|
|
|
|
|
       |
 |
|
|
The iDeal iSeries, Part 1 by Timothy Prickett Morgan Don't you wish some days that you could design your own iSeries product line? Me, too. I've been thinking about this a lot over the years. I've watched IBM launch 10 generations of AS/400 and iSeries servers. IBM used to do a pretty good job keeping the OS/400 platform in step with the industry. But around the time of the debut of the Apache machines, in 1997, the company stopped trying to keep up with entry and midrange alternatives in offering excellent bang for the buck.
This was a bad idea, and I have been saying that for five years, in one form or another. It does not take a genius to figure out the consequences of IBM's course of action--worldwide iSeries platform sales (including servers, memory, and disk storage) are about half their normal run rate. If we are lucky, IBM might hit $1.6 billion to $1.8 billion in sales for 2002. I say "we" because our careers are based on the iSeries, and IBM's is not. Whatever IBM's strategy was in the late 1990s, it hasn't worked out for either IBM or the customer base, even if that strategy of keeping OS/400 platform prices artificially high did allow business partners to cash in on their customer bases--at least for a while. Not anymore. Just as we were heading into a recession in late 2000, the current management team for the iSeries inherited these problems and the constraints that the business partner channel and IBM's bookkeeping impose. They have done the best that they can to keep the iSeries humming along without upsetting the business partner channel. This is, I think, one of the big the problems that has hindered the iSeries platform. Apparently, I'm not the only one who feels this way, because the Green Streak deal on the Model 270 and Model 820, which gives customers a 50 percent price break on two iSeries configurations, didn't happen because business partners wanted it or IBM's chief financial officer wanted it or Sam Palmisano wanted it. Green Streak is a trial balloon, meant to test the price elasticity in the OS/400 server market. Unfortunately, this is the worst possible time to do such a test, since no midrange shop wants to spend money on anything. From the outside--which is where I sit, and you do, too--the evidence suggests that there is something terribly wrong with the way the iSeries is packaged and sold, because it simply is not selling like it ought to. There are some probable causes for this. First, IBM is afraid to change its marketing pitch about the iSeries because, to put it bluntly, the iSeries is financing its nearly insane price cuts on pSeries iron, so it can try to kill Sun Microsystems and Hewlett-Packard in the Unix server market, which represents about half of worldwide server sales each year. Second, IBM is not putting its best technologies in the box to attract customers. It's that simple. The AS/400 used to get all the good stuff ahead of the RS/6000, and now the iSeries hardware seriously lags the pSeries hardware. The saving grace here is that the iSeries is a better operating system with better workload management and logical partitions. And this is what gets me hot under the collar. IBM knows how to build great chips and great software--and it used to know how to make great disks and other peripherals, until it started cutting corners somewhere--and it ought to know better than you or me about how to build an ideal iSeries. I shouldn't have to write this story and the one that will follow it. Nonetheless, because what IBM is doing is not working, we have to present an alternative vision for this platform that will not only get the iSeries back in synch with the server market but also get it out on the front end again. When I sat down to do this story, I had to resist the knee-jerk reaction to go look at how the existing iSeries line is configured and see how I could tweak and change what IBM already sells, to make the iSeries line more accurately reflect the products other vendors are selling in the entry and midrange markets. This is too much of a pain, so I just wiped the slate clean and started from scratch. Spiritually, this felt pretty good, and I highly recommend such an approach to IBM's iSeries developers and marketeers. Go back to the drawing board. To that end, I have created a suggested new iSeries product line, which I am dubbing the iDeal iSeries. The key word in that name is "deal," and by that, I mean that the 250,000 AS/400 and iSeries customers in the world should see a revamped iSeries product line and say, "Wow! What a deal!" The last time this happened was 1995, when IBM moved from CISC to PowerPC RISC processors and offered the biggest performance and price/performance bump in its history. My iDeal iSeries line is based on two assumptions. The first is that IBM can use the new PowerPC 970 chip it is making