Click here to load reader
Upload
chris-daemen
View
110
Download
0
Embed Size (px)
Citation preview
by Joel Starkman
A White Paper
The IBM zIIP Specialty Engine for
Mainframe FOCUS
A Benchmark of Significant CPU Cost Reductions
Joel Starkman is director of Operations for the FOCUS Division at Information
Builders, responsible for the creation, delivery, and support of all FOCUS
releases and their functional consistency with the back-end of iWay Software
and WebFOCUS. Joel is a 30-plus-year veteran of FOCUS application
development for Fortune 100 companies, using FOCUS for everything from
financial systems reporting to real-time control of an industrial robot.
Joel guided the functional implementation of zIIP support for the FOCUS
product, working closely with the programming staff to coordinate and test
the user-visible and efficiency aspects of its implementation.
Joel Starkman
Executive Summary
zIIP Benchmark Tests Recap
Background – Advantages of the zIIP Specialty Engine
Mainframe FOCUS Utilization of the zIIP
Is Implementing the zIIP for FOCUS Beneficial?
Local Adjustments That Improve zIIP Usage
Benchmark Tests and Statistical Rewards
Benchmark: Database Variations
Benchmark: Report Size Variations
Benchmark: Typical Application Activities
Conclusion
Appendix I
zIIP Usage Statistics
Appendix II
IBM Workload Manager Tuning
Types of Goals
Appendix III
Glossary
Table of Contents
2
5
7
8
9
13
14
14
15
15
1
3
4
5
6
11
13
14
15
Information Builders1
Executive Summary
As of Release 7.6.10, FOCUS is enabled to run on the system z Integrated Information Processor
(zIIP) engine. This white paper demonstrates the types and approximate amounts of FOCUS
workload transferred to the zIIP under various conditions, and the resulting cost reductions.
The zIIP specialty engine from IBM offloads CPU-intensive workloads from the central processors
(CPs). Since the MIPS capacity of the zIIP engine is not included in the overall rating of the
mainframe, all work incurred using the zIIP engine is free from IBM. Installing a zIIP can be a way to
delay a system upgrade, and it may also be a performance enhancer for products enabled to run
there and for the overall system.
About 80 percent of FOCUS code is able to run on the zIIP, which accounts for as much as 95
percent of CPU capacity, though it will more generally be in the 30 to 80 percent range. These
estimates are above the already documented 20 to 40 percent performance improvement of
Release 7.6 over prior releases due to our continuous improvement program using all of the newer
technologies available. During the zIIP enablement development effort, all of the technical and
legal implementation restrictions imposed by IBM were rigorously respected.
Some internal processes must continue to run on the central processor due to IBM specifications
regarding the types of workload the zIIP may process. Productive zIIP activity is monitored by
FOCUS, and FOCUS reacts accordingly to ensure that IBM-billable CPU charges are minimized
against zIIP-related overhead.
The actual amount of processing diverted to the zIIP is largely dependent on the goal settings
declared locally to the IBM Workload Manager (WLM), which specifies the priority of FOCUS access
to the zIIP in competition with other software. FOCUS provides methods for the user to query zIIP-
related CPU information and improve zIIP-related processing. There are also techniques available
to adjust the application to best take advantage of the zIIP engine when running FOCUS.
If you do not have a feel for how much a zIIP engine installed at your site can reduce the cost
of FOCUS processing, you may simulate the presence of a zIIP engine if none exists, or you may
simulate giving full access to FOCUS processing on an existing zIIP. System statistics will then tell
you how FOCUS would perform with full zIIP advantage.
The IBM zIIP Specialty Engine for Mainframe FOCUS2
zIIP Benchmark Tests Recap
Information Builders designed benchmark tests to evaluate the amount of workload transferred to
the zIIP under specific conditions and test scenarios. The results summarize as follows:
DB2 results vary, depending upon the local configuration of FOCUS and DB2.
CPU Savings
Type of Report or Activity Using zIIP Interpretation
Executive Summary Report 77% Roll-up, out to PDF
Operational Report 64% Medium size
Extract Report 66% Show 100% of file
Exception Report 51% Extract a few records
Database Type Variations 1-88% Various sort methods and destinations*
Reporting Scenarios 59-99% One million records, various complexities
Transaction Processing 42-95% One million records, load, update, etc.
