Main frames got so boring, who cares? They are really just a rack of processor cards that flash a red light when they break and you hot swap them. There are a couple hundred "paper mills" around the country that need I/O operators Banks are desparately trying to separate themselves from paper as is the government. Software is written in boiler rooms in india.
This is going to be a very small labor market but I imagine some folks will make lots of money, most would make more yanking Romex.
#132285 - 09/02/0507:43 PMRe: Who'll mind the mainframes?
That was a quite illuminating article, and this 37-year-old who has worked on mainframes a good deal cares a bit
A lot of it depends on what you are calling a mainframe. I reckon the article was specifically referring to a classic IBM site. But I've worked on Tandem and Stratus systems too, which deserve the moniker of Mainframe, and I learned to program on ICL 1904s, VAX 11/780s and PDP-10s. Shall we agree that where word sizes of 31, 36, 24 or some other eccentric number are the norm, and where character sets can be based upon 6-bit bytes, or radix-50, or EBCDIC, that's all mainframeland?
As to boredom. Depends on your boredom factor. I love writing code that actually solves a problem -- you know -- often it is back-office stuff. And as a general rule I'd say a mainframe is an excellent environment in which to do it. By a problem, I am talking about things like international equity/bonds clearance, settlement and custody systems that turn over 62 trillion dollars per annum, the European airspace control system, the administration of life insurence policies for some three million customers, or a well-known company who offers almost the whole globe the facility to use a blue plastic charge card to pay their way nearly anywhere. Nice solid mainframe solutions. If your definition of "more interesting" i.e. not "so boring" is to spend all day long patching away at rack after rack of blades running some varient of Windows, just to assure that there won't be some sort of security issue, or that the so-called TP monitor won't crumble to nowt after the first 3000 users, or the database product can actually do locking without calling on virtual memory even when there is 2GB free in physical memory, then I reckon your definition of "Interesting" owes a fair bit to Confucius. Sorry to rant
>They are really just a rack of processor cards that flash a red light when they break and you hot swap them.
Mmmm. I don't know about this. If so, channel-attachment of high-bandwidth devices, front-end controllers, Geoplex, FSIOP in the mid-range systems, proper partitioning and virtualisation, blah de blah de waffle just talk to your friendly local IBMmer -- well; if it's just a red LED, then why do we bother? That z-series architechture is a bit of genius, and it is still as valuable as ever for applications that need it. How can a rack of Intel or Sparc servers match the I/O bandwidth of even *two* IBM z-series channels, let alone the *eight* that are offered? And after channel; well, how many gigabit Ethernet controllers is it sensible to plus into a single PCI-style bus? It doesn't scale all that well. On the other hand, I've worked on mainframe applications that support 20000 users, all day every day. The total CPU load is nugatory -- it's I/O th at counts. The other thing is Business continuity. Mainframe offers Geoplex, where if an atom bomb falls on one data centre, the transaction rolls its way to the other on the other side of the earth, and processing continues. Came from the cold-war, and it is not something that PC type servers offer.
> There are a couple hundred "paper mills" around the country that need I/O operators Banks are desparately trying to separate themselves from paper as is the government.
I don't quite follow this.
>Software is written in boiler rooms in india.
Mainframe stuff seems to me to be less so, and I refer you to job sites like Jobserve for evidence. Plain old Windows apps often are outsourced, but because a mainframe is far more a high-value, bespoke investment, a lot of the development of software is in-house, even now. My mainframe skills are in assembly language and PL/I and there is *never* a shortage of work either here in Europe, nor in America and Canada either. (And I'd love to work in Canada one day: it seems to have a magnetism for me). A good PL/I contract attracts rates up to EUR 650 per day, and I don't think that's bad pay by anybody's standards, yanking Romex or not And apart from that, PL/I is a language from the cosmos. No reserved words at all Think of the badness )
I've done better this last ten years precisely by *avoiding* the Windows platform and specialising on IBM, the "high-availability" arena, and unix, than I would have done by looking at the rack-mount data-farm scenario.
All usual disclaimers apply; I just expect you got annoyed with doing
and then the remaining 10 cards that are needed just to rename a file using IDCAMS processing, and after typing your SUB in the ISPF you press the F9 key in TSO to look at the SDSF thing to see if it's run or not
For any MVSish (or VM/CMS) geeks, there's a rather beautiful piece of software that creates a virtual z-series (IBM mainframe platform) upon any Windows, Linux or Mac OS X system, and its name is Hercules. It is more than strong enough to run CICS, DB2 and a nice debug/development environment, and I recommend it very highly.
Ho hum bedtime. I wish it were so easy for those poor people in New Orleans
#132286 - 09/02/0507:53 PMRe: Who'll mind the mainframes?
Now you've set me off, I've IPLed MVS in Hercules on the ENIAC computer ( the Sun box.) IPL for non-mainframers is Inital Program Load, and means booting up. IBM calls a hard disk a DASD (pronounced Dass-dee), for Direct Access Storage Device.
Be back presently -- there is a special reason why
#132287 - 09/02/0508:08 PMRe: Who'll mind the mainframes?
