HP attacking the mainframe? Like a car vs. a truck

Well, HP is at it again. They are making more generalities about IBM’s venerable mainframe to scare customers off that platform. Check their facts and sources, though and you’ll find that something’s rotten in

Palo   Alto

. Their comparisons are just not realistic. In this note, I’ll be giving you some consolidation efforts that IBM has seen with its customers.


Before we get into that, though, let’s do a quick comparison benchmark to establish a baseline. Let’s compare a four passenger Mini Cooper car Minicooper to a two passenger Freight liner truck cabTruck . Benchmark 1: which is cheaper to commute to work in? Pretty obvious, but I’ll vote for the car. Especially given gas price vs. diesel now…the car is the “green” solution. Benchmark 2: We want to move the contents of our house. Most people would say the truck, but they’d be wrong. We need to accessorize and add a trailer to each vehicle. Now the Mini happens to put the tailpipe right in the middle of the car on many of their models. Why? You’d have to be a moron to put a trailer on their car. As for the truck, with a large enclosed trailer, you can put all kinds of materials in it. In fact, you might even put a couple of the Mini’s inside. So we’ve just proven that with the right benchmark, either solution is appropriate. But benchmarks aren’t reality either. Most people will move their family in the car and outsource to a shipping company to move the contents of their house. So continuing that analogy, there is no one computer that will solve all of a business’ problems, neither a mainframe nor a PC server will do the job by themselves. It’s all about collaboration and using the best servers for the right jobs.


So let’s get back to HP’s claims. I’m a little confused by Robert Frances Group claims right now. In the HP quoted report, they say you get less electricity and floor space with a PC server than you do with a mainframe. I’ve never seen a mainframe that only ran a single workload. Most of them will have transaction processing, batch, interactive, query and decision support running all at the same time. It’s true that you can take one workload off of a mainframe and run it on a PC server and then compare that PC server to a mainframe. The data might actually be real, but as information, it is “incredible”. A single PC server may be smaller than a mainframe and use less electricity (The car). But no single PC server is going to be comparable to a mainframe running multiple workloads. In fact, RFG published a paper in which they said a mainframe will use 3% of the electricity of a comparable PC server cluster attempting to accomplish the same workload. It will also use a fraction of the floor space. (The Truck). But don’t believe me….here’s exactly what they said:


RFG believes mainframe computing platforms have many of the characteristics that will ameliorate, if not eliminate, the current challenges data center managers face with power and cooling. First, mainframe power consumption and heat characteristics are, for many companies, the most efficient servers in the data center. This is true in an absolute sense, where the energy per square foot is lower than any data center system measured by our clients. More significantly, this is massively true in a relative sense, when comparing power used per transaction. On a total workload throughput basis, mainframe system power consumption is almost negligible when compared with distributed systems on a power per transaction basis. As power and cooling costs continue to rise, IT executives should reevaluate mainframe computers total cost and overall value in reducing data center operations costs.

Quote used with permission of Robert Francis Group.


So who are you going to believe? RFG or RFG? Well, in the HP cited paper, RFG just republished the results of a report done by HP. So don’t throw RFG under the bus. Just understand that it’s HP’s low quality and misleading information at work, once again.


As for the Alinean update, it’s a single workload in each example. And in them, they talk about the SAP application server. But what about the database server? Typically, if the application server is on z, the database server is in DB2 for z/OS. Did that move too? The labor costs for System z appear to be much higher than the norm for a business. The report discusses the price of an older mainframe and again, some incredible Software license charges. But what if SAP was added to a newer mainframe? How would that have compared in this report? What if it was added to an existing, newer mainframe, what would the incremental charges be as compared to net new computing servers?


HP mentions the BART system avoiding 50% of their paycheck errors. Wow…that sounds like a big number. They went to Peoplesoft, from what I guess was a homegrown application that was running on a mainframe…at least that’s what HP wants you to believe. So it sounds like the BART people are better running trains than they are at writing programs? I doubt it. That wouldn’t be fair to the hard working people at BART. But remember, if there are two paycheck errors a month and it goes down to one paycheck error a month, that’s a 50% reduction as well. (The Car). So sometimes the big numbers quoted are really just a meaningless indicator to scare you into thinking something else. How many errors a month was BART really seeing? I don’t know and neither do you based on HP's comments.


So let’s talk about something I do know about….consolidations of servers are occurring and System z has been a great place to do that. Nationwide and DGTI are two examples.

IBM has published a paper on SAP consolidation capabilities on System z. The HP press release described a customer that had mainframes and Windows servers. By eliminating the mainframe, they had a common skill set based on Windows. But how real is a customer with a single computing infrastructure? Maybe for relatively small customers, but not with larger ones. RENFE is the Spanish national rail agency. Prior to its reorganization into the two new operating companies, RENFE was composed of 18 separate business units, each with its own intranet system running various line of business applications. These included human resources systems, helpdesk applications and various internal communication portals. To drive better integration across the business and improve process efficiency, RENFE made a strategic decision to create a single information portal for all employees and that was based on System z.


IBM is eating its own cooking by consolidating many of it’s thousands of application and database servers onto System z. But that’s not the whole story either. They are also consolidating some onto System p and some onto System x. In each case, IBM is looking at underutilized stand alone servers, the baseline for the PC server marketplace and leveraging virtualization technologies to get a large reduction in physical server images. IBM is putting the right workload in the right place that makes sense for the business environment. (The Trucks).


