Don’t Believe the Myth-information about the Mainframe: Part 1

 Misconception about the mainframe continue in this article many of those are put to bed. Aivars Lode

By Janet L. Sun,
Immediate Past President, SHARE Inc.

When I heard about the HuffPost article highlighting a video debunking the myths that Hollywood has been repeating about the mainframe, I was cautiously optimistic.  Unfortunately, the writer chose to use only one reference book, and focused on the negative points.

Here at SHARE, we believe that the mainframe is the most secure, lowest cost and best performing mixed workload computing platform on the planet.  SHARE continues to serve the mainframe community, helping to show our members the best practices for managing the mainframe environment and optimizing the value that the mainframe delivers.  The mainframe is the core of the computing environment for many companies.

“Most companies don’t use mainframes” they said – Seriously?  SHARE, the mainframe community, represents more than 20,000 individuals from nearly 2,000 companies.  Those companies include: state and federal government agencies, universities, retail, energy, manufacturing, banks, and insurance companies.  More specifically:

• 96 of the world’s top 100 banks, 23 of the 25 top US retailers, and 9 out of 10 of the world’s largest insurance companies run System z
• Seventy-one percent of global Fortune 500 companies are System z clients
• Nine out of the top 10 global life and health insurance providers process their high-volume transactions on a System z mainframe
• Mainframes process roughly 30 billion business transactions per day, including most major credit card transactions and stock trades, money transfers, manufacturing processes, and ERP systems.

That doesn’t exactly sound like a technology that’s no longer in use, or even going away anytime soon.  So, what are the other most common myths about the mainframe?

1. Mainframes are old
2. Mainframes don’t run modern applications
3. Mainframes are expensive
4. The skills to manage mainframes are not available or you need more people

Mainframes are Old?

Well, the mainframe is celebrating its 50th birthday next year.  But, there have been generational differences between the mainframe that was introduced in 1964 and today’s mainframe.  The automobile is more than 100 years old, but no one suggests that automobiles are an old or outdated technology.

Are the cars of today different from the cars of 1964?  Absolutely.  Well, today’s mainframe is faster, has more capacity, is more reliable and more energy efficient than the mainframe of the 60’s, 70’s, 80’s, or even those delivered three years ago in 2010.

The new mainframe delivered in 2010 improved single system image performance by 60 percent, while keeping within the same energy envelope when compared to previous generations.  And the newest mainframe which shipped in 2012 has up to 50 percent more total system capacity, as well as availability and security enhancements.

It uses 5.5 GHz hexa-core chips – hardly old technology.  It is scalable to 120 cores with 3 terabytes of memory.  Clearly larger (more capacity) and faster than anything available in the 60’s, with a smaller physical footprint and better energy consumption characteristics.

IBM has a corporate directive for every generation of mainframe: each successive mainframe model must be more reliable than the previous one. Incremental and breakthrough improvements have been made over 20 generations of mainframes. Fault tolerance, self-healing capabilities, and concurrent maintainability are characteristics of the mainframe that are lacking in many other systems. The integration of mainframe hardware, firmware, and the operating system enable the highest reliability, availability, and serviceability capabilities in the industry.

Mainframes Don't Run Modern Applications?

Mainframes have been running Linux workloads since 2000 and the Linux workloads on the mainframe are growing.  From IBM’s 2012 Annual Report – “The increase in MIPS (i.e. capacity) was driven by the new mainframe shipments, including specialty engines, which increased 44 percent year over year driven by Linux workloads.”

The mainframe also has a specialty processor that is specifically intended to run Java workloads.  How about Hoplon Infotainment running their TaikoDom game hosted on System z?

You say that green screens are ugly?  There are graphical interfaces and even iPhone and Android apps that put a pretty face on the green screens for those who those who are trying to use business applications.  More and more, interfaces that the general public is familiar with and comfortable with are being utilized even in business contexts to make access to the mainframe easier and more transparent (how many people are accessing a mainframe on a regular basis today and don’t know it? - most of them!)

Those who manage the mainframe often prefer the green screens.  These are incredibly fast interfaces that can deliver sub-second response time.  When is the last time you clicked your mouse and got sub-second response from your Java application?