Actual Applications 46% Weighted average of five applications’ usage
Performance Benchmarks 76% Weighted average of targeted internal processes
* Several database types generate low zIIP advantage due to imposed record retrieval methods.
Information Builders3
IBM’s zIIP specialty engine is the newest of several specialty engines that IBM has implemented
for mainframe consumption. The zAAP and IFL engines have been highly successful for JavaTM
execution, XML parsing, and Linux applications, respectively.
The zIIP engine offloads workload from the CP, also referred to as the general processor (GP). The
MIPS capacity of the zIIP engine does not count toward the overall MIPS rating of the mainframe
image, so all CPU usage incurred on the zIIP is not chargeable from IBM. Effectively, all workload
run on the zIIP is free.
The zIIP engine is factory-identical to a central processor (CP). It is restricted at installation
time via micro-code to perform specific types of workloads, but always runs at 100 percent of
processor capacity. The zIIP offloads heavily CPU-intensive workloads, leaving the CP more time
to absorb otherwise queued workloads as well as to perform its dedicated tasks of running the
operating system, handling I/O interrupts and timer interrupts, initiating jobs, and controlling user
interactions with the operating system. Therefore, some overall performance improvement can be
perceived across all mainframe activity. The true benefit of the zIIP is its cost reduction effect and
its potential contribution to delaying a system upgrade.
The actual zIIP benefit achieved for FOCUS or any product depends on the response time
goals declared by the local system administrator to the WLM when defining the priority of the
software’s access to the zIIP in competition with other software.
Background – Advantages of the zIIP Specialty Engine
The IBM zIIP Specialty Engine for Mainframe FOCUS4
As of Release 7.6.10, the zIIP engine is accessible to FOCUS for processing much of the typical
workload associated with a FOCUS request. zIIP activation begins upon user issuance of a SET
command if all applicable conditions for zIIP use pass properly. If the zIIP is not available to the lpar,
then processing simply continues on the CP; FOCUS continues to run. See the Benchmark Statistics
section (pp.7-13) that examines the gains achievable by use of the zIIP.
FOCUS diverts eligible workload to the zIIP by switching from TCB mode (for instructions that can
run only on the central processor) to SRB mode that engages a preemptible enclave for secure
execution of enabled workloads on the zIIP engine. Though a large percentage of the workload
is eligible for execution on the zIIP, the actual amount permitted to run on the zIIP at any moment
(therefore the benefit achieved for the FOCUS user) largely depends on the response time
goals declared in the WLM by the local system administrator; see the IBM Workload Manager
appendix (p.15).
During the zIIP development effort, all of the technical and legal implementation restrictions
imposed by IBM were rigorously respected. The major factor that affects zIIP’s performance is
IBM’s restriction that the zIIP does not handle I/O interrupts. In applications that require significant
database interrogation, high-volume sorting, or the use of third-party tools or user functions
during processing (most of which are present in typical applications), passing information among
these environments requires switching out of SRB (zIIP) mode into TCB (non-zIIP) mode to
communicate, and then back again to continue processing. This switching can occur thousands or
even millions of times during a single request. Although each switch is miniscule, the cumulative
effect can absorb visible amounts of CPU on both the zIIP engine and the CP. So optimizing this
communication is critical to maximizing the savings attained by using the zIIP.
FOCUS implementation of zIIP-enablement has done just that. FOCUS buffers the records passed
to the system sort utility and some of the adapters, rather than performing record-by-record
transmission, thus generating only one switch for each buffer-full.
FOCUS constantly monitors zIIP usage. When FOCUS detects that the cost ratio of switching
modes compared to work redirected to the zIIP crosses a non-productive threshold during a
task, FOCUS dynamically may decide to no longer use the zIIP just for the duration of that task or
command. Such situations may include pure extracts with few calculations from certain database
types where non-zIIP-able I/O is the dominant activity, or a very low aggregation ratio (such
as PRINT might induce) while extracting from a highly disorganized file. In every case, FOCUS
endeavors to ensure that the user spends no more on CPU cost with the zIIP than they would have
without the zIIP.