We see constant examples of taking 100’s of underutilized standalone PC servers and consolidating through virtualization down to 10’s of higher utilized PC or RISC servers or individual mainframe servers. In each case, the customers are saving substantially on labor, environmental and capital costs. HP will tell you that 100’s to 10’s is good enough.

IBM mainframes, though, can get that down to single digits in many cases.


Look at HP’s Brazilian Navy example. A lot of folks may perceive that a mainframe could never go on a Battleship, Aircraft carrier, early warning aircraft or other military location. Well, those folks would be wrong. Today’s modern mainframe, the System z, going as far back as the zSeries z800 processor meets or exceeds the electrical, floor space, ambient temperature, humidity, air pressure and vibration specifications necessary to satisfy the locations in which those servers may be deployed. See page 12 to view a subset of these specifications. In addition, it provides operational redundancy built into the hardware architecture and operating systems that exceeds the availability requirements necessary to satisfy those particular business needs. And with its open programming models, including Java, J2EE, C/C++, in addition to the venerable COBOL and PL/I capabilities, it provides a hosting environment to capture those programming needs.


In fact, development belongs on the desktop. The most creativity and tooling is possible in that desktop and you can reboot the system at will to test your applications. IBM’s Rational Developer for System z (RDz) and Rational Team Concert suites provide an Integrated Development Environment that can leverage the simplicity of the open programming environment through its Eclipse.org tool base, but easily apply those skills and knowledge to mainframe application deployment. You want mainframe development skills? You have them in your hands already. Get the tools and put those people to work.


One of the principals of the mainframe has always been that the operating system, middleware and hardware are responsible for data locking, security, system resilience, storage management and capacity management. This enables multiple workloads to operate as individual processes and maintain the integrity of the system and the data. On other platforms, it’s typically the application that is responsible for many of these characteristics. In order to achieve these qualities of service, additional products must be acquired and additional code may have to be written by application developers to deliver these qualities. The point of this all is that a business might actually reduce the amount of code necessary to achieve their business objectives if it was targeted for deployment on System z and reduce their operational risk at the same time. To summarize this point, it can be the same code from distributed systems in a mainframe operational container and deliver superior operational performance. Same code, different container with superior operations model.


So this started by pointing out inaccuracies in the HP press release. How can a business use that information? Well, maybe to buy an individual compute server, that information may be helpful (The Car). But looking at an enterprise that needs to satisfy multiple business needs, it doesn’t appear too helpful at all (The Truck). They use Apples to compare to


. Customers continue to grow their compute power on IBM mainframes. New problems are being solved in creative ways, leveraging the best of the mainframe in collaboration with other systems. Like RENFE, get on board the IBM mainframe.



by JimPorell November 12, 2008 in Economics, Innovation, People, Systems Technology


TrackBack URL for this entry:

Listed below are links to weblogs that reference HP attacking the mainframe? Like a car vs. a truck:


Although anyone reading this will say I am biased since I am am the IBM sales executive for our major markets, it is a fact that System z gained 17 points of market share since 2000. Although IBM sellers are very talented our clients are very sophisticated and they would not have purchased these mainframes unless they deliver the differentiated value Jim claims on his blog (above). It is interesting that this share gain has come mainly from HP (and Sun). These facts are objective and undeniable; not biased opinions.

Posted by: Bob Hoey | Nov 13, 2008 1:25:07 PM

I wonder if some of HP's issues with IBM are related to the difficulty they had in getting down to a reasonable price in trying to outsource DHL's data center which has a large number of z MIPs. The rumor is that HP, as a direct IBM competitor, could not get a good deal on hardware & software from IBM. This does not bode well for EDS customers now that EDS is part of HP.

Posted by: | Nov 14, 2008 8:44:04 PM

John Hancock would be proud, Jim. :-)

Posted by: Timothy | Nov 15, 2008 7:45:45 PM

Keep in mind that bad-mouthing one for bad-mouthing another doesn't make you any more credible. It's like two young children bickering back and forth (a perfect example is your reference to x86/64 servers as PC servers in an attempt to belittle their capabilities - very juvenile). What IT shops want are facts. Vendors tend to be very fast and loose with facts - HP, IBM, et al. Take the famous Nationwide case study that IBM continually touts. Did you know that Nationwide does far greater consolidation on VMware than on Z? That they have chosen VMware for deployment of critical workloads over Z? That's not to say that they haven't derived some value from the Z platform. I would simply urge any IT organization to do your own research, your own comparisons. It's not a matter of who can you trust with respect to HP or IBM. You can't trust either one (especially anyone in Sales). Trust yourselves. Do your due diligence.

Posted by: | Nov 24, 2008 1:13:58 PM

Nice article, but i will never give credit to a company when the product is good. same time when the product is bad it is not the companies full fault. In IT various factors are affecting the product. it is just my thought.

Posted by: Car PC Guru | Sep 9, 2009 11:30:16 AM

The comments to this entry are closed.

The postings on this site are our own and don’t necessarily represent the positions, strategies or opinions of our employers.
© Copyright 2005 the respective authors of the Mainframe Weblog.