The Hitchhiker’s Guide to the Mainframe – The Basics

Pages: 1 2 3


In this article I will provide a brief history of the mainframe, with an emphasis on features that make a mainframe unique. I will not attempt to cover the architecture in any technical detail at this time; the intent is simply to set the stage for future articles. If you are in fact an experienced mainframe user then this article probably has no particular value for you, other than possible amusement. I will assume that the reader is familiar with general computer concepts and technologies, and will present the information in terms that should be familiar to PC or Unix geeks.

Why “Hitchhikers Guide”? I am a big fan of the works of the late great Douglas Noel Adams, especially his “The Hitchhiker’s Guide to the Galaxy”, a trilogy in 5 (6?) books. For those of you who have not read these books – DON’T PANIC – any references you don’t understand can be safely assumed to be Somebody Else’s Problem – they are not important – the answer is 42, and that’s all you need to know. Of course if you really want a better understanding of the galaxy and how it will affect your next party, then maybe you should read the books. ;-) I was given the nickname “Ford Prefect” (one of the story’s lead characters) many years ago by a group of friends who were also HHGG fans. So sit back, relax, and enjoy an Ole’ Janx Spirit while I introduce you to the world of mainframe computer systems. First, let’s define a few terms:

What is a mainframe?

The term mainframe was originally coined to refer to the very large computer systems that occupied very large steel framed boxes, and was used to differentiate them from the smaller mini- or micro-computers. While the term has been used in various ways over the years, it is most often used to describe the successive families of IBM computer systems starting with System/360. This would also apply to compatible systems built by other companies such as Amdahl and Hitachi Data Systems (HDS). This is the definition we will be use for the remainder of this series of articles.

Although some have used the term “mainframe” to refer to IBM’s AS/400 or iSeries systems, this is an improper use of the term; even IBM considers those systems as mid-range servers rather than mainframes.

Another more humorous definition of a mainframe can be found in Isham Research’s Devil’s IT Dictionary: “an obsolete device still used by thousands of obsolete companies serving billions of obsolete customers and making huge obsolete profits for their obsolete shareholders. And this year’s run twice as fast as last year’s.”

What is an I/O channel?

A mainframe channel is somewhat analogous to a PCI bus, it is capable of connecting one or more “controllers”, each of which controls one or more “devices” (disk drive, terminal, LAN port, etc.). One primary difference is that mainframe channels connect to controllers via either pairs of large “bus and tag” cables (for parallel channels) or, more recently, fiber optic ESCON (Enterprise System CONnection) cables (serial channels) and FICON (fiber channel). In the early days these channels were external boxes (about 6’x30″x5’H each), but are now integrated into the system frame.

These channels are a key part to one of the great strengths of the mainframe – its massive I/O capabilities. I will touch on the capabilities of the I/O subsystem later in this article.

What is DASD?

DASD is an acronym for “Direct Access Storage Device”; IBM coined this term to refer to any type of storage that was directly (randomly) addressable. Today the term is just what we call disk drives, but in the olde days it also referred to drums and datacells and such. What is a datacell? Well, before disk drives became cheap, fast, and popular IBM had a device that was basically a drum around which it would wrap one of many different magnetic strips (cells), then read/write information on data tracks on the spinning cell. The method of accessing the data was identical to that for a disk, but seeks involving cell changes were obviously measured in seconds. <G> Datacell devices also had a nasty habit of jamming when trying to dismount a cell into its storage slot, sometimes causing physical damage to the media – we’ve come a long way baby !

What is an LPAR?

An LPAR (Logical Partition) is a type of virtual machine implemented by PR/SM (Processor Resource/System Manager) a hardware/firmware feature on all recent mainframes. PR/SM allows you to run up to 15 LPARs on a single system, each having dedicated real storage (RAM) and either dedicated or shared CPU(s) and channel(s). Since the functions most important to performance are implemented in the CPU, there is very little performance cost. IBM has announced its intention to expand this support to more than 15 LPARs in the near future.

Pages:   1 2 3  Next »

Discuss (4 comments)