Hercules Emulator 3.07：系统安装指南（S/370, ESA/390 & z/Arch）

需积分: 9 169 浏览量更新于2024-08-01 收藏 1.51MB PDF 举报

HerculesInstallation文档详细介绍了Hercules系统，一种专为System/370、ESA/390和z/Architecture架构设计的模拟器。该安装指南适用于Hercules Release 3.07.0及后续版本，适用于不同软件级别的用户，强调了使用正确版本的重要性。文档结构包括序言、相关出版物、硬件和性能要求，以及Windows环境下的安装步骤。在序言部分，首先介绍了文档的版本信息，指出读者应根据其使用的软件级别选择相应的版本。该指南的目标读者是希望安装和使用Hercules模拟器的系统管理员、开发者和技术支持人员。文档假设读者已经具备一定的技术背景，以便理解书中内容。阅读和使用指南的方式也在此处给出，强调了可能的更新通知、读者反馈、法律声明和商标信息，以及对其他相关出版物如用户参考手册、错误消息与代码指南和快速参考的引用。第一部分着重于硬件和性能，其中详细说明了硬件配置要求，包括必不可少的PC硬件，并讨论了MIPS（每秒百万指令）和I/O速率等性能指标。还提供了Hercules模拟器的性能测量方法，帮助用户评估模拟器的实际效能。第二部分深入到Windows安装流程，列出了软件需求清单，包括操作系统兼容性、驱动程序、运行时环境（可能是特定的操作系统接口或兼容层），以及Hercules模拟器本身。此外，文档还指出了额外的必要和可选软件，以确保完整且顺利的安装过程。在整个安装过程中，读者可以根据文档提供的指导进行操作，注意检查并满足所有先决条件，以确保Hercules能在目标环境中稳定、高效地运行。由于篇幅较长，这部分内容还包括了组件组件安装的具体步骤和注意事项，以及可能遇到的问题解决策略。 HerculesInstallation指南为用户提供了一个详尽的安装和配置指南，旨在帮助用户充分利用这款强大的System/370、ESA/390和z/Architecture模拟器，实现兼容性和性能优化。

4. Performance

As described previously the performance of the Hercules Emulator depends heavily on the underlying PC

hardware. It is therefore not possible to give exact performance specifications of any particular operating

system running under Hercules. Some practical values from user experiences across several machines

presented here provide some guidelines though.

The performance of Hercules is measured in two values, MIPS and I/O rate. Both of these values are pre-

sented on the Hercules console and are refreshed every second independently of the PANRATE control

statement. See Hercules User Reference for details on the PANRATE statement.

4.1 MIPS

MIPS is an abbreviation of “Million Instructions Per Second” and presents a measure of the number of

instructions the CPU is executing in one second. MIPS is a measure of a computer's processor speed.

However this measure is useful only among processors with the same instruction set as different instruc-

tion sets take different numbers of instructions to do the same job. Many of the reported MIPS values

represent 'Peak' execution rates on artificial instruction sequences with few branches, whereas realistic

workloads consist of a mix of instructions some of which take longer to execute than others.

The performance of the memory hierarchy greatly affects processor performance, an issue also not con-

sidered in simple MIPS comparisons. In an attempt to address these issues researchers have created

standardized tests such as "SpecInt" to measure the real effective performance in commonly used

applications. The use of raw MIPS as a measure of overall system performance has fallen into disuse.

MIPS is sometimes pejoratively referred to as "Meaningless Indicator of Processor Speed" or "Meaning-

less Information Provided by Salespeople".

The Hercules console reports the MIPS rate for the emulated S/370, ESA/390 or z/Architecture instruc-

tions, not the underlying executed instructions of the hardware. As implied previously the MIPS rate can

vary significantly depending on the executed instruction and whether the instructions can be processed

entirely in cache, as can happen in a tight loop.

It is difficult to determine how the speed of the Hercules emulation corresponds to a real mainframe. This

is due to difficulties in comparing real mainframe hardware to PCs (or servers) as well as the actual per-

formance measurement itself. Hercules shows its processing speed in MIPS. Compared to the earlier

IBM System/360 and System/370 hardware, it is safe to say that Hercules will outperform them when it is

running on moderately powerful hardware, whereas newer IBM System/390 and z/Series hardware still is

much faster than the emulation.

On a Celeron 300 you should see an execution speed of 1 to 2 MIPS, which is enough to run OS/360

(MFT or MVT) or MVS 3.8 with a response time better than that of a 3033 from the 1970's. It's also fast

enough to run VSE/ESA with an acceptable response time. On a more recent system with a 2 GHz

Pentium processor, you may see the system peak at around 30 MIPS which is enough to run Linux/390

or z/OS with a light workload.

Performance on server class machines is now fairly respectable. For example, on a dual-core Intel Xeon

with "Hyperthreading" (4 CPUs) running at 3.46 GHz, you might expect to see a sustained MIPS rate of

40 to 60 MIPS. A dual-processor quad-core Mac Pro (8 cores, 3 GHz) will sustain over 150 MIPS. For

anyone who is prepared to spend a considerable amount of money on their Hercules system, there are

reports that a sustained 300+ MIPS rate has been achieved on an Intel Core i7 processor running at 3.75

GHz using all four cores plus "Hyperthreading" (8 CPUs).