What about “cloud”?  The “cloud” is actually an online computer environment consisting of components (including hardware, networks, storage, services, and interfaces) in a virtualized environment that can deliver online services (including data, infrastructure, storage, and processes), just in time or based on user demand. By this definition of Cloud Computing, System z has been an internalized cloud for decades.

System z has been “in the clouds” for more than 40 years!  Whether you are thinking cloud computing (e.g Infrastructure-as-a-Service) or simply server virtualization, System z is a great platform.  Instead of running dozens of virtual images, a mainframe can run hundreds.  And besides Infrastructure-as-a-Service, you could also implement Platform-as-a-Service or Software-as-a-Service.

Starting in 2007, IBM embarked on its own server consolidation project called “Project Big Green”.  They consolidated 3900 servers onto 16 mainframes decreasing energy and floor space by more than 80 percent.

The electrical power ($600/day vs $32/day), floor space (10,000 sq ft vs 400 sq ft), and cooling costs for those mainframes were less than those of distributed servers handling a comparable load.  In addition, those mainframes required 80 percent less administration/labor (>25 people vs <5 people); “Mean Time Between Failure” measured in decades for mainframe vs months for other servers.

Need more?  Ask City and County of Honolulu about their cloud implementation on System z.  They had issues with planning, deployment, and maintenance of hardware and services in a Windows environment which took weeks.  They created a mainframe cloud environment and offered ‘Software as a Service’ to other departments.

They saw immediate benefits:

• Planning, deployment and maintenance of hardware and services can be done inhours vs. weeks
• Lower costs enable the expansion of database services for a fraction of the distributed costs
• Improved performance and response time to end users
• Sharing of resources with other state wide jurisdictions

Ready for more mainframe misconceptions? Janet Sun continues the ‘Don’t Believe the Myths about the Mainframe’ series with the forthcoming parts 2 and 3 on the President’s Corner. Stay tuned!

Will customers buy into Oracle's modern-day mainframes?

With Oracles gouging of customers it will only be time before more and more customers move off this architecture. Aivars Lode

Oracle is making a big play to cash in on the simplification of enterprise IT environments with "engineered systems," the collection of vertically integrated appliances that debuted five years ago with Exadata.
While Oracle insists customers who enter into the Big Red cocoon will reap all kinds of benefits, some may be hesitant to buy into Oracle's one-throat-to-choke mainframe model.
Oracle's engineered systems are like modern-day mainframes. They include all the hardware-like processors, memory, spinning hard disks, flash-based hard disks, Infiniband connectors, and power cords and the software like kernels, operating systems, hypervisors, file systems, databases, middleware, management tools, and applications - all in one handy dandy container.

The IT giant says its engineered systems are "pre-integrated at the factory" in Oregon, where they're assembled using "best of breed building blocks".

While the hardware is mostly industry standard and commodity level, Oracle says its "secret sauce" is the design that goes into engineered systems, which are, in effect, private cloud platforms that Oracle will service and maintain for customers through a single "premiere" support agreement.

Since there's no need to cobble together IT pieces from different vendors, there's less validation and testing work involved, and future software upgrades are less likely to topple what would normally be a system very sensitive to change.

Patches are applied automatically by Oracle once a quarter. They also feature automated error detection at multiple levels of the stack, just like IBM mainframes, and a five-minute response time commitment from Oracle.

Best of breed

"We're taking those best of breed parts and vertically integrating them," Oracle president Mark Hurd said recently during an interview with Independent Oracle Users Group.

"We do that to deliver extreme performance and total cost of ownership. We design the technology explicitly for the stack. We fine tune it. We optimise it across all of those layers so that we can deliver better performance, better reliability, better security… more manageability, etc."

In an industry that breeds complexity, this mantra of simplifying enterprise IT components has a powerful appeal. And to be sure, Oracle isn't the only system vendor taking such an integrated approach. IBM, HP, EMC, NetApp, Cisco, and Intel all have irons in the consolidated server fire, which Gartner predicts will account for 35 percent of server sales (by revenue) by 2015.

