views:

211

answers:

5

My bank account, a lot of client information, accounting, insurance and whole lot of other mission critical data is still running on mainframes and/or AS400s (iSeries?). COBOL, RPG, CICS and a number of other older technologies still dominate the mission critical core data for most Fortune 500 companies.

The really big question is how much of the industry does this represent, and is this changing at all?

Twenty years ago, fresh out of school, we were going to rewrite all of the legacy systems with the next wave of technology. Twenty years later, it doesn't seem like much has changed.

Edit: Does anyone know of any statistics out there? Or well known numbers, or examples?

Paul.

+1  A: 

Mainframes are still alive and kicking. One of the reasons is that it is a stable platform with lost of extension possible.

The biggest problem with mainframes is to find skilled people to operate them. Most sysprogs I know are within years from retirement and they are no exception. So is the gap is not filled the dinosaurs have met their meteor.

Gamecat
I read an article about Cobol conference, 50% of the guys there were young, not old greybeards. Apparently you can make very, very good money.
gbjbaanb
Mainframes are fast, with unmatched uptime and reliability. They're the right choice in many business situations. They're also clumsy and annoying to use, and most people can't hack on a mainframe at home.
David Thornley
If you can't change it, you can't break it :-)
Paul W Homer
Cobol programmers is not the problem. It is the sysprogs.
Gamecat
+1  A: 

Not sure of the % but a lot of companies are leaving the core legacy code (millions of lines) and nibbling away at the edges.

e.g. for large COBOL installations, you can develop new interfaces by means of web services which can then be used by a ASP project.

So the core business engine remains but users can now use a browser instead of the old 80 x 24 "green screen".

nzpcmad
Yup that is also true.
Gamecat
A: 

I don;t know about large mainframe type legacy systems, I think that depends on the industry segment. Banks for example are much more likely to use them, as is large retail type places, after all, once you've spent millions of pounds on a mainframe, you don't dump it for a new webapp (that often doesn't do what the mainframe did anyway).

I have a codebase for modern command and control systems that has some new stuff in there, but the core of it dates to 1984. That core is the most stable part of the entire system! (perhaps that's obvious when you think about it).

As a result, legacy systems are the majority of systems in place today. Never break things that work. you may slap a different GUI on top (and Windows-based systems tend to be centred around this) but the real work is more likely than not to be legacy. Even web based sites can be described as legacy if you look at some of the old Java code that still powers them.

edit: quick note, according to Ovum in 2005 Cobol accounted for 90% of all financial transactions.

Plus IBM has made a commitment to train 20,000 graduates in mainframe tech and is committed to the platform (system z) including.. wait for it... PHP 5.1.25 on z/OS!! (and Perl, so its not so bad)

Legacy! Hah! Its more modern than all your "pretty pretty" managed languages on your toy platforms :-) (oh alright, you can get Cobol.net)

gbjbaanb
I know, legacy isn't a fair word. I would have used 'stable', but few people would have understood the question :-)
Paul W Homer
A: 

This isn't really answerable.. There's basically no way to quantify how many "systems" are around, never mind how many of those would be considered "legacy". The term legacy system alone is a bit vague

Basically, lots, and of course this is rising.. As time passes, existing shiney new code could be considered legacy, and as the "if it's not [too] broken..." saying implies, there's not much reason to rewrite.. Although, as gbjbaanb said, in $year_cobol_was_popular + ($avg_age_of_retirement - $avg_new_programmers_age) a lot of those system will (possibly) start to get replaced as the only people who know how to manage them retire.

dbr
I'm more interested in the mainframe and mini markets than in something like the C/Tuxedo, client/server systems from the mid 90s. That first set was scheduled to be removed, but it never happened. That second set seems to be part of a five year cycle that turns over regularly.
Paul W Homer
I wouldn't bet on those systems being replaced as people retire. As long as they can keep limping along, and the staff can kinda sorta do it, it'll be cheaper to keep them going than rewrite, test, migrate.
mpez0
+1  A: 

Code is legacy the minute you deploy it to a production system

A server is legacy the minute you drive it off the lot rack it up.

(if it's not obvious, tongue planted firmly in cheek)

Alan Storm
I tend to consider systems to be legacy when they're no longer in 'active development'. The term implies at least some desire to remove the system and replace it with something in a newer technology. Otherwise it just becomes another generic term for software (and we already have one of those).
Paul W Homer
"The term implies at least some desire to remove the system and replace it with something in a newer technology." Have you worked with these damn kids nowadays? ;)
Alan Storm