ISTEK Consulting, Inc.
|
z/Linux |
Around the year 2000, a grassroots movement began a port of the Linux operating system to the zSeries hardware platform. This culminated with fully supported distributions from both RedHat and SUSE on the mainframe. Linux can run on the mainframe in two modes: either as a standalone LPAR, or as a guest operating system under z/VM. There are pros and cons to running Linux on the mainframe, which are discussed here. |
|
IFL |
In 2001, IBM introduced new functionality into their processors which allowed them to run the z/Linux operating system, either standalone or under z/VM. Although these processors contribute to the multi-processor effect and the corresponding MIPS reduction of the system, they are not visible to any of the z/OS LPARs. Consequently, they do not contribute to any of the software licensing charges that typically represent the bulk of mainframe datacenter costs. |
|
zAAP |
In 2004, in order to further reduce the third party licensing costs of running Java on the mainframe, IBM introduced a new co-processor for Java. Since this is not a full fledged processor, work has to be dispatched to it from a traditional CP. There are some Java code that cannot run on the zAAP. Additionally, you are limited to at most the same number of zAAPs as regular CPs. zAAP details are described here. |
|
USS |
Unix System Services. This is a collection of several key operating system components which gives the z/OS operating system a Unix-like personality. The main components of USS are a hierarchical file system (HFS), a Unix process model, a Unix security model, and a Unix shell. A key point to note is that USS is a fully participating member of the z/OS environment. As such, the default code pages for USS are EBCDIC, USS processes can be launched through JCL, USS processes can read MVS files. Similarly, traditional MVS programs can read HFS files. |
|
LPAR |
A logical partition. An image of an operating system which is allocated to some share of the physical resources on the hardware (such as processors, memory, etc…). Multiple LPARs typically exist on the same physical hardware. |
|
CP, Engine |
Central Processor. On the mainframe, in addition to the CP, there are specialized co-processors for doing things like I/O and encryption. This enables the mainframe to achieve enormous bandwidth between the processors and the disk subsystems (DASD) |
|
DASD |
Disk subsystems. On the mainframe, these are usually very high end systems from IBM, Hitachi, or EMC. The disk controllers are usually fully functional computer systems in their own right. For example, the IBM controllers(code named Shark) are essentially AIX Unix boxes whose sole job it is to manage the disk farms. The DASD is often connected to the mainframe through high speed cabling – these days typically fiber. On the mainframe, the technology is called FICON. Analogous technology in the Open Systems world is called FibreChannel. Of course, these two technologies are not compatible with each other (See Corigin experience report). |
|
XCF |
External Coupling Facility. IBM provides the ability to physically interconnect multiple mainframes through an external coupling facility to essentially build a cluster. The XCF is a small mainframe computer whose job it is to facilitate interconnections an maintain global data structures that are used as a foundation for other IBM clustering technologies such as DB2 Data Sharing. |
|
Data Sharing |
The ability for multiple DB2 subsystems (instances) to share the same database. If the Data Sharing is implemented over multiple physical mainframes, Data Sharing relies on the XCF to coordinate global locking and buffer pools. |
|
Buffer Pools |
Memory used by DB2 to cache frequently accessed data and index pages. |
|
EBCDIC |
the code page for mainframes. Unlike most every other operating systems which use ASCII to do character encoding, the mainframe uses EBCDIC, which is a proprietory IBM codeset. The most important ramifications of this are the cost of translating between EBCDIC and ASCII/Unicode, and the potential differences in sorting and collating sequences. The new IBM processors try to alleviate some of the conversion costs by providing built-in instructions to handle codepage conversions. Also, version 8 of DB2 will store catalog information in Unicode. However, any large scale transition from EBCDIC is unlikely in any short term horizon due to the sheer volume of legacy code that depends on it. |
|
CICS |
A transaction processing monitor. In the last 5 years, CICS has been reengineered to add support for Java, EJB, CORBA, XML, SOAP. Due to the architecture of CICS, adding Java has been somewhat of a challenge. My experience with using Java in CICS are described in this experience report. |
|
TSO |
The interactive user interface to zOS. Allows the user to type in commands and submit jobs. Similar in purpose to the Unix shells. |
|
ACF2, RACF |
Two of the most popular security subsystems for zOS. RACF is an IBM product. ACF2 is a product from CA. |
|
ISPF |
The menu driven user interface that sits on top of TSO |
|
JCL |
Job Control Language. A way to run traditional programs on zOS. |