Mainframe FOCUS Utilization of the zIIP
Information Builders5
If you do not have a zIIP at your site and you are not sure how much a zIIP would benefit FOCUS
processing, then add PROJECTCPU=ON to your sys1.parmlib table and run several jobs. This parm
will make FOCUS believe that a zIIP is available and attempt to redirect its work there. You can then
measure what the advantage of a fully accessible zIIP would be by looking at the &FOCZIIPONCP
reserved amper-variable in FOCUS. This value shows the amount of CPU that would have been
used by FOCUS on the zIIP, but was redirected to the CP by the system because the zIIP was
unavailable. With these results, you can see how beneficial the zIIP would be if a zIIP of sufficient
capacity and priority directed to FOCUS were actually available.
In another situation, you may already have a zIIP accessible to FOCUS, but FOCUS has insufficient
priority access to it such that FOCUS is currently deriving limited zIIP benefit. Under this
circumstance, try: SET ZIIP=SIMMAXZIIP in FOCUS. This will allow the job to run while simulating
full zIIP accessibility to see how much zIIP access the job would have had if the zIIP were 100
percent available for FOCUS usage.
Local Adjustments That Improve zIIP Usage
User-controllable factors that can improve the effect of the zIIP on the FOCUS application include:
Declare policy and goals of WLM to allow FOCUS to access the zIIP with sufficient priority ■
(compared to other zIIP-competing software) to derive significant CPU benefit. Negotiate with
operations to maximize the priority of FOCUS for zIIP access
Maximize the block size of files that are read or written by FOCUS to reduce the number of I/Os ■
required to access the file and therefore the number of switches to non-zIIP mode that FOCUS
would have to make
Move non-Information Builders 3GL functions from DEFINEs to COMPUTEs to reduce the ■
switching to non-zIIP mode for each such call, or rewrite them as DEFINE FUNCTIONS to avoid
the switch entirely
Is Implementing the zIIP for FOCUS Beneficial?
The IBM zIIP Specialty Engine for Mainframe FOCUS6
The benchmark tests demonstrate the relative amount of FOCUS workload transferred to the
zIIP during processing of several scenarios varying the file-type and type of request or activity.
These scenarios are selected as approximations of typical customer applications, and several tests
represent more extreme boundary situations. From these results, you may project the savings for
your particular application in accordance with how it parallels these examples.
Under these benchmark tests, WLM was directed to give highest priority to FOCUS, ensuring that
all zIIP-eligible code actually ran on the zIIP rather than having some percentage diverted to the
central processor. Note that some parts of FOCUS processing must always be done on the CP due
to zIIP restrictions.
The following categories of tests were developed and implemented to demonstrate the degree to
which FOCUS uses the zIIP in each instance:
Database Variations
Report Size Variations
Typical Application Activities
Benchmark Tests and Statistical Rewards
Full extractions from various database types, all with
identical database structures, contents, and sort orders.
Each test varies only one factor from the original
Typically sized customer reports in terms of data volume
extracted and report complexity, performed against the
same file
Various activities of reporting, data manipulation, and file
maintenance, most of which exist somewhere in a typical
application, plus execution scenarios of several actual
applications
Information Builders7
Benchmark: Database Variations
These tests are full extracts from various database types, all with identical database structures,
contents, and sort orders. Each test varies only one factor from the first test in the list. FOCUS sort
is used unless otherwise stated.
Use zIIP No zIIP
Benchmark Case zIIP CP CP Save
(cpusec) (cpusec) (cpusec)
FOCUS file 36.94 13.73 38.74 65%
FOCUS file to Excel 52.11 7.45 60.89 88%
FOCUS file (DFSORT SORT) 18.14 19.35 27.47 30%
VSAM file 39.57 18.47 59.56 69%
Flat file 45.26 13.22 47.98 73%
IMS file 0.16 69.19 69.59 1%
DB2 file (optimized) 1.91 30.60 30.31 1%
DB2 file (unoptimized) 32.50 57.22 80.09 29%
Scenario: Four million records, all files with identical data, reduced down to 7,500 lines.
