Mainframe yearbook

Arcati have published the "Arcati Mainframe Yearbook 2006" - an interesting read with reviews of mainframe growth / how much data stored there vs how much spent on it etc etc.

by pwarmstrong May 17, 2006
Permalink | Comments (1) | TrackBack (0)

Inexpensive Mainframes

So IBM announces that the new price of the System z9 Business Class starts "about $100,000." Presumably that's for a model A01. Or is it with one Linux engine? Anyway, anybody know what effect that announcement is having on the price of z800 and z890 systems?

Can I get a real 64-bit mainframe with 8 GB for the price of an average car now? Who wants to go in with me? I bet we could start running this blog on a z800 from my basement data center....

by Timothy Sipples May 14, 2006
Permalink | Comments (9) | TrackBack (0)

Mainframe code presents problems?

A discussion thread from the IMS-Listserver, that has been going over the last few days:

Did anyone get a chance to read the following article?

This seems to imply that big banks will be migrating off the mainframe (IMS) real soon.

Also, it mentions SABRE which is American Airlines TPF (Transactions Processing Facility) also known as ACP (Airline Control Program).  Most TPF applications are written in assembler.  They are very small and very fast.  TPF programs track everything from reservations on millions of airline flights, advanced seat selection and on-time performance of aircraft.  I wonder how well a network of distributed servers can handle what is essentially a huge inventory application with very perishable product?

Any comments?


WOW. As an alumnus of the long defunct Eastern Airlines; this brings back memories that are older than I admit to being.  Eastern collaborated with IBM on the original ACP and sold a copy to American for what was then the 'world record' price for a piece of software. It may still be the 'record'; depending on how you define 'single function'


Notably, AMR only speaks of the high cost of conversion.  It doesn't sound to me like they are seriously considering replacement.  If anyone from SABRE monitors this forum; it would be fascinating to hear how many transactions per second they are processing these days.  That would spawn a discussion in this forum of whether or not IMS could yet handle that volume.

IMS transaction integrity would be a HUGE leap forward.  When an ACP transaction died, the application program was responsible for backout/recovery/cleanup/everything.  Ponder that the next time an airline can't find a record of your reservation.

I pursued converting Eastern's reservation system to IMS Fast Path in the mid 1980s.  Our ACP group viewed that as a call for jihad and Eastern was spiraling into bankruptcy so rapidly that the only positive outcome was giving me something to discuss with Peggy Rader.

ACP/TPF was a brilliant piece of code for it's time; but, that was a long time ago.  It never approached IMS Reliability.


P.S. 'Financial' institutions will likely migrate away from IMS whenever someone produces a database & transaction processor that is more reliable than IMS (sometime after pigs learn to fly).


Great discussion!  As an Alumnus of the defunct Piedmont Airlines and USAIR, I could not agree more with Dale.  Brilliantly put.  However, the sad news here is that the Mainframe is NOT a strategic direction for my corporation no matter what you say or what proof you have to show otherwise.  No way to get anyone to look at the tremendous cost of running everything on a gazillion servers or how many people it takes to support those servers not to mention the non-24x7 availability.

The most puzzling thing to me is the three-day reorg.  When was the last time one of those ran on the mainframe, IMS especially?  Probably only one prior to converting it to make-sense products like MAXM Reorg or even Online Reorg if using HALDB.

I feel IBM let us all down.  In technical conferences of the past, I have brought that topic up.  IBM Marketing expected the individual corporate techies to drive technology and save the mainframe.  That went away in the 80's and early 90's.  Who listens to us now?

Now the CIO's get the common trade journals.  Where was the mainframe being marketed there?  Where was IMS, MVS, CICS or DB2 being marketed?

No where.  That, my friends, is the problem. Newer technologies came and crept up on us and our naivety felt the dependable nature of the Mainframe would win out.  Wake up.

We all know we have the most stable systems around.  You can't compare the response times or reliability with Open Systems.  But let's make way for all those servers which are now beginning to take up areas in the data center that resemble the old DASD and CPU farms.  My goodness, has anyone seen the new servers?  They look just like old 360's.  Try to convince upper management the Mainframe is the best investment and they now look at you like you just swallowed poison.  How very tragic.  I worked on both TPF and IMS at


and USAir.  I hate to sound like a skeptic, but the writing is on the walls, at least here.