Trends are cyclical in the IT business. Mainframes were supposed to die off 20 years ago, when the rise of the PC-led organisations to distribute IT processing capacity. Now organisations want to centralise IT processing again, hence the shift back to the mainframe model that made IBM a ton of money between the 1960s and the 1990s.

"Oracle's strategy, at the end of the day, is they want to be IBM," says R "Ray" Wang, principal analyst and CEO of Constellation Research. "On the high end, there's only a few players left [in the mainframe business], and IBM is one of them. And Oracle is going to be the other one if they keep doing what they're doing."

Do the sales numbers bear them out?

During the second quarter of fiscal 2012, the company says it sold 700 all-in-one servers, a 70 per cent increase over the previous quarter. Most of this is probably the five-year-old Exadata database appliance, but Oracle didn't break down the numbers.

Sales of Exalogic, which was designed and tuned to run Oracle Fusion middleware and Java applications, is also basically doubling every quarter, according to Oracle. Exalytics, which is pre-tuned at the factory for analytic workloads, was just released at the end of 2012.

The smaller Oracle Database Appliance is making headway with smaller customers, but the three other systems - Oracle Big Data Appliance, the ZFS Storage Appliance, and its latest Sparc T4-4 SuperClusters - have yet to make a big splash.

Engineered systems: a possible turnaround solution

The company is betting heavily on its engineered systems to turn around its hardware business, which has dropped precipitously since it acquired Sun Microsystems. While some of the components are there, especially around database, storage, and clustering, it still has some work to do with Fusion middleware and Fusion applications, analysts say.

When the stack is built out, it could be extremely lucrative for Oracle, which is banking that the "attach rates" of software, support, and service deals that go along with engineered systems sales will drive a lucrative, long-term stream of maintenance revenue. And that, of course, is where the real money is for Oracle, which recognises nearly half of its total revenue from software maintenance and support.

The benefits of engineered systems are clear for Oracle. As customers entrust more of their enterprise IT stack to Oracle, the California-based company benefits through growth in its high-margin maintenance revenue stream. It gets a shot at replacing HP, IBM, and Dell gear in Fortune 500 accounts that already run its software, and consolidating its footprint. It's a big time mainframe play, with a 21st century cloud twist.

But the benefits to customers are not as clear. Oracle contracted the Edison Group last year to write a whitepaper that tries to quantify some of the customer benefits of taking a vertical stack approach. In the paper, titled "The Optimized Stack: Reducing TCO through Vertical Integration," Edison compared Oracle's pre-integrated two-socket and four-socket servers against similarly equipped servers from HP and IBM.

Edison concluded that the three-year total cost of ownership of Oracle's two-socket machine was 61 percent cheaper than the TCOs of HP and IBM servers.

The bulk of the savings came from lower licensing and maintenance costs for the operating system and hypervisor components of Oracle's systems (Oracle Linux and Oracle VM) compared to the IBM and HP servers, which used Red Hat and VMware.

The benefits

"On a two-socket system, these two expenses alone [OS and hypervisor] can represent as much as 71 per cent of total TCO over a three-year period," the report states. "Why pay for virtualisation when it comes for free?" Oracle asks in its brochure.

However, it's unlikely that bundled price savings alone will entice customers to adopt Oracle's engineered systems. Despite the sales pitch, Oracle isn't giving away its Linux OS and hypervisor software. Big companies never pay list price, especially when buying from Oracle, so it's not fair to assume they wouldn't bargain with Red Hat and VMware for better terms.

Industry observers say the biggest challenge that Oracle faces with engineered systems is convincing customers that putting all of their enterprise IT eggs in Oracle's basket is a good thing.
"The danger is, it's one throat to choke, and you're becoming more and more dependent on Oracle," says Wang, CEO of Constellation Research.
"Oracle wants to be able to get to this point where customers are saying, 'Look, you're already a super strategic supplier to us' and here's what you do.' [But] it'll make it harder to get off, because everything is there at the infrastructure layer."
Wang says the customers who are most receptive to "one throat to choke" are Oracle's die-hard Red Stack fans who face huge integration headaches and costs.
"These folks are never going to leave Oracle,” he says. "They started with a database. They adopted the financial apps and then they bought the middleware. For them now, it's a complete IT consolidation play. Soup to nuts, these are die hard fans, and they're saying, 'Oracle, thanks for putting this all together. We're eliminating five other vendors, and delivering it for less than the cost of five other vendors'."

