One elephant that arguably should be in the room labelled 'cloud computing’ is the mainframe computer. Yet it still remains largely absent. But there are occasional sightings of services that are in the mainframe space.
The latest comes from Cloud Compiling, which plans to use next month’s mainframe-centric SHARE conference in Boston as the launch vehicle for a range of virtualized language compilers for IBM’s primary mainframe operating system, z/OS.
These will include virtualized versions of its COBOL, PL/1, Fortran and C/C++ compilers for z/OS. The company believes they are the first examples of cloud-delivered compiler services for IBM’s flagship behemoth systems. They can be installed either in a private cloud or externally hosted with Cloud Compiling as a SaaS-delivered service. It is expected to bring significant cost savings, greater systems controls, increased flexibility, and the ability to scale quickly.
Cloud Compiling is easy to install, because there is no migration to another platform, as there has been with previous attempts to provide other applications development options than the mainframe itself. So there are no changes required to IPL’s, authorized libraries, supervisor calls, z/OS parmlib or existing JCL. The company claims it can be installed in just a few hours and allows users to run what are, in effect, the same compilers they have always used. One unnamed user has been quoted and stating that the only change they found was the saving of $100,000 a year.
This, for mainframe users, could represent a tempting option. The costs associated with developing applications for IBM’s mainframes is a long-running issue. Running compilers on mainframes, often with several iterations before the code is finalized for production use, can be an expensive process because of the licence structure attached to mainframe utilisation by IBM.
This has led over the years to companies springing up offering low-cost alternative platforms on which to conduct applications development. This approach, however, has had its problems. Early attempts at using PC-based software proved largely inadequate when scaling up from a pilot development, while all other platforms have involved some degree of code conversion or simulation/emulation. It has also led to confrontations with IBM’s legal department, keen to ensure that its licences are not breached by either users or third party suppliers.
It will be interesting to see how a cloud-based approach is received.
[1]