isham research
Platform Solutions emphatically rejects the suggestion that its product will be an IBM zSeries emulator - but if it looks like a duck, walks like a duck, swims like a duck, flies like a duck, quacks like a duck and lays duck eggs, it's pretty duck-like.
z/OS does not run natively on Intel processors.
And it looks like Platform Solutions is determined to take over where UMX Technologies left off, with a stream of disingenuous statements. Their press release on 2 December 2003 referred to them as a "supplier", for instance, although GA is not planned until sometime in 2005 - if then. Surely a supplier is someone who has supplied?
Platform Systems Inc, in conjunction with key business partners, is a leading provider of systems for mainframe applications
How can a company be a provider - let alone the downright ludicrous "leading" - when it hasn't shipped anything? Shades of UMX all over again. How gullible can journalists - and venture capital companies - be?
The company is a spin-off from Amdahl's "Stingray" skunkworks project, which was an attempt to implement the /390 architecture on Intel's notorious Merced. Notorious because Intel failed - spectacularly - to meet every Merced deadline. So Amdahl pulled the plug - even though the system was working well under simulation - and some of the team involved elected to spin the project off into a company they founded.
That was in 1999 - it's now 2004. Five years. So where's the beef?
Platform Solutions is a somewhat secretive company. For the last four years the only tangible sign of its existence was a couple of desultory pages on the web. This has changed with a well-publicised second round of external financing, a presentation at SHARE, and an article in Linuxworld. So what is going on?
The Linuxworld article - printed on 21 August 2004 - was appallingly researched (if indeed it was researched at all) and creates some very misleading impressions. Among other inanities it contains the following:
There's been no one to challenge IBM's unprecedented mainframe monopoly since Big Blue ran off the last of the PCMs years ago when it got light years ahead of them on the silicon front, forcing them to close up shop and slink away.
Well, hardly. Hitachi was actually using IBM's chips in its mainframes - and built the z800 both for IBM and to market itself in Japan - so how could IBM have been "light years" ahead? In any case, light years are a measure of distance, not time.
At the same time IBM saved mainframes from being entombed in the La Brea tar pits along with other fossils, an exercise that has proven to be worth the effort.
Actually, it was IBM's early-1990s decision to rescue its mainframe via aggressive adoption of CMOS processors that ultimately doomed the PCMs. The PCMs' demise was a result of, and not a precursor of, IBM's increased competitiveness within a constrained market.
In its latest quarter, mainframes were credited with contributing 15% of IBM's profits, the tip of a huge multibillion-dollar business that has been IBM's to do with what it will.
Mainframes? 15% of IBM's profits? Want to buy a bridge? It may well be that the mainframe sector contributed 15% of IBM's profits, but the majority came from the 85% and higher gross margin on mainframe middleware such as DB2 and from services. Mainframe hardware, these days, is heavily discounted to miniscule margins and it's entirely possible that some of IBM's smaller mainframes are sold at a loss on the expectation of future MES revenues.
Until now.
Enter Platform Solutions Inc (PSI), a company that was started in 1999 by Amdahl's old core engineers using an estimated half-a-billion-dollars worth of Amdahl IP and bent on making an industry-standard Itanium 2 system capable of running MVS, z/OS, S/390, Unix, Windows and Linux simultaneously.
"Some of" Amdahl's old core engineers would create a more accurate impression. "One or two" might be even closer. Think one stall of a diner. Half a billion? Imagine the wrath of Amdahl's stockholders if $500,000,000 in intellectual property walked out of the door. And the TIDA material that Amdahl bought and licensed from IBM was most scrupulously destroyed (formal shreddings and burnings in Sunnyvale and Dublin with lawyers present) when Amdahl left the business. In any case, intellectual property's value is capped with what you can achieve with it. There simply isn't half a billion in gross margin in the market - even if PSI got 100% of it. When IBM bought Lotus for $3.5 billion, the intellectual property was valued at a mere $88 million. The true value of Platform Solutions' intellectual property is much closer to a good lunch than to $500,000,000.
And they may yet regret the choice of Itanium - its future seems by no means assured: "In the second quarter, Hitachi shipped 15 Itanium servers, Unisys moved 11, Fujitsu shipped 5 and NEC moved 38, according to Gartner. HP sold 4,789 of the grand total of 5,665 boxes. We couldn't help but wonder if the Fujitsu box on stage was one of the five it sold in the quarter."
Itanium is hardly an industry standard - Microsoft has even dropped support for it in Windows 2003 Server - but the more worrying aspect is the lack of mention of IBM's Itanium offering. Business partners selling mainframe emulation have found that building their systems on IBM xSeries servers has many advantages - the customer is not losing the IBM hardware connection; one IBM machine is simply being replaced with another. Although HP's products are reliable, the IBM label still counts for a lot in mainframe sites.
PSI is playing for the half of the $5 billion-a-year mainframe hardware business that IBM barely cares about, all those mid-sized and middling large companies that aren't the Fortune 100.
