What the changes at Sun mean for the customer

In the short term - meaning tomorrow- the major changes at Sun: a renewed focus on Solaris/SPARC, make to order mfg with direct shipping, more emphasis on Sun Ray, customer focused ready-to-rock appliance computing - are all good news. Look a bit deeper, however, and the handwriting on the wall says much of what Java brings to the server is going, going, gone.

For Sun employees - and customers: Sun is the brand, innovation is the product, Oracle is the company - but what's that really mean for customers?

Obviously not everything's settled, but three things are very clear for the short term:

  1. Make to order and direct ship are long, long, overdue - and great for customers.

    In a make to order system, the money you don't spend on distribution and parts inventories goes to sales compensation and customer support engineering. I doubt customer costs will be much affected, but the structure of Sun's costs making up the customer price will change to put more bucks behind keeping the customer happy, informed, listened to, and on side.

    Most importantly, the guy you talk to about buying a Solaris/CMT machine, will actually know what that is - and he won't try to sell you a PC instead.

  2. Selling appliance computing focused on application delivery is what Sun should have been doing for years, and the best possible news from Sun for the tens of thousands of mid range businesses previously forced to choose between living with Wintel's limitations or accepting IBM's pricing.

    Two weeks ago a customer with 500 employees and a real need for a lot of ERP/SCM features didn't qualify for Sun's attention and so faced a choice between laying off staff to pay for an IBM solution or going Wintel and moving a lot of staffing dollars from revenue generators to IT support.

    Today (well, RSN anyway) Oracle will sell that customer what he needs: a pre-configured rack with applications, site specific configuration, installation, and support services all on one bill - just like IBM, but with at least one zero chopped off on the right.

  3. Pushing Sun Ray is the right thing to do - for customers (especially big application customers), for the company, for the entire tech industry - and Oracle is going to do it. That appliance I mentioned above? they're going to be asking the customer "do you want rays with that?" and probably offering a cost break on support for those who do.

These changes are happening now, but there are big, longer term, changes coming down the pipeline too - and I want to speculate a bit about what one of the more distant might be.

I have a personal interest in a Sun research project called Wonderland - because the ideas behind it have the potential to become integral to any advanced enterprise communications environment. What Wonderland does is a create a networked virtual world environment you can carry your iPad (or whatever) into just as you would in real world meetings to take notes or share materials with others. Since it's largely scriptable, Wonderland lets you spread meetings across both space and time - giving subscribers to a particular instance almost office-next-door access to others without much regard for time zones or distance.

Wonderland uses the Darkstar gaming engine - and because Oracle's killing Darkstar, the wonderland folks are feeling a lot of pain.

Darkstar does what its creators set out to do: provide a gaming platform enabling developers to back end hundreds of thousands of game users with a grid of little machines running sub-scale OSes without limiting the interactions between players.

I suspect Oracle killed it mainly because they saw no market for it, but the reason it deserved to die was that it successfully solves a problem Sun doesn't have - and my speculation is that we'll see a lot more work based on this reality entering the R&D pipeline this year and next with product effects two or three years down the road.

The issue is this: what darkstar achieves, providing seamless horizontal scaling through box count, can also be achieved, although currently only at a smaller scale, using Solaris/CMT directly - and the hardware transaction commit in the ROCK technologies coupled with ZFS block level control will later reduce overheads for this kind of application by at least another order of magnitude.

So here's the speculation in chief: when the key players start to figure this out, they're going to see it applies to a lot of other things. Most politically in the Oracle world, the same logic applies to the core components in the BEA products Oracle bought for over $8 billion a few years ago.

That's a long term change, of course - but the bet is simply this: we're going start seeing a slew of major Oracle product overhauls and rationalizations as it becomes obvious that the most complex, over head intensive, (and commercially valuable because proven), components coded into those products are simply not needed on Sun gear.

Show Comments