Analysis
Savings against a FOCUS file using FOCUS sort reach a respectable 65 percent ■
Outputting to Excel (or other styled formats) requires intensive additional CPU. Since virtually all ■
of that work is performed on the zIIP, the percent savings is much better
Using DFSORT means that the records to be sorted are transferred between FOCUS and the sort ■
package, requiring some buffered switching between zIIP and CP. There is still significant savings
over non-zIIP execution, but not as broad a percentage drop as with FOCUS sort
Newly buffered VSAM file data movement appears to be very efficient in terms of zIIP ■
participation
Flat file processing is highly effective, especially with a high blocking factor, which has always ■
been typical and prudent when reading or writing such files
Extracts from IMS files gain virtually no benefit from the zIIP since records exchanged between ■
IMS and FOCUS must be handled individually, i.e., there is no buffering. So switching between
zIIP and CP to satisfy the I/O restrictions becomes too expensive very quickly. The FOCUS zIIP
Monitor detected this situation early in the extract phase and completed the majority of the
extract task on CP, protecting the user from excessive costs; note that the CP costs of zIIP versus
non-zIIP were approximately the same. However, if the result had been directed to a styled
output like Excel or PDF, that CPU-intensive part of the processing would have occurred on the
zIIP, making the overall savings ratio of the request somewhat more skewed toward the zIIP
The IBM zIIP Specialty Engine for Mainframe FOCUS8
Both DB2 and FOCUS are independently zIIP-enabled products; either product’s choice to use ■
the zIIP is not directly influenced by the other product. For an optimized DB2 request, DB2 does
the majority of the work (extract, sort, aggregation) and FOCUS performs the post-retrieval
computations, report formatting, and any output destination processing, most of which runs on
zIIP. Unoptimized DB2 requests show more FOCUS/zIIP usage because FOCUS performs more
of the processing after DB2 merely returns all extracted records to FOCUS. CP usage is higher
for unoptimized cases (and always has been in a pre-zIIP environment). But assuming that a
particular request must be unoptimized due to its requirements, running it via zIIP still costs
much less than without it. DB2’s usage of the zIIP depends upon how it is locally configured
relative to FOCUS
Other database adapters supported by FOCUS (such as Adabas, IDMS, Millennium, Model 204, ■
Oracle, Teradata) have not yet been benchmarked, but are expected to be relatively consistent
with IMS results for much the same reasons as shown in the table for IMS in the first release.
Continuing work is being done and improvements are expected
Benchmark: Report Size Variations
These reports were specifically designed to approximate typical customer usage in terms of data
volume and report complexity. Tests were run under FOCUS Release 7.6.10.
Use zIIP No zIIP
Benchmark Case zIIP CP CP only Save
(cpusec) (cpusec) (cpusec)
Executive Summary (roll-up, out to PDF) 2.49 0.91 3.91 77%
Operational Report (medium size) 3.18 1.82 5.08 64%
Extract Report (show 100% of file) 5.99 2.93 8.69 66%
Exception Report (extract few records after full table scan) 1.10 0.86 1.75 51%
Scenario: One million-record FOCUS file
Analysis
Savings are consistently over 50 percent ■
As the Executive Summary noted, PDF output and aggregation are highly CPU-intensive ■
activities, resulting in the largest savings
The Exception Report does very little formatting or aggregation since it is a short-resulting ■
report. Most of the cost is in extracting the data, which requires repeated switching to the CP, so
there is relatively reduced zIIP benefit
The Operational Report and Extract Report show intermediate savings and are close only by ■
coincidence. Their major activities (in addition to a full table scan) occur in temperate doses,
such as aggregation, sorting, HOLD output (in the case of extract), computations, and number of
fields requested
Information Builders9
Benchmark: Typical Application Activities
These tests represent characteristic requests in mainstream customer applications including
TABLE, MODIFY, MAINTAIN, and other generally observed activities. As typical scenarios, they are
reused here for zIIP comparison purposes.