$2,500,000,000 a year? Wrong planet - a few percent of that at most. IBM has had no true low end mainframes for well over a decade and there is practically no churn in the market - none of the operating systems are supported and no one wants to convert to the ones that are. 8 MIPS OS/390 users, for instance, are stopped in their tracks by IBM's refusal to license z/OS under ESL Ts&Cs. Germany had over 1200 small IBM mainframes a few years ago - that number is today less than 500 and shrinking fast, and the overwhelming majority are VM/VSE, not MVS. This market sector is now handled by IBM zSeries Business Partners - some of whom (e.g., Comparex) are starting to give up on the market.
IBM can't be bothered dirtying its own hands on anything less than a Fortune 100, according to PSI president and CEO Michael Maulick, himself an IBM veteran with over 20 years of service (ADSM is his). IBM passes their business on to stand-ins.
Such is IBM's focus on the rarified high high-end and such is its fascination with 450-MIPS building blocks that it has stranded less prestigious accounts without an affordable migration path past the G4 and G5, Maulick says.
Things being what they are, he says, second-tier companies are being forced to pay $8,000-$10,000 per MIPS where their better-off betters are paying only $2,000.
Hardware is not the issue in most cases - most companies in this sector couldn't afford a z890 if it was given to them. The problem is software costs - and especially ISVs.
With a Flex-ES license at around $25k and a server to get about 30 MIPS out of it somewhat less, the "second-tier companies" are paying around $1500/MIPS or so. $10,000/MIPS? Someone needs to read the "z/Series Remarketed Hardware Pricing Guide For Solution Providers/Tier 1 & Systems Integrators". The problem is zSeries software which is indeed hideously expensive in $/MIPS terms at the low end. IBM cuts deals to "grandfather" software costs over up to four years, but the ISVs don't.
Knowing that PSI is coming, IBM threw these second-tier accounts a bone in July [2004] when it introduced the dumbed-down z990 it dubbed the z890. It pulled a few chips out of the 990, lowered its clock speed, crippled its cache, reused its packaging and, knowing that PSI would start at 26 MIPS, said the 890 would do that.
Actually, it was announced on 7 April [2004]. A significant date - the fortieth anniversary of IBM's first System/360 mainframe announcement. You'd think a date like that would be noticed. Someone's not paying attention and someone else isn't doing the research you would expect of a responsible journalist. Or even an irresponsible one. But that's Unixworld for you. There's a more plausible explanation for the 27 MIPS entry point, which some IBMers were discussing with analysts as early as April 2003. But if Platform Solutions is limiting itself to a 26 MIPS start point, it will be taking on IBM head-to-head all the way. Fundamental Software sells an awful lot of smaller systems - such as 8 MIPS systems that qualify for ESL licensing and Partnerworld for Development systems - and this helps enormously to build critical mass for support.
Maulick claims it's not a solution and, at roughly the same price as the 990, certainly not a reasonable one.
"Roughly the same price as a z990"? Laughable.
PSI aims to address this disenfranchised market, believing that it's also elastic. It figures, for instance, that there are high-end Sun users that would love to go back to the safety and security of the MVS RAS and can't get there from where they are.
If the market were elastic, IBM would lower software prices to increase hardware volumes. It hasn't.
Oddly, even though it would benefit immensely from the software revenue, IBM denies that z/OS RAS is available under emulation. In a single-minded focus on box numbers (which is how the relevant IBM executives are measured) IBM will FUD against any z/OS solution that does not run on zSeries hardware by saying that z/OS RAS is only available where there is "true synergy between the operating system and the platform" - i.e., z/OS is only stable on real IBM hardware.
Besides that, of course, migrating a Sun user to pSeries is one hell of a lot easier, cheaper and more competitive than migrating to zSeries.
At this point, PSI is only thinking about carving off a piece of the mainframe iron market, not the $24 billion service and software business connected with it.
PSI is playing for big stakes - and ultimately a big IPO - remember that 60%-70% of the world's mission-critical data is still on the mainframe - and it has a few little friends helping it take on the IBM monopoly.
It has fewer and littler friends than it had last time around. The list of backers in the first round was Blueprint Ventures, Fujitsu, Intel Capital, InterWest Partners, and InvestCorp - but in the second round only Blueprint Ventures, Interwest Partners, and Investcorp were named. Where are the technology companies - Fujitsu and Intel? And note the downright crafty sequence the press releases on PSI's site are listed in - check the dates.
(Note: Intel joined the second round a little later, amount undisclosed)
It just got an undisclosed amount of money in second-round financing - undisclosed because it doesn't want IBM to know what its resources are and try to drain them, but it must be a tidy pile because it's supposed to take PSI past profitability in, oh, 2006.
A more likely explanation is the precise opposite. Fujitsu is absent - it seems most likely that the VCs have promised just enough to cover PSI for the current product plan, to see if they can get any of their money back. That must include the promise to ship in 2005 - anyone remember Encore out of Fort Lauderdale? Let the (first) customer beware.
The money came from Goldman Sachs, which led Blueprint Ventures, InterWest Partners and InvestCorp, all of which must be delirious at the thought that PSI will have no competition.
No competition? Recent information suggests that IBM is still around.