for Apple Computer to use in its Mac desktops and X-Serve servers (which should be ready for market sometime in the middle of next year) in a new entry iSeries server line. IBM will make millions of PowerPC 970 chips, and it will be a high-volume, low-cost component that can and--here's the important bit--should be the heart of a new low-cost iSeries entry machine. The second assumption is that IBM should just stop fooling around and rebrand the entire pSeries line--including the pSeries 650 that is coming in a week or so--and slap iSeries labels on them and get this technology out there in the base. And if the business partner channel says it can't handle the launch of a new machine, then IBM should start selling them direct, right off its Web site and right into customers' hands. It is better to sell a great iSeries than to listen to excuses about why a less-than-great iSeries didn't sell. If there is no reception on this channel, then maybe it is time to change channels. Let Dell sell them if IBM can't. The important thing, as EMC has learned the hard way in the midrange disk array market, is to sell, and that is why Dell is codeveloping and cobranding a line of low-cost disk arrays with EMC. When you do these two things, you get a line of iSeries machines that might look like the line I have created in my iDeal iSeries table (click here to view table). There are not yet performance stats on the PowerPC 970 processors, but they are true Power4 processors, and I assumed that, given the limited amount of L2 cache and main memory I put in the entry iDeal iSeries boxes, their CPW ratings would be artificially limited enough. The only governor on OLTP transaction processing in this box is main memory, but I nonetheless assumed that only about a third of the theoretical processing capacity of the PowerPC 970 chip would be available to these hypothetical iSeries machines because of memory constraints. There are no interactive governors or any other kinds of governors in these entry boxes, which I am calling the Model 381s, Model 382s, and Model 384s--that is an increment up from Model 270 on the first two digits, plus the last digit telling how many processors are in the box. I abhor any kind of governor, as you well know, but IBM would have to do something to differentiate the capabilities of the entry iDeal iSeries from the Power4-based iDeal iSeries machines I show as the Model 831, Model 851, Model 871, and Model 891. These latter machines are exactly and precisely the pSeries Model 630, 650, 670, and 690, right down to the base configuration and base list prices. The rule I want to impose on iSeries pricing is simple enough--whatever IBM charges for the pSeries, it charges for the iSeries. I have, as the price/performance metrics in my table show, priced the PowerPC 970-based iDeal iSeries machines very aggressively. (Note that these prices are for based hardware only. We'll get into software pricing on these machines next week.) An alternative to putting a memory governor on these entry iDeal iSeries boxes would be to do something totally unexpected, such as shipping each machine with base two logical partitions equipped with Linux-based Samba print and file serving or--and I really like this idea--shipping the machines from the factory with a high availability solution bundled right in the box. Why shouldn't the iSeries be the first platform with high availability built in, under the covers? Let customers choose their own high availability solution, and dedicate half of the processing capacity in the box for this purpose. If customers don't use the high availability software, they can't get at that extra capacity. The other differentiator I put in my iDeal iSeries machines is that the entry models can run either OS/400 or Linux natively. What I mean is that customers can, if they want, completely wipe out OS/400 on the box and install Linux on it, and it works like any other Linux server. And if customers buy one of these machines to support Linux and decide they want to run OS/400 on it, they buy an OS/400 license from IBM, wipe out the Linux on the box, and load OS/400. Similarly, the Power4-based machines support either OS/400 or AIX natively--I am not talking about OS/400 PASE here--and support Linux within OS/400 or AIX partitions, since Linux cannot scale to the heights in the Power4 machines. I'm sure I am not the only one with ideas about how to reinvent the iSeries. Send me your ideas at tpm@itjungle.com. Next week, I will tackle the thorny issue of software pricing on these iDeal iSeries machines, and further down the road I will present a price/comparison of these mythical machines to the current iSeries line and Wintel, Lintel, and Unix alternatives.
|
Editor
Contact the Editors |
|
Last Updated: 11/4/02 Copyright © 1996-2008 Guild Companies, Inc. All Rights Reserved. |
BCD INT'L