Use zIIP No zIIP
Benchmark Case zIIP CP CP Save
(cpusec) (cpusec) (cpusec)
TABLE
Detail extract of 1M records (PRINT * HOLD) 3.1 4.4 7.6 42%
TABLE using IF for nonexistent record (full database scan) 1.1 0.6 1.4 57%
1M recs. SUM’ed to top level – 10 lines 2.4 0.5 3.0 83%
1M recs. SUM’ed to top level, DEFINE on every record 2.7 0.6 3.3 82%
1M recs. SUM’ed to 3rd level – 200 lines 2.4 0.5 3.1 84%
1M recs. SUM’ed to 3rd level – DEFINE on every record 3.2 0.5 3.8 87%
200K recs SUM’ed with COMPUTE 2.3 0.6 3.2 81%
COUNT on keys of all 4 levels – read all pages 2.6 0.6 3.1 81%
1M records SUM’ed to 1st level using MIN. and MAX. on 4th level field 2.6 0.5 3.2 87%
IF test on 4th level non-key field retrieves 10,000 records, but reads all 1M 1.1 0.6 1.5 60%
JOIN 100K record flat HOLD (400 bytes wide) to 1M record FOCUS file 1.7 0.4 2.1 81%
JOIN 100K record FOCUS HOLD to 1M record FOCUS file 1.6 0.4 2.1 81%
HOLD 100K records, then MATCH to 1M record FOCUS file 2.3 0.5 2.6 81%
Table 100K 400-byte records once, then HOLD PDF, HTML, EXL2K, WP 16.2 0.3 21.0 99%
Non-IF’able WHERE against 1M recs 2.9 0.5 3.37 85%
Extract 50K records from VSAM file 0.3 0.4 0.7 43%
MODIFY and MAINTAIN
COMBINE and MODIFY 21 files 0 .8 1.3 2.2 41%
MODIFY include 100K transactions into new FOCUS file 1.5 0.2 1.6 88%
MODIFY update 100K same sorted transactions selected from 1M 2.3 0.6 3.0 80%
MAINTAIN load 100K sorted 400-byte transactions 2.1 0.1 2.4 96%
UTILITIES AND MISCELLANEOUS
? FILE and ? FDT statistics on 71,000 FOCUS pages 0.4 0.4 0.7 43%
REBUILD DUMP/LOAD 1M records restrict to 100K recs 0.8 0.2 1.1 82%
COMPILE four 500-2500 line MODIFYs with up to 50 CRTFORMs each 3.8 0.1 5.0 98%
Load flat file via 10,000 Dialogue Manager – WRITEs 6.7 0.0 8.2 99%
Weighted average of 77 activities on 5 in-house production FOCUS appls 26.7 19.8 36.6 46%
Weighted average of FOCUS General Benchmarks for Performance 3.3 1.3 4.4 70%
Analysis
We see superior results on TABLE commands, most residing in the 80-plus percent savings range. ■
On normal FOCUS file sizes of one million records, the effort of I/O work to read the file trades off
well with average calculation density and output formatting requirements
The full table scans (one for all records and one for no records) both conclude with 40 to 60 ■
percent cost savings since they encounter few calculations and minor report formatting. Most of
the cost is in performing the I/O, which incurs switching overhead
The HOLD outputs to various styled formats are highly CPU intensive, calculation-based ■
operations. So the vast majority of the post-extraction processing is performed on the zIIP,
resulting in a 99 percent CPU cost savings
The IBM zIIP Specialty Engine for Mainframe FOCUS10
MODIFY transaction processing did well, CPU savings-wise. The voluminous COMBINE structure ■
incurred more I/O as it navigated the files, thus it produced a somewhat smaller return. The
MAINTAIN test is intentionally similar to the MODIFY test, so it does not even exercise the CPU-
intensive strengths of MAINTAIN stacking, yet it performed impressively
? File and ? FDT are entirely I/O operations, so their CPU savings are merely respectable. REBUILD ■
is largely an I/O operation, but the record constriction induced many calculations, so the 73
percent savings is a compromise
The COMPILE of a MODIFY is entirely oriented toward syntax parsing and internal instruction ■
flow generation, so 98 percent savings is fully expected
The five in-house production applications are several of the custom tools that drive the daily ■
business at Information Builders, and are highly complex in terms of exercising the FOCUS
language and pushing its capabilities. 46 percent falls in the projected range of cost reductions
for most customer applications
The General Benchmarks for Performance tests intensively target low-level internal operations. ■
Seventy percent cost reduction is in line with the application results above since the applications
by definition all use some combination of these low-level operations
Information Builders11
The zIIP engine offers tremendous cost savings to the FOCUS user by offloading FOCUS workload
from the CP to the zIIP engine on which all CPU utilization is free of charge. FOCUS provides this
capability as of Release 7.6.10.