It is quite nonsensical, of course, to imagine companies such as Fujitsu being absent from such an ostensibly lucrative opportunity when there is supposedly no competition. Just how stupid are the executives who run one of the most successful IT companies on the planet? In fact, of the potential 4,000 to 5,000 sites world wide, Fundamental Software's partners already have 800-odd and this will only increase with "real" DASD attachment in FLEX-ES Release 7.
As Maulick points out, it's not exactly a field where VCs are funding a hundred start-ups. The barrier to entry, he figures, is roughly $500 million, the value he puts on the tools, diagnostics and testing widgetry PSI got from Amdahl.
A $500 million barrier to entry is frankly ludicrous. The Hercules Open Source effort has built a zSeries emulator for little more than the cost of coffee, beer and Internet bandwidth - and also made it portable. UMX did the same thing commercially, but still under the $10 million mark, and actually shipped product to customers before going bankrupt.
PSI has wrapped core microcode and patented virtualization mojo into the Universal Server's architecture and Goldman is evidently heady with delight at the prospect of running z/OS on an open systems server. It speaks of the widgetry as "reinvigorating the PCM market in a 64-bit world."
Perhaps IBM is not so heady with delight at the prospect of running z/OS on an open systems server. So far it has expressed great reservations - not only because the integrity of open systems is hard to assure, but also because its software licensing policies (and internal audit) require strict and precise guarantees of an individual system's performance. Software capping is a major issue - and until you get IBM to agree to price and license its code, you can't sell a box.
And so far, IBM has refused to license z/OS for commercial use in 64-bit mode under emulation. Perhaps it can be induced to do so - the ISVs may force the issue - but that may take time, and capital injections can be burnt up quickly.
Maulick describes it as "reincarnating Amdahl" without Amdahl. Amdahl used to have 24% of the mainframe market. If PSI gets just 10%, Maulick will count it highly successful.
It will be nothing short of a miracle. Stand-alone mainframe PCM companies such as Amdahl are no longer viable. Those that are still active, such as Hitachi Data Systems and EMC, derive the majority of their income from other segments and would not be viable if the IBM mainframe market were all they were serving. Building a completely new company based on a mainframe-only business case is utterly out of the question.
Which leaves the other alternative - using IBM's existing Business Partner structure. Fundamental Software's major partners do this - but some of the battles can be hard to win. A zSeries Business Partner with a 27 MIPS prospect will naturally face pressure from IBM to sell a real IBM mainframe. The FLEX-ES alternative - even if built on an IBM xSeries and using properly licensed IBM software - will not count to the zSeries quotas upon which the executives are paid. In contrast with the PCM years, two battles have to be won - one at the customer and another one at the Business Partner.
The number of such Business Partners is also somewhat limited. In the UK, for instance, there are really only two - one now owned by major Fundamental partner T3 Technologies - in Germany around half a dozen. Most had their fingers burnt by UMX, which makes them even more cautious. They operate to an extent geographically - so if you want to sell a system in a particular city, you HAVE to sign up the relevant Partner. What if they are (as most are) already a reseller for a Fundamental partner?
PSI, which will sell multi-processor servers that range into the thousands of MIPS, thinks it's got a world-class alternative to IBM's z/Series and p/Series. It believes it can argue against the wisdom of investing in a single-purpose proprietary platform that, it says, limits the exploitation of open technologies and eats up IT resources trying to solve management issues.
It's an elegant theory - but you can do it today with FLEX-ES and almost no one does. An n-way Intel-based server can be licensed for zSeries emulation on one processor with UNIX or Linux workloads running on the others. There is terrific flexibility to do this in the FLEX-ES solution - but the simple fact is that very few exploit it. Nobody cares. The vast majority of FLEX-ES systems are stand-alone replacements for older and smaller - mostly obsolete - IBM mainframes running just a few critical legacy applications.
The original Stingray was a Gerry Anderson creation - an fantasy aircraft that became a submarine in a children's puppet show. Such multi-function devices are usually failures. Although there are such things as amphibious cars, very few are sold. Cars make lousy boats and boats make lousy cars. Putting cameras in phones now seems to have peaked as well.
Thousands of MIPS sounds good; but systems in that space are usually in Sysplexes - which implies Coupling Facility support - with huge farms of shared storage devices and massive channel connectivity.
In fact, IBM is not PSI's main competitor. Fundamental Software is. And Fundamental has a head start with rich functionality. Emulating a zSeries to execute z/OS is only part of the job - the real cost savings come from eliminating maintenance costs for older peripherals such as external DASD and telecommunications gear, and permitting extensive automation at low cost. Internal DASD storage has been a feature of IBM's lowest-end mainframes almost from the start. The way it works in the market is that users examine IBM's hardware offerings and decide for or against. Once they decide against, they invite in the emulation suppliers. UMX is now gone, so their choice is effectively between multiple suppliers of Fundamental's technology - and there are indeed differentiators such as service and support. This is the space in which PSI will need to be active.
And for the final inanity:
Ironically, if IBM Software weren't so captive to IBM's hardware interests, IBM Software chief Steve Mills would probably be beating a path to PSI's door.
You would think that the "IBM Software chief" would be the one who decided whether or not IBM Software were captive to anything? Or does the division determine policy for its Vice President and General Manager?