OK, the first thing I'm getting from this discussion is that at one time or another, just about every IMS DBA or systems programmer on earth must have worked for a major airline that has merged, gone belly-up, or come close.  Count me in that group as an ex-US Airways alumnus.

<<begin rant>>

Like others, I find this discussion fascinating.  I'd been working as an IMS developer for less that a year in the early 80s when "friends" and trade publications began advising me that IMS was a dinosaur and would soon be eclipsed and replaced by DB2.  The fact that we're all still discussing it tells you how accurate that prediction turned out to be.

During that same period, I attended night school at the




and was advised by the head of the CompSci department that mainframes and COBOL applications would soon disappear and be replaced by DEC and VAX clusters and languages like Pascal and


.  Likewise, I've watched and watched as client server applications and DBMS became hotter and hotter, causing the same publications to write the mainframes off long ago.  How much of any of this has turned out to be true?  I'll leave that to all of you.  Has the short sighted mentality that makes executives want to convert everything to the hot new servers and DBMS systems eroded IMS' domination?  Sure, the same way that DB2 did at one time.  Yet, IMS continues to evolve and becomes even stronger and more dependable with each release.

At one of my recent employers (and there have been MANY), during an orientation session the CIO discussed the technology the company was using and would use in the future.  There company hires a significant number of IROCs (Idiots Right Out of College), and most of them know only that CompSci educators have told them mainframe technology is outdated, and one questioned the CIO's statement that the mainframe was absolutely integral to the company's plans.  The CIO put it into perspective by suggesting that rather than focusing on the terms "mainframe" and "servers" and seeing the two as being different species, that people unfamiliar with mainframes instead view them as the biggest, fastest, most powerful, most dependable, most cost effective servers imaginable, and understand that the mainframe is simply the best choice for many businesses processing high volumes of transactions and requiring speed and dependability above all else.  Isn't that what IMS, z/OS and mainframes are all about?  At most places I've been where both IMS and servers were used, the server application required far more DBAs and far more outages than IMS.  When company executives make decisions to arbitrarily convert to server applications because that's what they see being hawked in magazines and trade journals, they're making technology decisions the same way they'd shop for a car, and that's a disservice to the companies for whom they're supposed to be guardians. 

<<end of rant>>

I agree with Ivan, who said earlier that the demise of mainframes and IMS has been greatly exaggerated.



I really like the discussion this brings out, with Avram comments, I started out my career in 1969 with the state of


on a 360 model 50 with 512K. we ran applications written in assembler to all the processing for the states unemployment compensation systems. You know what, these systems are still running today, I do not know how, but they are, and there was never a requirement to change or rewrite anything due to hardware or software changes made by IBM. They were always downward and upward compatible. My old job has long been outsourced, but IMS is still there, for how long I do not know. I'm sure someone is looking at how they can rewrite all these OLD applications to run on the much cheaper hardware that will take who knows how many servers to support and how large of a staff will it take. How reliable will it be. Zelma mentioned 3 day reorgs, I have seen 2 day upgrades for this other hardware. Has it ever taken 2 days to upgrade an IMS release, I think NOT.

After the end of the Vietnam war, I can remember working 7 days just to get 5 days work done, unemployment went sky high in the early 70's, but we still got the claims processed.

Today these systems even have GUI front ends on them, to give them the modern look. They are kinda of like the old Timex commercials, they take a lickin' and keep on ticking.


Re the question: "Anyone have any idea from IBM side for example, what's the relative revenue and profit between various *series?  If mainframe is still profitable for IBM..."

Earlier someone posted a link to a NY Times article from this month that stated this;

"all mainframe-related hardware, software and services account for a quarter of its [IBM's] revenue and, more important, about half of I.B.M.'s total operating profit"

So one would assume pretty important then!