It is important for the local system administrator to allocate sufficient WLM priority for FOCUS’
access to the zIIP engine to take full advantage of the capability and savings offered.
FOCUS always ensures that zIIP usage for a specific request is of value in terms of cost, and will
protect the user from incurring higher costs due to the zIIP-related resource requirements of that
request.
Conclusion
The IBM zIIP Specialty Engine for Mainframe FOCUS12
zIIP Usage Statistics
zIIP-related usage statistics can indicate the benefit gained and CPU costs saved when the
zIIP engine handles some or most of the job’s workload. Properly interpreting the statistics is
important to this analysis. FOCUS provides several reserved amper variables that deliver pertinent
statistics about zIIP processor usage:
&FOCZIIPCPU CPU incurred on zIIP engine, normalized to the speed of CP
&FOCZIIPONCP CPU of work enabled for zIIP but sent to the CP by Workload
Manager and the dispatcher
&FOCCPU CPU incurred on CP; includes &FOCZIIPONCP
&FOCZIIPONCP indicates the amount of CPU spent by code that FOCUS had enabled for
execution on the zIIP but was diverted to the CP by Workload Manager and the dispatcher due
to circumstances of system load and competitive priorities at that moment. Optimally, this value
should be zero for the most savings in execution time and CPU cost. Note that any non-zero
amount of &FOCZIIPONCP is already included in &FOCCPU since it was actually incurred on the CP.
The total CPU for the step is the sum of &FOCZIIPCPU and &FOCCPU, intentionally excluding
&FOCZIIPONCP from the calculation.
Assuming &FOCZIIPONCP is zero or very small, then high amounts of &FOCCPU compared to
&FOCZIIPCPU implies that much of the processing required by the request was not eligible for zIIP
execution. Examine the recommendations in the “Local Adjustments Affect Actual zIIP Impact”
section for possible ways to shift more of the execution of this request to the zIIP engine.
Appendix I
Information Builders13
IBM Workload Manager Tuning
WLM prioritizes queued workloads for distribution among the central processors and zIIP
processors based on a complex set of goals and rules established by the system administrator for
each product or other definable category. This applies to all workloads from all sources, not just
for FOCUS. All of these goals described here combine to influence the probability that FOCUS
requests are directed to the zIIP engine at any particular moment.
Types of Goals
Average Response Time Goals ■ declare that all jobs must finish within a declared average
amount of time – a continually recalculated value
Percentile Response Time Goals ■ state that a certain percent, say 90 percent, of all jobs must
finish within a given amount of time. The advantage of this type of goal is that it is not affected
by rogue jobs (the other 10 percent) that run much longer and effectively skew the Average
Response Time Goal toward improper expectations
Discretionary Goals ■ are for lesser priority jobs that given resources only when spare time exists
among higher priority workloads
Velocity Goals ■ complement the other goals by declaring an acceptable time by which any
job may be delayed once it is ready to run, ranging from “run immediately” to “keep it plodding
along until it is done”
System Goals ■ do not specifically belong to WLM, rather they are general operating system
targets for recognized types of work in several service classes
Appendix II
The IBM zIIP Specialty Engine for Mainframe FOCUS14
Glossary
3GL Third-generation language such as C, COBOL, PL/1, FORTRAN, or others
(including Assembler, which is a 2GL) in which subroutines may be written
and called from FOCUS (considered a 4GL).
Central Processor The general purpose chip of the IBM mainframe that typically handles the
mainframe workload.
Client SRB Created and executed like an ordinary SRB, but runs with client (scheduler)
dispatching priority and is preemptible.
CP Abbreviation for Central Processor, also known as General Processor. See
Central Processor.
Dispatcher Distributes workloads among available processors, depending upon
current workload conditions and target goals of WLM. Responsible for
determining which zIIP-eligible tasks may be offloaded to the zIIP engine
at any specific moment based on competing task priorities and goal
achievement targets.
Enclave Entity that encapsulates the execution units (TCBs and SRBs), which
execute programs on behalf of the same work request.
