In the last five years, the System i has made impressive performance gains, picking up faster processors, more memory, more disk, faster I/0, and Capacity on Demand — all of which has been crammed into boxes with smaller and smaller footprints. Of course, the System p, which shares much of the same hardware guts, has also made notable gains during the same time period.
Most experts peg the performance growth for multiple server types, including PCs, at similar rates of improvement during the last five years. Thus, IBM’s System i effectiveness hasn’t gained or lost at rates much different from the rest of the industry. Let’s forget System i competitors, though, for a more troubling question:
Is the System i scaling faster than its customers?
Just a couple of years ago, you could buy an iSeries Model 825 with three- to six-way POWER4 microprocessors, up to 48 GB of memory, 58 TB of disk, and a CPW that could range from 3,600 to 6,600. Today you can get a System i5 520 that handles almost as much memory and disk (32 GB and 39 TB) but that boasts a CPW that can top out at 7,100.
The 825 fits the middle of IBM’s lineup in terms of size, and the scalable 520, which comes in a wide variety of editions, fits the lower end of IBM’s array. Now, assuming that a customer would upgrade within the same relative tier, an 825 user would most likely move into the 550, which tops out at 14,000 CPW.
For that kind of transition to make sense, a customer would have to grow a heckuva lot with more orders, more employees, bigger databases, and a bunch of new workloads. Otherwise, a relatively stable and slow-growing organization wouldn’t need such a big box.
“The processor, the memory, the disk — if you compare it to where we were at even a few years ago, a customer can now buy a box that’s twice as powerful for half the price,” says Stan Staszak, director of System i products for Sirius Computer Systems, an IBM Premier Business Partner. “If you want to post the same revenue numbers, you have to double your volume — you have to sell more workloads on the server just to break even.”
This leads to an even more troubling question: Do these changes have anything to do with IBM System i financial declines in recent quarters?
At first glance, the numbers look nasty. Compared to the previous year, System i revenue was down 22 percent in 3Q06, down 7 percent in 2Q06, down 22 percent in 1Q06, and down 18 percent in 4Q05. The gains from the first three quarters of 2005, by the way, only boosted revenue 1 percent over 2004.
Looking more closely at 1Q06, smaller System i models actually sold pretty well, IBM says, but they also generated smaller profit margins and less revenue than IBM’s big iron. In addition, IBM reduced pricing to improve price/performance at the low end, which gave IBM another revenue-producing challenge over previous quarters.
In his 1Q06 report, Mark Loughridge, IBM senior vice president and CFO, notes, “System i had strong volumes and revenue growth in its low-end and Express models, which target the small- and medium-business market space. We expect to see improvement in the midrange and high end of the product line as the year progresses.”
Did the pickup ever really happen? It's hard to say because the second- and third-quarter financial reports in 2006 don't show it. Might increases in overall power factor in?
Going back to our 825-520 example and customer organization scalability, I posed the question to Jim Herring, IBM’s director of System i Product Management and Business Operations team: Does IBM have customers who “upgrade” into a relatively smaller box, for example, from an 825 to a 520?
“Most customers who are upgrading out of 825s are going into 550s because they want the extra headroom,” Herring says. “There are a few customers [who move from an 825 into a 520], typically customers who didn’t activate many of their standby processors, who can fit pretty handily into a new 520. But most of the 825s are upgraded into the 550s, and we have quite a few upgrading into the 570s as well.”
Clearly, most Enterprise Resource Planning (ERP) vendors want to sell into small- and medium-size businesses, primarily because that’s where they can find new sales opportunities. IBM wants in, too. There just aren’t that many large companies that haven’t already committed to their ERP platform of choice. When companies used to look at ERP solutions, those solutions generally came with very large price tags. Part of the cost was the software, of course, but another part of the cost traditionally came from the big hardware needed to run it all. So what happened when IBM began selling System i editions such as the System i 520 Solution Edition for SAP, which as of October started at a rock-bottom price of $35,000?