I know of a large company in the


that chose pSeries as its platform of choice for a large new application, despite having been heavily invested in zSeries already and understood the scalability, performance and resiliency advantages of z and liked the mature management procedures that came with it, however they saw pSeries as the cheaper option. IBM had also recommended pseries to them. Now a little way down the line as the application expands and evolves they are finding that pSeries is actually proving less economical than zSeries! can assume that gap will only widen as the application grows yet further.

So perhaps IBM think they can get more money from their customers by selling them non-mainframe products!?!

I think IBM need to encourage their mainframe customers to leverage their mainframe investment. All too often I think IBM is sending consultants who know nothing about mainframes through the doors of their customers.


My concerns:

- Marketing for mainframe in individual shops, more and more shops the application people have access to business and not system people.

Applications tend to market to business in silo view and don't need to worry about integrated application advantage at all. Business naturally will favour the slick views and the appearant savings.

- Marketing for mainframe as a whole - Missing.  Anyone have any idea from IBM side for example, what's the relative revenue and profit between various *series?  If mainframe is still profitable for IBM, some money should be channel into making a better image. Or does IBM think a wild mazed IT environment is really the best thing for IBM?

- Charge back scheme in various shops are probably more mature in the mainframe, and I wouldn't be surprised that many distributed system cost other than equipment is buried within the mainframe chargeback.

- Software cost on mainframe - what a impressive system, that alone can bury mainframe, can't say anymore!

- When your application enviornment start to migrate little bits to distributed enviroment, this creates an interesting inbalance because they will take all the easily done and cheap parts with them, leaving the difficult parts in the mainframe, thereby making the mainframe with higher percentage of mutants. After a few iterations, the applications left on host all look like dogs breakfast, and guess what, the system people have to deal with them. And then the application people say 'see, z is difficult and unmanageable' - a nice twist from 'the application is a mutant'.

On the other hand, I myself believe that distributed computing for high volume integrated environment is got to be hoax, we will be in high demand when the wave is over.

by pwarmstrong May 12, 2006 in Application Development, Contest, History
Permalink | Comments (2) | TrackBack (0)

Mainframe Transactions Will "Easily Double" Within 43 Months

Last week's IBM zSOA conference, attended by James Governor, attracted lots of press, including an article in the New York Times. There's also an IBM press release with some bullish news: IBM expects that, thanks to SOA, mainframe transactions will "easily double" within the next 43 months.


by Timothy Sipples May 9, 2006
Permalink | Comments (7) | TrackBack (0)

Wachovia mini reference at industry analyst SOA on z conference

Wachovia is going end to end mainframe SOA.

The presentation was interesting- basically Wachovia had an ageing proprietary middleware infrastructure it chose to overhaul and pretty much completely replace... with a mainframe-based system running WebSphere tools. Really this can be seen as a "new" mainframe customer, in that respect.

Keith Harris, VP of retail architecture and integration, Wachovia:

Wachovia is the
4th largest bank in the US
has 3,900 financial centers
and is the 3rd biggest brokerage in the US (another best kept secret?

Core SOA services, such as: Authentication, logging services etc.

Wachovia has some aggressive uptime requirements. the SLA is 99.9% uptime including scheduled outages.

Keith said: "basically, It can't be down."

Results- 7 months since install

Wachovia has exceeded SLAs, and now runs millions of transactions a day

zAAP will provide 92% offload.

WebSphere on z gives: Proximity to data, Achievable SLAs.

Thanks for the info Keith...

by James Governor May 4, 2006 in Application Development
Permalink | Comments (5) | TrackBack (0)

more liveblogging: Robert LeBlanc on zSeries SOA

From the IBM SOA on z industry analyst conference:

I have known Robert LeBlanc for a few years now. He did a good job at Tivoli, turning the organisation once and for all into a true IBM division, rather than a maverick Austin outpost. Now he is running WebSphere as SOA kicks into high gear at IBM.

Robert claims IBM was able to spin out its PC business to Lenovo seamlessly, without interrupting the supply chain, by taking a SOA approach, which is a story I hadn't heard before.

So this mainframe SOA thing. What are the numbers?

