• Status: Solved
  • Priority: Medium
  • Security: Public
  • Views: 1092
  • Last Modified:

Future of COBOL and Mainframes

Hi,

I would like to activate a discussion about the subject " Future of COBOL and Mainframes".
What do you think? Will these environment live on? The death of COBOL and Mainframes were promised many times? By trend I think there is no chance for Mainframes and COBOL (for a very long future ) but how long will this dinosaur live on?

Regards

Michael
0
michael_ch
Asked:
michael_ch
1 Solution
 
giltjrCommented:
Well, I got into IT in 1982 and went into mainframe (IBM mainframes running MVS) and my friends laughed at me.  They said the mainframe was on the way out and departmental computing would take over.  Then it was client server, then it was thin client computing.  All of these are based on breaking up the functions on a large central computer and running them on  many smaller distributed computers.

Today, what is the big thing?  Virtualization.  Taking all those low usage distributed servers and moving them to fewer larger physical boxes.  

Most distributed platform are setup to run "a task", not that they are single tasking, but if you are running an application you are running a single application on one or more servers.  Why?  Because typically distributed platforms don't do well in balancing diverse workloads.  So you have one OS per application or function.  Typically you do not run your database instance on the same OS as you are running the application that uses it.

In the mainframe world (at least MVS) it is very good at balancing diverse workloads so you can have your database and application servers on a single box, even in a high transaction enviroment.  Now its not just the OS that allows you to do, this, it is the hardware design.

From a hardware point of view mainframes will not only survive, they will grow.  Virtulization will push this.  IBM has their zSeries which can be logically partitioned into multiple logical system, hardware based virtualization.  Their PR/SM technology has passed the EAL 5 requirements.  This basically means that a LPAR is just as secure as a separate physical box.  Why is this important, because it means in about 40 sq. ft. I can have the equivalent of 60 servers that when needed I can "move" hardware resources from one server to another.

Now take IBM's z/VM (the original virtulization OS).  It can handled thousands of virtual machines.  A few years back on a LPAR of a IBM mainframe one site fired up 41,000 (yes forty-one thousand) Linux systems before VM ran out resources.  This LPAR had 256 MB (yes MB) of "RAM" and two phyiscal CPU's allocated to it.  Now the Linux systems were not doing much of anything, but you get the idea of what it limitation are.  This was the 32-bit version of VM, not the newer 64-bit version, which should be able to start even more virtual images.

From a real world point of view there are many companies that are now putting virtual  Linux images on mainframes.  Nationwide Insurance Company is one.  IIRC they currently have about 700 virtual Linux images running on two IBM mainframes, both with only 5-9 CPU's.  They have saved millions of dollars in software, hardware, and environmental costs and prevented a $100 million data center upgrade.  

There is a gaming company Brazil that is using virtual Linux images on zSeries mainframes.

The biggest advantage I see from large scale mainframes is resource sharing.  IBM's biggest mainframe right now, th z10, can scale up to 64-CPUs (4.5 Ghz) and 1.5 TB of "RAM" and have thousands of I/O devices.  You can have a single OS image with all of these resources or thousands of OS images sharing these resources.  If you run IBM's z/OS you can run Parallel Sysplex and have 32 z/OS images acting as a single image with self load balancing and resource management.  Each image can be on its own z/10.  So you could have 2048 CPU's, 48 TB of "RAM" being treated as a single system.  A z/10 takes up about 40 sq. ft. of floor space (this include service area) and the power draw and heat generated is less than two standard racks full of blade servers.

I know of companies that had 5 year plans to get rid of their mainframe, they are now 15 years into those 5 year plans and they are scrapping the plan.  They will have mainframes forever.   Some of these companies are companies that have very, very small mainframe, less than 500 MIPS.  

The company I work for started 5 years ago with a 5 year plan to migrate everything off the mainframe.   The 1st part of the plan was to take 3 years.  We are now 5 years into the 5 year plan and are still in the middle of the 1st part.  We have had to upgrade the mainframe twice since then.  The number of servers distributed enviroment that was to replace the mainframe has at least doubled and most of the servers have had to have the number of CPU's and amount of "RAM" doubled in order to handle the workload that is being offloaded.  The software costs and environmental costs have skyrocketed and we are now looking at moving it back onto the mainframe to reduce costs.

Now, for COBOL.  Unfortunately that may be on the way out.  Schools are not teaching it and so if a company has COBOL programs they have to train their own programmers.  IMHO COBOL is easier to learn, but I'm a systems guy that when I do code I code in assembler.
0
 
michael_chAuthor Commented:
Thanks a lot for this detailed description.
0

Featured Post

Hire Technology Freelancers with Gigs

Work with freelancers specializing in everything from database administration to programming, who have proven themselves as experts in their field. Hire the best, collaborate easily, pay securely, and get projects done right.

Tackle projects and never again get stuck behind a technical roadblock.
Join Now