The answer, which includes similarly priced solution editions for other application suites, such as Oracle’s JDE EnterpriseOne, is that IBM has to sell more volume to maintain System i revenue. IBM has little choice in the matter. It must sell to smaller customers or barely sell at all.
Has IBM’s successful server consolidation play affected System i sales? If a company previously owned several distributed System i boxes around the country or the world, what happened when that company consolidated System i solutions onto a few larger System i boxes? The current consensus in this situation is that it’s not happening enough to change IBM’s revenue picture, and even when it does happen, there’s a tendency for such System i customers to add other workloads to their systems anyway. Basically, if you’re savvy enough to look at server consolidation within the System i, you’re smart enough to grow your overall System i solution to include new workloads. So the theory goes.
Still, in the third quarter of 2006, IBM lost revenue on operating systems, which isn’t surprising given IBM’s Linux push and the increasing ability of an i5/OS to do more than ever before.
Is the System i world somehow less able to recover from processor improvements? As the PC server world, for example, has tended to grow horizontally, has the System i world grown up — vertically — into a bigger machine that can do more than before? Yes. Although this doesn’t mean that System i customers are downgrading into less expensive (but more powerful) boxes, it does mean that the growth tends to stick within fewer, larger boxes. Capacity on Demand is a prime example of this. Take, for example, the notion that one Windows server is used to running one basic workload even if overall CPU utilization is fairly low. If a company wants to grow those Windows applications, there’s a good chance that the business will simply add another server, whether it’s a standalone or blade, and keep adding new servers to meet new growth. It’s kind of an organic process that seems to make sense. Got a new need? Meet it with a cheap new box. My point isn’t that server proliferation is a bad management practice; rather, I'm saying that server growth in these kinds of organizations rewards low commodity pricing by manufacturers. Does it ever and will it ever come back to bite these manufacturers? You bet.
In the System i community, on the other hand, you simply use Capacity on Demand. Turn on another processor, and boom, you have an instant upgrade. Here’s where it bites IBM: If a company knew that it would experience 20-percent growth every year in processing needs, it would lease a system that was powerful enough to last three years, allowing a bit of headroom in case something unforeseen came up. So for the first two years, processor utilization would be pretty low, which would mean that the business was paying for processing power it wasn’t using — bad for the customer, good for IBM. Now that a customer can make a lease decision based on performance pictures for the first, second, and third years, by turning on additional processors only when necessary and by factoring in things such as retail holiday sales bursts, the customer can pay less overall for performance.
So, yeah, IBM’s own innovations can decrease the company’s bottom line.
Some critics of IBM have accused the company of overcharging for the System i, of using it as a cash cow and milking it for all it’s worth for as long as possible — essentially bilking its most loyal customers. This theory has several holes. First, if the System i is so profitable for IBM, why isn’t IBM pushing it harder? Second, if customers are overpaying, why are they so loyal? Third, if this theory is true, how has the System i consistently managed to have the lowest annual operating costs, producing the lowest cost per unit of work done? (See “Simply the Best,” September 2006, article ID 20639.)
So, in addition to delivering faster processors, which IBM uses in the System p line to compete head-to-head in the Unix world, IBM strives to boost overall performance all the time even if it seems to adversely affect the company's System i quarterly reports.
“It behooves profitability to meet the needs of your most loyal customers, so if you’ve got customers who have been with you for years, you realize that sales aren’t a quarter-to-quarter issue so much as a generational issue,” notes Charles King, principal analyst for Pund-IT, Inc. “You might take a short-term revenue hit, but you’re doing something to show your customers there are benefits to being loyal, and in the long run you win.”
So, back to our original question of whether the System i is out-scaling its own customers: It sure seems that way, but in the long run, does it really matter? No, not as long as IBM continues to deliver innovation. The best and brightest companies will eventually catch up.
Links:
[1] http://systeminetwork.com/author/chris-maxcer