But what do the acquired customers think of the lock-in?

Less receptive to the engineered systems approach are customers Oracle obtained through acquisition.
"These guys say, 'I bought the best of everything. There was a reason I wanted Sun or I wanted IBM or HP,'" he says. "Then Oracle comes in, like a rude guest, and says, 'Oh, by the way'."
"That vendor lock in is scary," says a former VP in Oracle's engineered systems program, who asked not to be named in this story. "I don't think that's any customer's real desire. To be locked in scares customers in a significant way. They want to have strategic partners. They want somebody who understands all the different components in a highly virtualised, highly heterogeneous environment. [They don't want somebody] saying 'Hey everything has to be a Red Stack or a Blue Stack.'"
The most likely Oracle customers to give engineered systems a try are net-new customers, and those ready to make the move to Oracle Fusion Apps. After all, two-thirds of Fusion Apps deployments are being made on Oracle's cloud, which is composed in part of engineered systems.
But there are only 400 Fusion Apps customers out there, compared to tens of thousands of E-Business Suite, PeopleSoft, JD Edwards, and Siebel customers. These are the customers that Oracle needs to adopt engineered systems, but don't expect rapid adoption.
Whether it's fair or not, Oracle has a reputation that precedes it, and the question of lock-in is not something to be taken lightly. While it backed off its initial plan to eventually consolidate its business apps on Fusion when Fusion launched in 2005, it has ruffled the feathers of some of its acquired customers, notably JD Edwards EnterpriseOne customers who run their ERP on IBM fully integrated Power Systems (AS/400) servers.

Blue Stack

In 2010, Oracle announced that, at the end of 2013, it will no longer sell the Blue Stack of software (WebSphere and DB2) needed to run the ERP system on IBM gear, and from September 2016 will no longer take support calls for Blue Stack software issues. This is a powerful incentive for EnterpriseOne customers to move to Oracle's stack.
JD Edwards and PeopleSoft customers will approach the consolidate Oracle stack slowly and cautiously, says Jeff West, the president of the Quest International Users Group. "Replacing any or all of the components can be expensive for existing customers and has to be weighed against the potential benefits of a unified Oracle stack," West says.
Existing customers probably won't consider the engineered systems approach until their next hardware refreshes and system upgrades, he says. "For net new JD Edwards or PeopleSoft customers, the value proposition may be slightly better and an easier transition to make since they are already making a sizeable investment in a new ERP and may be able to realise the unified stack benefits quicker," he says.
Oracle has the size and funding to play the long game with engineered systems. That may actually benefit its enterprise software customers, which are stretching their system upgrade cycles to five years and beyond, and like making strategic architectural decisions about as much as a root canal.
"Whatever evil image of Oracle people have, they're not forcing customers," Wang says. "Oracle is one of the few [tech] companies with a big R&D team. They're actually developing and building. Other people buy companies and don't build on top of them. They just let them die.
"Whether you're looking at cluster or storage, or looking at the database appliance or Exadata, what Oracle is really saying is, 'Look, we're just dropping these in for you. And if you're not getting cost savings, nobody would buy it to begin with'."
The Register approached Oracle to comment on this article, but it was unable to respond by the time of publication. ®

Disruptive technologies

Very interesting that Eric Schmidt Googles Executive Chairman acknowledges that the cloud is now what the mainframe was and still is. Aivars Lode. Avantce

http://www.mckinsey.com/insights/high_tech_telecoms_internet/disruptive_technologies

Total Cost of Epsilon E-Mail Data Breach Could Reach $225M, Including Up to $45M in Lost Business, According to New Report by CyberFactors

Cost of having a data breach. Aivars Lode

Epsilon and Amazon episodes demonstrate how risk associated with cloud security issues is not being adequately addressed, company says

NEW YORK--(BUSINESS WIRE)--E-mail services firm Epsilon will face years of repercussions and up to $225 million in total costs as a result of its recent data breach, a massive event that indicates the often overlooked risk of cloud-based computing systems, according to a research report released today byCyberFactors™, a cyber risk analytics and intelligence company.