Enclave Service Enable workload manager to create and control enclaves
Enclave SRB Created and executed like an ordinary SRB, but runs with Enclave
dispatching priority and is preemptible.
General Processor Another name for Central Processor. See Central Processor.
GP Abbreviation for General Processor. See General Processor.
IFL IBM specialty processor for Linux applications.
I/O Input/Output. The act of reading or writing part of a file from disk or other
storage device. That part may be an individual record, or a block of records
determined by the blocksize of the file, or an entire track.
Preemptible Allows the dispatcher to interrupt a task at any time to run other work at
the same or higher dispatching priority. Non-preemptible units (like local
SRBs), once dispatched, continue to run until they complete or incur a
voluntary interrupt like suspend/page fault.
SRB Service Request Block. Dispatchable unit (DU) runs at supervisory priority;
not preemptible (zIIP-able). This is the mode in which FOCUS runs on
the zIIP.
Appendix III
Information Builders15
Switch In the context of zIIP-related workloads, the intentional movement of a
portion of code processing from zIIP to non-ZIIP execution or vice versa as
required by IBM zIIP processor restrictions.
TCB Task Control Block. Dispatchable unit (DU) runs at dispatching priority of
address space; preemptible (non-zIIP). This is the mode in which FOCUS
runs on the Central Processor.
Velocity The acceptable amount of delay a process can be allowed to incur when
competing for execution among other tasks.
WLM Abbreviation for WorkLoad Manager. See Workload Manager.
Workload Manager z/OS monitoring tool that actively adjusts priorities of workload tasks so
the dispatcher can effectively distribute them among available processors
with the objective of achieving WLM performance goals.
zAAP Engine An IBM specialty processor for Java execution and XML parsing.
zIIP Engine An IBM specialty processor; accronym for z Integrated Information
Processor. Configured from the Central Processor chip via selective
microcode adjustments at installation time, the zIIP engine offloads CPU-
intensive workloads from the CP. The CPU capacity of the zIIP engine is not
counted toward the MSU capacity of the mainframe images, thus all work
processed by the zIIP engine is effectively free of charge.
Worldwide Offices
Corporate Headquarters Two Penn Plaza, New York, NY 10121-2898 (212) 736-4433 Fax (212) 967-6406 DN1101086.0909
informationbuilders.com [email protected]
Canadian Headquarters 150 York St., Suite 1000, Toronto, ON M5H 3S5 (416) 364-2760 Fax (416) 364-6552
For International Inquiries +1(212) 736-4433
Copyright © 2009 by Information Builders. All rights reserved. [85] All products and product names
mentioned in this publication are trademarks or registered trademarks of their respective companies.
Java and all Java-based marks are trademarks or registered trademarks of Sun Microsystems, Inc.
in the U.S. and other countries.
Printed in the U.S.A.
on recycled paper
North America
United States ■ Atlanta,* GA (770) 395-9913 ■ Baltimore, MD Professional Services: (703) 247-5565Boston,* ■ MA (781) 224-7660Channels, ■ (800) 969-4636Chicago,* ■ IL (630) 971-6700Cincinnati,* ■ OH (513) 891-2338Dallas,* ■ TX (972) 490-1300Denver,* ■ CO (303) 770-4440Detroit,* ■ MI (248) 641-8820Federal Systems,* ■ DC (703) 276-9006Hartford, ■ CT (860) 249-7229Houston,* ■ TX (713) 952-4800Los Angeles,* ■ CA (310) 615-0735Minneapolis,* ■ MN (651) 602-9100New Jersey* ■ Sales: (973) 593-0022New York,* ■ NY Sales: (212) 736-7928 Professional Services: (212) 736-4433, ext. 4443Orlando,* ■ FL (407) 804-8000Philadelphia,* ■ PA Sales: (610) 940-0790Phoenix, ■ AZ (480) 346-1095Pittsburgh, ■ PA Sales: (412) 494-9699St. Louis,* ■ MO (636) 519-1411San Jose,* ■ CA (408) 453-7600Seattle, ■ WA (206) 624-9055Washington,* ■ DC Sales: (703) 276-9006 Professional Services: (703) 247-5565
Canada
Information Builders (Canada) Inc.Montreal* ■ (514) 421-1555Ottawa ■ (613) 233-7647Toronto* ■ (416) 364-2760Vancouver ■ (604) 688-2499
Mexico
Information Builders MexicoMexico City ■ 52-55-5062-0660
Australia Information Builders Pty. Ltd.
