Jump to content

ZIIP

From Alfawiki
IBM System z Integrated Information Processor (zIIP)
Company typeTechnology / Hardware component
IndustryMainframe computing
FoundedJanuary 24, 2006; 20 years ago (2006-01-24)
Headquarters
Armonk
,
United States
ProductsSpecialty processors for IBM System z mainframes
ParentIBM

In IBM System z9 and subsequent mainframe models, the System z Integrated Information Processor (zIIP) is a dedicated processor engineered to transfer particular workloads away from the general-purpose mainframe central processors (CPs). Originally developed to alleviate processing demands on IBM Db2, its functionality has since expanded to include other z/OS tasks. This approach builds upon earlier specialized processors such as the zAAP, which handles Java workloads, and the Integrated Facility for Linux (IFL), designed to run Linux and z/VM but not IBM operating systems like z/OS, DOS/VSE, or Transaction Processing Facility (TPF). System z processor units are categorized as CPs, SAPs (System Assist Processors), or specialty processors like the zIIP. While these specialty processors do not incorporate microcode or hardware acceleration enhancements, they enhance overall throughput and help lower software licensing costs by offloading work from general CPs.

The inaugural application to leverage the zIIP was DB2 for z/OS V8, with support now extended to a variety of IBM and third-party software solutions. The zIIP requires a System z9 or later mainframe platform. Compatibility is provided starting with z/OS version 1.8 and DB2 9 for z/OS, and IBM has released Program Temporary Fixes (PTFs) enabling earlier versions such as z/OS 1.6, 1.7, and DB2 V8 to utilize the zIIP. Native support is integrated within DB2 9 for z/OS.

IBM publicly announced the zIIP technology on January 24, 2006; hardware became commercially available in May 2006, followed by supporting software PTFs in June 2006.

zIIPs deliver cost-efficient capacity enhancements for four key DB2 workload categories:

  • Remote Distributed Relational Database Architecture (DRDA) access via TCP/IP protocols, including JDBC and ODBC connections across Logical Partitions (LPARs) through HiperSockets—for example, Linux running on IBM Z. While most DB2 V8 stored procedures are excluded, DB2 9’s native remote SQL procedures utilize the zIIP.
  • Parallel query processing in DB2 9 benefits from increased parallelism when executed on zIIPs.
  • XML parsing operations within DB2 can be assigned to either zIIP or zAAP processors.
  • Selected DB2 utility functions.

Support for zIIPs

Although initially designed for Db2 on z/OS workloads, the zIIP specialty CPU also supports IPSec processing within TCP/IP stacks, certain XML processing tasks, and IBM’s Scalable Architecture for Financial Reporting. In August 2007, Shadow middleware—now part of Rocket Software—introduced the first non-DB2 integration eligible for offloading onto the zIIP, thereby extending specialty engine advantages to environments including Adabas, CICS, IMS, IDMS, and VSAM. Since then, numerous independent software vendors (ISVs) have adapted their products to leverage the capabilities of zIIPs.

Prominent ISVs encompass Software AG, Compuware, CA Technologies, BMC Software, GT Software, Inc., and Phoenix Software International.

For instance, CA’s NetMaster Network Management for TCP/IP can perform both core network management tasks and packet analysis using a zIIP processor. Rocket Software’s Shadow server asserts that as much as 99% of integration-related processing—such as executing SQL queries against non-relational databases and handling Web services or SOA workloads—can be shifted onto zIIPs. GT Software’s Ivory Server for z/OS supports XML parsing, payload assembly, and data transformation on the zIIP; it also accommodates the zAAP processor through IBM’s XML Services and utilizes IFL processors running Linux on IBM Z. Ivory Server combined with Ivory Studio offers clients tools to efficiently allocate workloads between general-purpose CPUs and specialty processors.

Commercial software developers may obtain access to detailed IBM technical information under Non-Disclosure Agreements to optimize their applications for effective use of the zIIP processor.

With the introduction of the IBM z13 system, all former zAAP functionalities have been consolidated into the zIIP processor (known as zAAP on zIIP), thereby enabling all previous zAAP workloads to execute on the zIIP. Furthermore, both IFL and zIIP processors in the IBM z13 incorporate the advanced IBM z13 microprocessor architecture featuring simultaneous multithreading (SMT) capabilities.

Use of zIIPs

IBM Z Operational Log and Data Analytics, along with IBM Z Anomaly Analytics with Watson, both support the utilization of zIIPs. IBM Z Operational Log and Data Analytics gathers IT operational data from z/OS environments, converts it into a format suitable for consumption, and streams this data to either external analytics platforms or its proprietary analysis tool, enabling the identification of operational challenges. Meanwhile, IBM Z Anomaly Analytics leverages historical IBM Z metrics and log data to establish models of normal system behavior and then examines live data streams to identify and notify about irregularities.

Within these solutions, when the Z Common Data Provider transmits operational data, the zIIP offload capability can be activated. The System Data Engine module shifts eligible processing tasks from general purpose processors (GCPs) to zIIP processors, thereby decreasing MIPS usage on GCPs and contributing to a reduction in overall ownership costs.

It is important to note that this offloading process may introduce additional CPU overhead. In cases where zIIP capacity is limited, z/OS reroutes eligible workloads back to general CPUs if zIIPs are fully utilized. Consequently, the extra CPU time required might surpass the savings gained from offloading, which could lead to an increase in general CPU consumption.

See also

External links