“While the attractiveness of the cloud model is hard to refute, the economics of business risk for cloud providers and their customers can no longer be ignored”

The recent breakdown of Amazon’s cloud computing services that disrupted services to popular sites like Foursquare and Quora is another example of a cloud failure that could prove extremely costly in the long run – and a hint of more troubles on the horizon, CyberFactors asserted in its CyberBrief on Epsilon.

According to research conducted by CyberFactors, the Epsilon breach may have affected 75 companies or 3% of Epsilon’s customers, not 2% as previously reported, and could eventually cost these companies as much as $412 million, for a total event cost of $637 million. Further, CyberFactors conservatively estimated the number of affected e-mails in the Epsilon breach at 60 million.

The total cost of the Epsilon breach – including forensic audits and monitoring, fines, litigation and lost business for provider and customers – could eventually run as high as $3 billion to $4 billion, according to CyberFactors, given that the compromised e-mail addresses could be used by hackers and phishers to gain access to sites that contain consumers’ personal information.

“While the attractiveness of the cloud model is hard to refute, the economics of business risk for cloud providers and their customers can no longer be ignored,” said Regina Clark, Research and Analytics Director, CyberFactors. “With the cost of technology failures rising at an accelerated rate, the Epsilon event suggests a much more profound financial risk environment is now upon us. Cloud companies would be wise to think more like banks, insurance companies and hedge funds, and not just aggregators of the world’s precious data and technology dependencies.”

Some other results of CyberFactors research on the Epsilon breach:

• 51% of the costs related to the Epsilon data breach will occur in year one, 42% in year two, and 7% in year three and thereafter
• Loss of revenue related to customer churn as part of the Epsilon breach fallout could range from $6.1 million if just 1% of customers left, to $30.7 million if there were 5% churn.
• CyberFactors research shows that since 2005, data events have cost individual affected companies in the range of $5.5 million to $12.8 million, depending on the industry and assuming no liability claims.

Ellison aims his first Oracle 'mainframe' at Big Blue

Even Oracle now knows the only way forward is a mainframe class server. Aivars Lode

T5 takes the lead from Power and x86, Big Larry claims

By Timothy Prickett Morgan 

Larry Ellison has launched the first mainframe-class machine that he can correctly say he made sure came to market, and now he is going to take a run at IBM's mainframe and Unix server businesses.
What's more, it looks like he will to be able to make some credible arguments as to why customers running Oracle software – and indeed any mission-critical app that runs on any Unix – will run better on the new Sparc T5 and M5 servers.
Oracle announced the new Sparc servers at an event in San Francisco today, and El Reg already gave you the feeds and speeds on the new T5 and M5 processors and their respective systems. At that event, Ellison touched on some of the salient characteristics of the new servers, but he spent most of his time explaining how Oracle would keep the pedal to the metal, pushing performance even further after taking the lead from Intel and IBM in terms of throughput-per-processor and bang-for-the-buck against IBM's Power Systems lineup.

At first blush, it seems a bit peculiar that Ellison should care so much about hardware – but maybe not.
When Ellison cofounded Oracle decades ago with Bob Miner, IBM mainframes and DEC VAXes were the main machines people used in commercial computing. And Oracle has always run its software on IBM mainframes and is well acquainted with their virtues: security, I/O throughput, reliability – and their excessive costs.
Ellison paid $7.6bn to buy Sun Microsystems a little more than three years ago to get Java and Solaris, but also to get hardware engineers who could build systems that would push back against the onslaught of IBM in the Unix space. And contrary to lots of talk, Ellison has maintained his commitment to both Sparc and x86 iron. Ellison likes to build and control his entire stack, and Sparc processors and systems let him do that.
When you are a multi-billionaire, you can indulge.
The Sparc T Series chips have more threads than any other processor out there, to be sure, but they have not been very good at raw, single-threaded, integer workloads. The situation got better with the Sparc T4, and it has apparently got quite a bit better with the Sparc T5 – and, presumably, its Sparc M5 "mainframe-class" big brother.

