, or Compaq Computer Corp are going to do with their Solaris, AIX, HP-UX, or Tru64 operating systems.
It's not that operating systems are not control points - they sure as hell are in a world that has spent more than $1 trillion in 35 years developing applications running on proprietary and Unix operating systems - it's more that companies are sick of caring.
The Cobalt Qube and RaQ servers, while appropriate for small enterprises and service providers, are not as expandable a product as competitive general purpose servers. So, Sun will have its Cobalt engineers cook up a Red Hat-compatible Linux distribution used only for its own X86 servers and will drop the one Sparc-Solaris mantra in exchange for a new one: Linux is Unix, so Sun still wins. (Many Sun execs, when cornered, will admit that Linux is an inferior implementation of Unix, and they sure won't volunteer that Linux is just as incompatible with Solaris as it is with AIX, HP-UX and Tru64.)
That Sun - and indeed any of the Unix vendors - has to embrace Linux at all is their own fault. The ghost of Unix wars past comes back to haunt. Sun, and perhaps all the Unix vendors, would have been better off if a united Unix had killed Windows before it was even born.
And now, no matter how proud they are of their Unix operating systems, they all have to employ Linux as a means to fight against Microsoft's Windows platform. Sun has no choice in a world that is asking for Linux and is not asking for Solaris. The latter you have to sell, but increasingly - and particularly for infrastructure workloads - the former only requires sales reps to book orders.
© ComputerWire.com. All rights reserved.