He said CICS saw the fastest version to version upgrade by customers in 35 years… to take advantage of new XML and Web Service functionality.

WebSphere Application Server (WAS) on System z grew 21% in 2005. That is pretty impressive growth. IBM also claims 25% of z accounts are now running WAS.

What is the news?

CICS Transaction Server v 3.1 has been enhanced. Who knew this product would hot enough to justify six month revs?

One enhancement is to integrate CICS Service Flow runtime with Websphere Process server. That potentially means "workflow virtualisation" between the two environments.

I had a short conversation yesterday with John Shedletsky, "Mr Mainframe TCO", about language. One big problem for the mainframe is that mainframe people speak a different language from distributed people. Many architects literally don't know what their mainframe people are talking about. John pointed out something very insightful.

CICS is just a container. Say that to a Java architect and they are going to grok it straight away. Sounds simple doesn't it? I would like to see John pursue this line. Maybe we need a mainframe to distributed translation wiki?

But back to Robert.

He spoke to IMS enhancements: IBM last year offered an IMS SOAP interface. The new announcement is an XML adapter on IMS connect, enabling direct connection to IMS without needing WAS an an intermediary. Expose the transactions. Note: I need to get some clarification here. SOAP is one thing-but in many respects native XML is more interesting because it potentially enables more RESTFUL development and data access models. If I am a mainframe shop that wants to offer an online information service, its likely to get more consumers if I offer a more lightweight API, rather than forcing everything through SOAP. SOAP is anathema to many Web 2.0 developers, and IBM definitely needs to position z for Web 2.0. IBM needs to think more about XML data sources not just XML Web Services.

But for those that do like WS-* and metadata-driven development Robert said IBM is going to bring Registry/Repository to the mainframe by year end.

Its also interesting to hear IBM's approach in that regard. While IBM has been saying that its Reg/Rep architecture is federated, what does that mean in practice, and why is it relevant to mainframe?

("The return of the data dictionary" -is this one for the mainframe to distributed babelfish wiki)

The point is that IBM plans to position z as a master registry manager (like master data management), across multiple SOA end points. Maybe it’s a "Registry of Record" but I think 37Signals already owns that acronym ;-)

All in all it seems z and SOA is really happening. Wachovia, a customer, also came in and presented. More on that soon.

by James Governor May 4, 2006 in Application Development
Permalink | Comments (6) | TrackBack (0)

A little live blogging: IBM's SOA z analyst event

IBM has a room full of people here in New York, industry analysts that usually cover the software market, to learn a little bit more about why the mainframe makes a good service oriented architecture (SOA) platform.

Steve Mills, IBM Software Group supremo kicked off, before handing off the Jim Stallings, GM of the mainframe division.

Stallings says a new business context is emerging, that suits the mainframe. Security, risk management, and even power management are all traditional mainframe strengths. He even claimed to talk to Google lately - I wonder if we're going to see System z in the Googleplex anytime soon?

We have spoken on this blog before about China scale, which is a real and fundamental context. China is going to have all the biggest businesses in the world by customer... no competition. Try scaling to 500m customers in five years. You're going to need a pretty heavy engine to handle that.

So what about new workloads. Q1 MIPS grew 25%, with 35% of that growth across IBM offload processors such as zIIP, zAAP and IFL.

How about a customer example? Farmers Insurance runs 45m transactions a day, supporting 50k users. 90% of its revenues are driven by the mainframe running WebSphere. So what about dollar savings? Well, as any mainframer knows customers pay per MIP.

By using the zAAP offload engine Farmers cut its MIPs from 1200 to 700.

That's the kind of price break that should interest any mainframe shop.

Something interesting is definitely afoot in mainframe economics. Quite simply - every year IBM is finding new ways to reduce mainframe cost of ownership. IBM calls this trend The Mainframe Charter. Its also good to see IBM focusing so squarely on third party ISVs. Successful platforms needs developer communities.

Well I guess I will post this now. I hope the Farmers example was a public reference. IBM didn't say not to report any of this material, so here goes...

by James Governor May 4, 2006
Permalink | Comments (2) | TrackBack (0)

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.