Melbourne* ■ 61-3-9631-7900Sydney* ■ 61-2-8223-0600
EuropeBelgium* ■ Information Builders Belgium Brussels 32-2-7430240France* ■ Information Builders France S.A. Paris 33-14-507-6600Germany ■ Information Builders (Deutschland) Eschborn* 49-6196-77576-0Netherlands* ■ Information Builders (Netherlands) B.V. Amsterdam 31-20-4563333Portugal ■ Information Builders Portugal Lisbon 351-217-217-400Spain ■ Information Builders Iberica S.A. Barcelona 34-93-344-32-70 Bilbao 34-94-452-50-15 Madrid* 34-91-710-22-75Switzerland ■ Information Builders Switzerland AG Dietlikon 41-44-839-49-49United Kingdom* ■ Information Builders (UK) Ltd. London 44-845-658-8484
RepresentativesAustria ■ Raiffeisen Informatik Consulting GmbH Vienna 43-12-1136-3870Brazil ■ InfoBuild Brazil Ltda. São Paulo 55-11-3285-1050China ■ InfoBuild China, Inc. Shanghai 86-21-5080-5432 Beijing Xinrong Software Technology Co., Ltd. Beijing 86-10-5873-2031Denmark ■ InfoBuild AB Kista, SE 46-735-23-34-97Egypt ■ Al-Hisn Al-Waqi (AHAW) Riyadh, SA 996-1-4412664Ethiopia ■ MKTY IT Services Plc Addis Ababa 251-11-5501933Finland ■ InfoBuild Oy Vantaa 358-207-580-840Greece ■ Applied Science Athens 30-210-699-8225Guatemala ■ IDS de Centroamerica Guatemala City 502-2412-4212Gulf States ■ Al-Hisn Al-Waqi (AHAW)■ Bahrain ■ Kuwait ■ Oman ■ Qatar ■ United Arab Emirates ■ Yemen Riyadh, SA 996-1-4412664India* ■ InfoBuild India Chennai 91-44-42177082
Israel ■ SRL Group Ltd. Tel Aviv 972-3-7662030Italy ■ NessPRO Italy S.p.A. Genoa 39-010-64201-224 Milan 39-02-2515181 Turin 39-011-5513-211Japan ■ K.K. Ashisuto Osaka 81-6-6373-7113 Tokyo 81-3-5276-5863Jordan ■ Al-Hisn Al-Waqi (AHAW) Riyadh, SA 996-1-4412664Malaysia ■ Elite Software Technology Sdn Bhd Kuala Lumpur 60-3-21165682Norway ■ InfoBuild Norway Oslo 47-48-20-40-30Philippines ■ Beacon Frontline Solutions, Inc. Makati City 63-2-750-1972Poland/Central and Eastern Europe ■ InfoBuild SP.J. Warsaw 48-22-657-00-14Russian Federation ■ FOBOS Plus Co., Ltd. Moscow 7-495-926-3358Saudi Arabia ■ Al-Hisn Al-Waqi (AHAW) Riyadh 996-1-4412664Singapore ■ Automatic Identification Technology Ltd. Singapore 65-6286-2922South Africa ■ InfoBuild South Africa (Pty.) Ltd. Gauteng 27-83-4600800 Fujitsu Services (Pty.) Ltd. Johannesburg 27-11-2335911South Korea ■ Unitech Infocom Co. Ltd. Seoul 82-2-2026-3100 UVANSYS Seoul 82-2-832-0705Sweden ■ InfoBuild AB Kista 46-735-23-34-97Taiwan ■ Galaxy Software Services Taipei 886-2-2586-7890Thailand ■ Datapro Computer Systems Co. Ltd. Bangkok 662-679-1927, ext. 200Venezuela ■ InfoServices Consulting Caracas 58-212-763-1653
Toll-Free NumberSales, ISV, VAR, and SI Partner Information ■ (800) 969-4636
* Training facilities are located at these branches.