LinuxGram

IBM’s paranoid defense of its mainframe monopoly is going to be tested again but this time if it blows the start-up out of the water, or denies it clearance to compete, it’ll be messing with an open source operation, and a European open source start-up to boot.

The European Commission, an admitted open source advocate, is already entertaining at least one known complaint about IBM’s allegedly anti-competitive behavior in the mainframe market.

The new would-be mainframe player is a Franco-American start-up called TurboHercules SAS.

It wants to commercialize the Hercules open source mainframe emulator that’s been in development by a community of IBM mainframers for the last 10 years, and sell it to mainframe shops so they can use x86- or Itanium-based PCs ostensibly for the test and development of mainframe applications.

Hercules is a software implementation of IBM’s System/370, ESA/390 and z/Architecture instruction set, but not IBM’s companion operating systems, and although Hercules has never been commercialized before, it is, for all intents and purposes, the only existing alternative to IBM.

But right away TurboHercules has got a problem because it needs an IBM operating system and IBM ties its mainframe operating systems to its hardware, a fact that should make it the subject of an antitrust investigation but hasn’t lately.

See, IBM is specifically forbidden to tie under the lingering terms of its now-dissolved 1956 consent decree with the United States government.

But IBM has so far evaded capture. Last year it bought Platform Solutions, one of the last companies to sue it for antitrust for tying the OS to the mainframe, to hush the complaint and get PSI’s cheaper, z/OS-capable Itanium boxes off the market.

TurboHercules president Bill Miller, an American who used to work for Softway Systems, the Interix Unix-to-NT interoperability shop that Microsoft bought way back when, says that, being incorporated in France, TurboHercules wrote to the head of IBM France in July asking that IBM loosen up its license enough to let the start-up function.

It wants IBM to let its mainframe customers use their 64-bit z/OS, z/VM and z/VSE – or even older 31-bit and 24-bit – operating systems on a TurboHercules configuration as IBM used to permit before its change of policy in 2000. They would pay IBM a reasonable licensing fee for the privilege.

TurboHercules hasn’t heard back yet.

Miller says there are mainframe shops and schools that would kill for such a device. Universities that can’t afford a mainframe could start teaching mainframe programming again. Mainframe sites could use it for ancillary workloads, training, demonstrations, pre- and post-processing, data preparation and archiving.

Anyway, in the absence of an answer from IBM TurboHercules turned up at Intel’s Developer Forum this week claiming to have found a niche it can occupy in mainframe disaster recovery/business continuity without violating IBM’s myriad proscriptions.

It says there’s a provision in the IBM license that allows IBM operating system software to run temporarily on an alternate machine in the event the customer’s mainframe is inoperable.

TurboHercules doesn’t need to touch the IBM OS to back up a system, merely to restore it.

TurboHercules probably can’t make a living at such a thing, which would likely be popular only with mainframes that had gone off maintenance, or sites that can’t afford a new mainframe, and none of its people have given up their day jobs yet but it’s hoping IBM will either voluntarily relent or be finally pushed to it by the regulators.

The start-up says it a press release that it “hopes to benefit from IBM’s long-standing support of open source software.” In other words it’s trying to stick IBM none too subtly between a rock and a hard place.

It wouldn’t take much to banter about the conclusion – voiced by T3 Technologies (T3T), once the second-largest mainframe systems integrator in the world, in a still-pending antitrust suit lodged against IBM at the end of 2007 – that Big Blue is nothing but a big fraud and hypocrite and “postures itself as a champion of open systems and standards,” demanding that competitors like Microsoft provide reasonable and non-discriminatory access to interoperability information, but won’t do it itself.

IBM is said to be sensitive about its image.

Anyway, TurboHercules co-founder Tom Lehmann observes that “In these tough economic times smaller mainframe shops, as well as state and local governments, are looking for ways to reduce costs and still meet their obligations in the event of a disaster.”

An unidentified mainframe site has got a prototype TurboHercules backup system installed (see http://www.youtube.com/watch?v=q8km-xOD2vc).

Due to IBM’s licensing restrictions, the demo TurboHercules ran at IDF was reportedly limited to IBM’s older MVS 3.8j mainframe operating system and SUSE Linux Enterprise Server for System z. There’s no problem running z/Linux on the thing. The demo was hosted on Windows Server 2008 on Nehalem and Itanium processors.

Hercules is reportedly spiffier in some respects than its mighty clone.

The widgetry, which can run under Linux, Windows (98, NT, 2000 and XP), Solaris, FreeBSD and Mac OS X (10.3 and later), is supposed to be able to run mainframe apps with the performance of a multi-hundred MIPS mainframe on a single Intel-based server.

Otherwise, TurboHercules means to provide corporate users with commercial services, and provides free downloads and documentation for Hercules installed on Windows and Linux host systems. Its voyage to a recent Share user group meeting suggested that there may be more use of TurboHercules in the corporate world than anybody suspects.