Oracle says it has passed IBM on integer throughput performance

"These machines deliver better integer performance than the IBM Power series," proclaimed Ellison. "The T5 microprocessor itself delivers better integer performance than IBM's PowerPC chip. Now that is really extraordinary, because IBM has had that lead for a very, very long time for integer rate performance, but that lead now moves over from IBM Power to Sparc T5.
"A lot of people are surprised by this," continued Ellison. "When Oracle bought Sun, a lot of people thought the Sparc microprocessor was a real laggard. There were a lot of people who believed that we would never catch up. Well, we have done better than catch up. We caught up, and then we passed the competition. We passed x86 and we passed IBM Power."

Larry Ellison's CPU roadmap throwdown to Intel and IBM

But, Ellison said, playing catch-up is easier than trying to go and double performance again as it has done with the Sparc T3 to T4 to T5. So what is Oracle going to do for an encore? Add database, Java, and other accelerators to its chips to make its software run faster, and free up those Sparc processor cores to do other tasks. Just like it added vector math units and encryption/decryption units to chips, Oracle is going to add database and Java accelerators.
Here's one example of a kind of database search acceleration that Oracle will cook into its processors in 2014:

Oracle is going to accelerate database functions directly on Sparcs

In this example, you want to extract a range of matching data from the database. So you load the data into main memory and the processor runs the database algorithms to make the comparisons to find matches. When you find a match, you save it, and when you don't find a match, you ignore it. The processor cores are obviously busy through this whole process. Now, etch that database search function in the chip. You drop the data you are searching for into a buffer and the database search accelerator looks through the entire database without invoking the processor at all.
"We think that this will give us a greater and greater advantage going forward," Ellison said.
It doesn't hurt that Oracle owns some of the most popular systems software in the world, so it has customers who will be eager for these accelerated functions even if they are probably getting nervous about vendor lock-in. But that said, with Oracle claiming up to a factor of 10X improvement in bang-for-the-buck versus IBM's Power Systems machines, it's already trying to claim it has a huge lead in value. The question is, can Oracle start penetrating IBM accounts, particularly those that used to be Sun accounts? That remains to be seen.
But with the kind of numbers that Ellison was throwing around at Tuesday's event, you can bet a lot more CIOs and CFOs are going to listen. Ellison claimed that the new Sparc M5-32, which has 32 sockets using six-core M5 chips and scales up to 32TB of memory in a single rack, offers three times the bang-for-the-buck compared to IBM's top-end Power 795 system, and has an order of magnitude better value-for-dollar compared to the high-end Sparc Enterprise M9000 machines from Fujitsu, which are indeed several years long in the tooth.
Of course, the real question is how the Sparc T5 machines stack up against x86 systems in terms of price/performance – and Ellison didn't have a single thing to say about that. Larry & Co. are obsessed with taking share back from Big Blue in the midrange and high-end of the Unix market. ®

For software giant Oracle, the cloud of doom

The search for an enterprise solution continues. Aivars Lode

Oracle CEO Larry Ellison delivers a keynote address during the 2012 Oracle Open World conference on September 30, 2012 in San Francisco, California.

by Gigi Douban

Global software giant Oracle will release its earnings report today. In the past, the company has made its money selling traditional hardware and software to businesses and governments, including servers and database software. But experts say that while smaller startups are already staking their claim in the cloud space, Oracle is slow to follow.

Scott Pierce this year launched a cloud startup called 45000 Feet. He says one of the benefits in moving to the cloud is that a company pays only for what it needs.

"If you do only get 10 users, you're only going to pay for enough infrastructure to support those 10 users," Pierce says. "If you get 100,000, you're going to pay more but you're gonna have the ability to service those 100,000 users the way they expect to be serviced."

One mission of Pierce's company is to convince businesses that they don't need a big room full of servers, a concept that goes directly against Oracle's business model.

Krishnan Subramanian, a technology analyst, says when it comes to shifting to the cloud, Oracle is way behind its competitors, even old-timers like IBM.

"They are doing a catchup game right now and they have a long way to go before they can become a leader," Subramanian says.

Oracle declined to comment for this story.

Subramanian says for now, at least, the tech giant is bundling its cloud services so that customers also have to buy database software, which for Oracle, is still where the money is.

