The stack, from metal to your program
Between the physical machine and the COBOL program you will eventually write, there are about five layers. Knowing their names makes almost every other conversation on this platform easier to follow.
The layers
What each one does for you
- The machine — often called the CPC or 'the box'. One machine holds many processors, a very large amount of memory, and dedicated I/O hardware.
- PR/SM — carves the machine into LPARs. This is firmware-level virtualisation that has been in production since the 1980s, and it is the reason one machine can safely run production and test side by side.
- z/OS — the operating system most business workloads run under. It manages memory, schedules work, drives I/O and enforces security. Other operating systems exist on the same hardware: z/VM, z/VSE, z/TPF and Linux on Z.
- Subsystems — long-running services that your programs use. CICS runs online transactions; DB2 stores relational data; IMS does both in a different style; MQ moves messages.
- Your program — usually small, usually called by something else, usually doing one clear job.
A concrete path
When a customer taps their card, the request typically arrives over the network into a CICS region running under z/OS in one LPAR of a physical machine. CICS starts a transaction, which runs your COBOL program, which issues SQL to DB2 and returns an answer. All six layers above are involved in something that takes a few milliseconds.
Common mistakes
Being precise about the layer — LPAR, z/OS, CICS region, DB2 subsystem — is what makes you quick to help and easy to work with.
A single machine commonly runs a dozen or more LPARs: production, development, test, and often other companies' work at a service provider.
What you will see at work
- Your site will have LPAR names you have to learn, often four characters and cryptic. Ask for the list on your first day; there is always a list.
- Production and test usually live on the same physical machine in different LPARs. That is normal and safe — but it means 'it worked in test' is not a statement about hardware.
- You will hear 'sysplex': two or more z/OS systems clustered so they share data and work. Its purpose is that maintenance never requires an outage.
Key terms
Check your understanding.
Take this lesson's quiz and save your progress. Free.