On its web site the start-up says that in the next few months it will be releasing “fully tested commercial distributions to match several real-world scenarios. Our preliminary assessment is that there are several important – albeit not necessarily mission-critical – ways in which Hercules can be deployed in mainframe shops.” It will be offering SLAs.

Hercules is available under the Q Public License to avoid what its community called the “political baggage” of the GPL.

Roger Bowler, the creator of Hercules, is a TurboHercules co-founder.

By the way, IBM filed two motions for summary judgment against T3T on August 30. T3T has until October 14 to respond. That’s likely the last obstacle IBM can throw at it. If it survives, its antitrust case could get to trial sometime in Q1. T3T is also waiting to hear whether the European Commission will open a formal investigation of IBM on its complaint.

Figuring that the cloud may be the “most significant shift in technology since the outset of the Internet,” IBM is moving to ensure it gets its piece of what could be a $66 billion business in three years.

It’s taking a workload-by-workload approach.

To start, it’s targeting the enterprise – meaning big, largely Blue accounts – with offers of a virtual desktop – either a Microsoft desktop or some stand-in – and the widgetry to move test and development to the cloud.

Unlike Amazon, IBM’s not interested in the individual developer; it wants all of a major account’s test and development, which it argues absorbs 30%-50% of a company’s infrastructure though 90% of the time it’s idle.

Blue says its widgetry is based on two years of research and hundreds of client engagements. There’s no mention of interoperability with other clouds like Google or Salesforce.com; that will have to be left to third parties like Vordel.

IBM will be offering to run these cloud services from inside its own data center or to set up a private cloud using a client’s own infrastructure.

And, for those who want to go it alone, it’s got CloudBurst 1.1, a pre-integrated cloud-to-go with the hardware, storage, virtualization, networking and management needed to build a private cloud.

Users running their own clouds will be able to access IBM’s cloud for short-term additional resources.

CloudBurst will ship at the end of the week at prices starting around $200k. The entry-level configuration includes a 42U rack, a BladeCenter chassis, eight Intel cores on three blades plus an eight-core management blade, attached storage and middleware.

IBM’s public cloud will be located at its recently announced production-level “Cloud Delivery Center” in Raleigh, North Carolina, where it’s got a bunch of Intel blades, VMware virtualization and its own management widgetry.

The hardware there will come to include mainframes and IBM’s own p blades. Over time that will mean its Rationale and Eclipse tools and the Jazz platform as well as Xen and KVM virtualization.

IBM’s cloud CTO Kristof Kloeckner has found that large accounts aren’t as reluctant to share cloud infrastructure if they know the guys they’re sharing it with are, say, fellow Fortune 5000s and not riffraff.

IBM estimates that the cloud can save up to 73% of the energy used to power traditional desktops and laptops and up to 40% of the support costs.

It reckons its Smart Business Test Cloud can save customers 50%-75% on capex and licensing expenses and 30%-50% on operating and labor costs while reducing defects from faulty configurations and poor modeling by 15%-30%.

IBM is also setting up a subscription public cloud version of the Desktop Cloud. It’s partnering with Desktone, Quest and Wyse.

Kloeckner said IBM would offer 99.5 SLAs on the cloudware, upgradeable to 99.9 – “infinitely better than what you get with desktops today.”

Back about 15 years ago after Lou Gerstner saved IBM from falling off a cliff and Sun was still a comer Gerstner and Sun CEO Scott McNealy used to have breakfast, lunch or dinner more than ever so often and Scooter invariably pitched Lou on buying Sun. Sun’s stock was then, if memory serves, around $72 a share. Scott wanted IBM to pay $125.

Well, if reports are to be believed Sun is finally going to realize an even better 125% markup from IBM except the end numbers are way smaller.

Somebody leaked a story to the Wall Street Journal Wednesday saying IBM was negotiating to buy Sun and that the pair could come to terms this week if the deal didn’t blow up.

The AP waded in hours later with a source of its own, perhaps the same as the Journal’s, saying the same thing.

The reputed price – before it started inching up in one retelling or another – was $6.5 billion in cash, a bit less than double Sun’s beaten-down market cap on Tuesday and more than IBM has ever spent on an acquisition before. The most money it ever spent was $4.9 billion last year for Cognos.

Reuters fixes the total value of the deal at $8 billion less the $1.4 billion in cash on Sun’s balance sheet – (we thought Sun had more like $2.6 billion but we may have lost track of how many times it had to tap its bank account) – and we can all sit around and debate whether Sun is worth any of it, sort of like the Obama bail-out.

A lot of folks think Sun could be vacuumed up for a lot less than the now reputed premium of $10-$11 a share; a lot of other folks wonder why IBM would be interested at all.

Maybe there’s a rival for Sun’s hand or maybe IBM thinks so.

Sun has reached a stage of permanent eclipse – its customers – it had a lot of the Wall Street firms – have disappeared with the recession and it desperately needs to get bought; reportedly it’s been peddling itself around town for the last year and HP turned it down.

