RTC is short for Real Time Computing, not Reaming The Customer, which is what a lot of the early versions of capacity on demand and utility pricing certainly did when the idea made its debut in the Unix server market in late 1999 and was perfected in 2000 and 2001. By announcing capacity on demand pricing, Hewlett-Packard and Sun Microsystems were trying to do the right thing, and IBM followed up a year later in its Unix line.
While being able to activate capacity that is sitting idle in a server is a great idea – and a lot more appealing than either going through a normal upgrade process or provisioning a new server, both of which can take weeks or months – the initial premium that these three companies charged for capacity on demand configurations ranged from 20% to 25% above the cost of just acquiring the capacity. This was a big premium to ask even during the dot-com build out, but one that could be justified. When the bubble burst, there was no way to justify the extra cost.
Moreover, capacity on demand was a one-way deal. You could turn on capacity and use it, but once you did, you were stuck with it. Having the ability to activate a processor to cover peak workloads was a big improvement, to be sure, but workloads go down as well as go up. Over time, after plenty of complaining from customers, the big three Unix vendors figured out that they needed to have capacity on demand for other components (such as memory, I/O, and storage) to balance out added processors; they also were told point blank by customers that they wanted to have truly metered usage, like an electric utility, where they truly pay for the processing capacity they use.
The big three Unix server makers have stopped talking about capacity on demand and utility pricing, they are no longer able to command that premium, and they are more focused on trying to promote the idea of grid-utility computing as a service. Sun and IBM want you to buy excess capacity from the grids they have built in their data centers to do number-crunching jobs, and they want you to buy servers just like you used to for enterprise applications.
IBM, Sun, and HP are, of course, perfectly happy to create a leasing or financing arrangement that smells like utility pricing, but these deals are generally too complex for busy data center managers and chief financial officers to weigh against the way they have always done things, which is to buy more server than they need right now so they will have it in the future when they do need it.
But that doesn’t mean VPDPs and CFOs want to do it this way. In fact, says Mark Feverston, vice president of platform marketing for enterprise servers at Unisys, the company has been talking to customers in its ES7000 and mainframe bases as well as within other companies’ installed bases to get an idea of how it might do a better job of putting together a capacity on demand offering on Xeon and Itanium servers aimed at Windows and Linux shops. The ES7000 RTC boxes are the result of its studies.
The IT industry has been talking for years now about how the average server utilization is in the range of 15 to 20%, and Mr Feverston says that the CEOs and CFOs have found out about this and they want something done about it. He says further that the problem gets more acute as chip makers like Intel roll out faster and faster processors; most workloads don’t scale as well as theoretical performance boosts for processors. People just want to buy the amount of capacity they need that the moment – and they do not want to buy any capacity they do not need, explains Mr Feverston.
Unisys has offered very granular capacity on demand and metered pricing on its Clearpath mainframes for three-and-a-half years. On the ClearPaths, customers could activate temporary and permanent capacity on demand (based on 24-hour usage increments) as well as spare processors (which normally ship in these machines, even if customers don’t know it) for use as a disaster recovery option. The option allows customers to use the spare processors for 30 days in the event of a disaster, like a primary mainframe engine burning out. Unisys was able to offer this on the ClearPaths many years ago because it has absolute control over both the hardware and operating system in its mainframes.
With the Windows and Linux operating systems and the Xeon and Itanium platforms, Unisys has anything but complete control. But its engineers have put a sophisticated service processor inside each ES7000 server, and with the right tweaks to the microcode, you can create a way to activate and deactivate processor capacity in a running Windows or Linux server. This is exactly what Unisys has done with the ES7000 RTC boxes. But as history has shown, being able to turn capacity on and off is only half of the battle. Customers want this technology, but they do not want to pay a premium for it, says Mr Feverston, and that means Unisys had to come up with a pricing mechanism that is easy to use and makes sense to bean counters as well as IT managers.
With the ES7000 RTC boxes, Unisys is making one thing simple and clear: it will never cost more to add capacity on an ES7000 RTC box than it does to buy it on a regular ES7000 machine. It is up to Unisys to come up with a scheme to price processing capacity and meter it to make that sentence true. So what Unisys is doing is selling processing capacity in 15-day increments (it will not divulge the price, which varies depending on the speed and type of processor).
This is a fairly large chunk of processing capacity. Unisys keeps track of how the processors in the ES7000 RTC are used, and after you have activated the processor four times, you own it. There is also an initial capacity planning and setup fee that amounts to a 10% surcharge on the ES7000 server, but Unisys subtracts this from the fees for processor activations.
So the net-net is that customers are essentially renting processing capacity to own it, and the 10% fee is like a security deposit that is good toward the last month’s rent. Companies have no contractual obligation to activate the spare capacity in the box, ever. (But that means you would be out of that 10% setup fee, which is purposefully meant as an economic barrier to sort customers who really don’t need utility pricing from those who do.)
Equally importantly, companies do not have to go to a Unisys salesperson or their CFO to get approval to do an upgrade. All of this is handled on the front end with the ES7000 RTCs, and a select number of personnel in the data center are authorized to activate capacity that their companies agree to pay for.