Here is for your delight the very worst bit of PL/I code that has ever been written.
proc: proc; entry: proc (entry) returns (bin (71)); if returns < entry then return (call (entry - returns) * entry); return (1): return (1); dcl entry bin (71), end entry; end entry; dcl dcl char (char) based (based), based ptr ptr, ptr bit (144) init ("33184c674379c8d26136080bc32308f4bc64"b4), if defined defined (1), 1 then, 2 else (0:1) options (variable) variable init ((2) ioa_), ioa_ options (variable), defined (0:1) bin init (1,0), end char (16) based (addr (ptr)), char fixed (21), cu_$arg_ptr entry (bin, ptr, prec (21)) returns (prec (35)), returns defined defined (0), call generic (cu_$arg_ptr when (fixed, ptr, binary (21)), entry when (prec (17:71))); if if ^= call (returns, based, char) then do; goto: call else (returns) ("I need a numeric arg > 0 & < 23"); return; end; if if ^= verify (dcl, "1928374650") then goto goto; if bin (dcl) - 22 > if | if = bin (dcl) then goto goto; call then.else (if) (end, dcl, call (bin (dcl) ---- if)); end;
PL/I these days is more or less exclusively an IBM mainframe language, which to me seems a pity, because it is easily a more civil language than C, and it is easily a more comprehesible one than Java. It was only IBM's silly license terms which stunted it. C'est la vie.
#132288 - 09/02/0511:40 PMRe: Who'll mind the mainframes?
I was your friendly IBMer. I started fixing 1401s in 1966, I went to hardware/software school for 360s in 68 and was plucked out for field program support on DOS, the real one. They killed that career path when they pulled all the assembler level support back to a Hudson valley cube. Not for me. I went back to hardware and ended up as region support for 135, 138, 145, 148 and 158. That was fine. We had the documentation and they broke a lot. When I was in Endicott support for the 4341 I started understanding mainframes were never going to be fun again. They seldom broke and the field didn't have any documentation when they did. I kept myself busy with check sorters and big laser printers. That trend got worse in the 80s and I moved to Installation Planning, contract services, building computer rooms, connectivity (LV) and TP support. Computers may not break but they were big honkers that needed a special big room with lots of interesting stuff in it. It was fun for a while. When I saw the first announcements for the RAMAC (a rack full of PC hard drives) and the 9672 (the first blade server type machine, still running MVS) I knew the whole business was going away. These days, that acre of raised floor is a small room and nobody really needs to go in it. Mainframes are basically just big file servers. The raw number of mainframe shops is probably 5% of what it was in the 90s.
A clue, in 1987 there were 47 IBM folks in Ft Myers Fla. Now there are FOUR and 2 of them are exclusively printer or desktop guys. None of them do much with anything we would call a mainframe. There are NO IBM software guys here. That support comes from a guy with a funny accent named "Bob" from halfway around the planet.
I made myself very happy being an electrical inspector.
#132289 - 09/03/0506:54 AMRe: Who'll mind the mainframes?
Aah - so your criteria for how boring they are is how frequently you need to get in there and start fixing them -- Now I understand
I still don't quite agree that they are only file servers. The application development infrastructure on the mainframe is much much richer than that -- there's CICS, and MQ Series, and DB2, and all the rest of that excellent middleware. Basic fileserver platforms would not offer any of that. And even considering only the file system part of it, mainframes are streets ahead, with support for proper record types -- fixed, varying, indexed etc, decent file generation control, automatic disposition/expiry etc and a heirarchical storage management system right there in the operating system, quite unlike the simplistic "everything is a stream of bytes" approaches of the Unix and Microsoft families of systems. Mainframes are a true joy to write applications upon because all of this infrastructure is already there, and it is tested, proven, and works dependably for years on end.
What I noticed in the clearing house machine room was that as the footprint of the mainframe steadily shrank over the years, the space was being filled up at an equal or almost faster rate by racks containing Compaq PC servers running NT and Win 2000. They were much much more wasteful of space than the IBM because there was not (and still isn't) any effective way of partitioning Windows systems, so, for safety, every application needed its own server to prevent one crash or one period of scheduled downtime from impacting all the other applications running on that box. We were running something like 150 critical applications, and then there were the multiple environments -- development, volume test, production, contingency. The heat and noise those racks and racks of compaqs produced was shocking.
This was a misguided management choice, when after "high-level talks" certain decision-making managers (i.e. budget setters, not technical guys) being flown across for the hard sell at Microsoft Seattle, the marketing did the trick and NT / Win 200x was for a time supposed to be the platform of choice for implementing new applications. In fact Microsoft was using the site as a "showcase" for the new platform.
Unfortunately, none of the NT applications they actually put in (1) supported the required throughput or user load, (2) had good enough uptime, (3) came in on budget, (4) were implemented on time, and (5) as mentioned above, the servers themselves where using too much space, slowly cooking the aircon, and costing a fortune. Unfortunately the marketing case studies that sold them on the platform in the first place failed to account for these extras. The Microsoft showcase thing was quietly dropped. The applications are slowly being phased out and migrated, and they've gone back to the mainframe as the platform of choice
I know a few mainframe IBMmers here in Belgium that work for Global services, and I know a couple more who work at Hursley Park in the UK on CICS, so I still reckon there *is* still life in the old beast yet.
The article mentioned that the MF population seems to be ageing, and I suppose it's true: I know probably about 10 in their late 30s early 40s (like me) and the rest are all about 5-10 years away from retiring. At least zOS now has very good Java support, which makes it a bit more attractive as a platform for the younguns coming from colleges these days. What I do know is that to keep a good job in IT, offering skills in a good mix of platforms is best, so I've kept myself up-to-date on mainframe and I've also got Unix and such stuff on my CV as well. Hopefully that will keep providing enough work for me until I make it down to Nice for my retirement
[edited to fix a typo]
[This message has been edited by jooles (edited 09-03-2005).]
#132290 - 09/03/0512:49 PMRe: Who'll mind the mainframes?