IBM is supposedly interested in Sun because with Sun it would again be bigger than HP, which is currently the biggest computer company on the face of the earth. Sun did $13 billion last year.

The acquisition of Sun would make IBM a big-time hardware peddler again; it could suck up another 9% market share in servers; importantly the high-end servers IBM is so fond of, albeit Sparc-based, especially now that Cisco’s coming in; and every computer dollar taken in makes for seven dollar in service revenues – so much for HP’s uppity $13.9 billion acquisition of EDS.

With Sun in tow, IBM would have a 45.6% server market share to HP’s 29% using IDC’s numbers.

There are many reasons why an IBM acquisition of Sun seems foolhardy – it’s counter-directional for starters – but presumably antitrust concerns aren’t one of them at this point. That’s why places like the Metropolitan Club exist to discretely sound out reactions. The regulators, at least this side of the pond, may be satisfied that with Cisco coming into the high-end server market numerical balance is maintain. And both IBM and Sun have friends in the European Commission since they ratted out Microsoft.

The tape sector, another Sun business in secular decline like its Unix business, is another possible concern to regulators.

One of the reasons IBM is believe to be interested is Java, which IBM largely developed. While not a big moneymaker, Sun’s continued ownership of the stuff has been a big irritant to IBM. Maybe it knows how to make it pay even at this late date.

Sun brings a host of other IP and software assets to the equation like its MySQL acquisition and OpenOffice – Big Blue’s big on open source – and maybe IBM might be willing to trade its frumpy AIX operating system for Sun’s Solaris. Together the two would have 65% of the dwindling Unix server market with only HP left to compete again, but then Linux’ cannibalization of Unix would be even more of an in-house sport.

And Sun scientists with their over-generous $3 billion R&D budget may feel more at home at IBM than the perpetually non-monetizing Sun. It’s idle to get into cultural differences of ponytails and pin-stripes now when people need jobs. Anyway they share a common distaste for Microsoft, whose boss is already cherishing the idea of the pair struggling with dislocating integration for the next couple of years.

Users could stop worrying Sun will crash.

People close to IBM say if it does do the deal it would initially let Sun be before it dumping people right and left. However, Thomas Weisel Partner told the Wall Street Journal IBM might have to cut a billion dollars in costs out of Sun to bring it in line with the profitability of IBM’s own hardware business. Barclays Capital thinks the number is closer to $400 million and another 5,000 jobs lost to get accretive in a year.

Sun of course is in the process of dumping another ~6,000 jobs, roughly 18% of its workforce, after losing $209 million in the December quarter on sales down 11% to $3.22 billion. Its stock, pre-rumor, was down around 70% over the last year.

Sun generally gets 40% of its money from servers, 40% from services and 15% from storage with software contributing next to nothing. Revenues peaked in 2000 and it’s lost money in five of the last eight years and cut 7,000 jobs.

IBM had $13 billion in the bank a few months ago.

In a fascinating move to co-op the cloud, IBM is proposing to rate other people’s clouds, people like Amazon, GoGrid, Mosso and FlexiScale.

Hopefully it will do better than Moody’s and Standard & Poor’s did with the bonds created out of sub-prime loans.

It’s unclear why anyone would submit to such a search by IBM, but – if they do and pass – IBM is proposing to reward them with a “Good Housekeeping Seal of Approval” or “Resilient Cloud Proven” logo.

This resiliency stuff means IBM validates their facilities, apps, data, staff, processes and business strategy. IBM claims it will weed out untrustworthy providers.

It says the influx of newfangled cloud services – where everybody’s middle name is suddenly cloud – has created a challenge for customers evaluating the move to the cloud.

It’s promising reassurance – benchmarking and checks on design, infrastructure hardening, redundancy and ongoing monitoring and management.

IBM says it’s done this kind of stuff forever and has strict standards for service quality – everything from infrastructure design to process excellence.

Allscripts, the healthcare player, is IBM’s first validation customer. It’s delivering a data recovery service through the cloud.

Meanwhile, IBM Global Services is going into the cloud consulting business.

Cloud computing should save 80% on floor space and 60% on power and cooling costs, and deliver 3x asset utilization, but that doesn’t mean people are comfortable with the idea.

Enterprise customers still have security, data portability, compliance, privacy and reliability concerns.

IBM says it’ll do both industry-specific work to assess the TCO of building private clouds and help people set up their own clouds – or move data and application off-site or into a hybrid model.

Its research suggests public and private cloud won’t compete – they’ll be complementary – but naturally it’s going to steer people into its own cloud centers.

Under the code name Project Yun, Chinese for cloud, IBM is also working on new menu-driven, cloud-delivered, horizontal and vertical business services like healthcare in the first place and CRM and supply chain management in the second.

It says the widgetry is being piloted with customers by its China Research Lab and that Wang Fu Jin Department Stores, one of China’s biggest retailers, is trying some of the services.

IBM will use the stuff to create industry-specific clouds aimed at mobile and telecom, social collaboration, finance and banking (well, what’s left of it anyway), government, healthcare, education and IT management.