On Amazon, cloud service companies put themselves at risk

The cloud has dangers. Aivars Lode

Analysis The power that Amazon Web Services wields over its cloud partners illustrates the new business reality brought about by pay-as-you-go rentable IT – and it's not a pretty picture.

Last week we reported on allegations made by Amazon partners that the cloud king was using its third-party ecosystem as a proving ground for products it could knock-off.

The article spurred a debate, with some reactions to it pointing out that businesses have always expanded into new fields, and that the companies we spoke with were always destined to have Amazon create low-cost derivatives of their products.

"I think for all of us entrepreneurs, it's incumbent of us to protect our assets and innovate and execute," Tom Lounibos, chief executive of Soasta, a cloud-based mobile and web application testing company and a long-time Amazon partner, told The Reg. "You live with this constant fear of someone catching up ... and that's what propels innovation."

Lounibos is unruffled because although Soasta uses AWS, his is a difficult business to clone since it depends on the use of multiple clouds. It is one of thousands of businesses that have built successful operations by using AWS's rentable resources.

But some businesses are not so lucky. They have built tools and devices for the AWS cloud itself, and are ripe for what the tech darlings term "disruption" by Amazon.

With Amazon pulling in revenues of well over $1bn a year, it's no wonder that companies are flocking to a business opportunity – however short term it might be.

These businesses will be aware that their model makes them ripe for cloning – all they do is automate some bit of business process that Amazon hasn't yet presented as a product, or they hold the hands of companies confused by Amazon's management interface or setup processes.

By participating in the AWS ecosystem, these companies help to suck more people into Amazon by making it a more attractive platform on which to develop. But they also give Bezos & Co. a form of costless research and development, which lets Amazon benefit from the hard work done by third-party companies working on its cloud.

Once a company has established a viable AWS business, then El Reg imagines some process kicks in within Amazon's HQ that sees an assessment of the technology get drawn up, then handed to some engineers to create a mock-up.

Soon, Amazon produces a low-cost derivative of the product, as has happened with the launch of products such as OpsWorks, Trusted Advisor, or the many language additions to its free PaaS Elastic Beanstalk.

This clonerama costs Amazon little, as the market research and prototyping have been done by third-party companies.

In the old days of building for operating systems – or before that, mainframes – it would take the mother-company significant amounts of time to develop its own version of a third-party's technology.

Major organizations such as IBM, Microsoft, or Oracle would have to either buy a company to effectively mimic its IP, or enter into cross-licensing agreements, as Microsoft did with DriveSpace.

What sets Amazon (and other cloud operators) apart is that owning a cloud platform gives a business a vast amount of live data about how software is being used and what makes the most money – information that was much slower to bubble up to the platform operator in the days of the mainframe or PC.

This means that cloud operators can become aware of successful third-party products very quickly, and can easily clone them because tech that services AWS must work well with Amazon's technology stack.

Cloud operators can always produce these types of business-process or ease-of-use services at a lower cost than third-parties, as they own the infrastructure on which the services are delivered.

Amazon's strategy of creating a huge ecosystem of third-party services has benefited the company by outsourcing its research and development costs, while bringing in more cloud punters.

The question El Reg thinks businesses should ask themselves is whether they want to help cement Amazon's cloud dominance.

"If you're not paying, you're the product being sold," is what people say to pooh-pooh the concerns raised by users of social networks when they complain about being served ever more ads.

Similarly, in the cloud if you're not paying for the underlying infrastructure, then you're nothing more than a trial business for the data center operator.

Every time a company yokes itself to a larger one, it puts itself at risk of being cloned – but in the past this would happen through acquisitions or cross-licensing, or the slow process of R&D on the part of the operator.

In Bezos's big yellow cloud, the chances of a company being bought or partnered with are slight when compared to the likelihood of it suddenly finding itself competing with a derivative product. (Even if the company is a partner and has paid Amazon for advance briefings, it will typically get no advance information on the competitive product.)

All users of AWS exist in an ecosystem defined by Bezos & Co, so companies should think carefully before linking themselves closely with Amazon itself. If they do, they may find themselves in a business race that they cannot win. ®