The speed however depends greatly on the executed instructions. Instructions that are very expensive to

emulate can even on fast systems still be around 1 - 2 MIPS, see the tables further below for details.

Hercules Emulator - Installation Guide Page 16

4.2 I/O Rate

The second value that gives us performance data is the I/O rate. This is the average number of SIOs

(Start I/O per second) occurring to active devices, including DASD, TAPE, CTC (channel-to channel

adaptor), local SNA and non-SNA devices etc.

This value also varies depending on the underlying hardware. A RAID system will easily outperform a

non-RAID system. Many PC systems today can be ordered with a RAID-0 (Striping) or RAID-1 (Mirroring)

adapter. While disk mirroring (RAID-1) gives fault tolerance it does not improve performance. However

Data striping (RAID-0) spreads blocks of each file across multiple disks and can nearly double the per-

formance of a single disk.

On a recent PC system as described above, incorporating two Serial-ATA (SATA) disk drives connected

to a RAID adapter card with activated RAID-0 has shown peak I/O rates of more than 3250 SIOs per se-

cond. The sustained I/O rate delivered from such a system was more than 1500 SIOs per second.

4.3 Hercules Performance Measurements

The following sections show some performance index tables under various system configurations. Indi-

vidual measured performance may vary from the figures shown as performance depends heavily on the

hardware used and the Hercules release level.

All tests have been performed with the following measurement tools:

 IMON (Rate CPU Instruction Speed)

 CPU Instruction Timing Tool (INSTRATE program)

Tests showing the performance of Hercules itself include measurements of the Hercules Emulator soft-

ware and the difference between working with CKD or CCKD (compressed CKD) DASD emulation. Other

tests, like the host CPU on which Hercules runs or the disk types of the host system, show the influence

of the hardware used. The last test ("Hercules Overall Performance") represents the practical perfor-

mance growth of the Hercules Emulator as both hardware and software have improved.

4.3.1 Hercules Emulator Performance

The following diagram shows the software related performance index of the Hercules Emulator software.

The graphic shows the performance improvements that have been made over recent releases. The base

version (index = 100%) for all measurements was Hercules V3.00.

All these tests were performed using the same hardware configuration. The results therefore demonstrate

the influence of the improvements in the performance of the Hercules Emulator software only.

Hercules Emulator - Installation Guide Page 17

Figure 1: Hercules Emulator Performance

Although there have been massive performance improvements in the Hercules Emulator software in re-

cent releases, the current rate of improvement cannot be expected to continue. On one hand, the maxi-

mum performance improvements available via software improvements will be reached, on the other hand

the developers of Hercules have architectural frameworks that impose certain limitations, as explained

below.

Firstly the Hercules software should be portable and is therefore written entirely in "C". The use of assem-

bler to give additional performance is not exploited. Secondly the developers try to emulate real hardware

as exactly as possible in accordance with IBM's "Principles of Operations" documentation. This leads to

code that may be not optimal for the underlying hardware, but provides perfect emulation of real main-

frame hardware, avoiding problems when running mainframe operating systems.

The performance of the Hercules Emulator software depends heavily on the set of executed (mainframe)

instructions as detailed in below ("Emulated Instruction Performance").

4.3.2 Emulated CKD / CCKD DASD Performance

The type of the emulated DASD devices, either CKD or CCKD, has direct influence on performance. If

using CKD devices there is more data to be read from the disk and transferred to memory, but the data is

directly usable and no further processing needs to be done. Using CCKD (compressed CKD) devices the

amount of data to be read from the disk and transferred to memory is greatly reduced but once in memory

the data has to be uncompressed before it can be processed.

The performance characteristics of DASD types are irrelevant if host disk space limitations impose the

choice of using compressed CKD only.

On a fast machine with fast disks CKD is the best choice. On a fast CPU with slow disks however, CCKD

can give better overall performance. A fast processor can uncompress the data in less time than that

saved by transferring the data uncompressed. In the less common case of a slow CPU with fast disks

Hercules Emulator - Installation Guide Page 18

剩余137页未读，继续阅读

lzldnc

粉丝: 0
资源: 1

Hercules Emulator 3.07：系统安装指南（S/370, ESA/390 & z/Arch）

HerculesInstallation IBM模拟器

全球与中国柔性装配线自动导引车市场现状及未来发展趋势（2024版）.docx

数据手册-STC51RC-RD简介-datasheet.zip

基于MATLAB汽车框定系统GUI面板.zip

基于ASP+SQL SERVER 实现的红旗汽车修理厂物资流通管理系统毕业设计（论文+源代码）

技术资料分享TL1838电子技术资料.zip

大学生软件测试2024（练习赛）12306测试成果截图

施耐德控制专家 Control Expert DFB Password 解密工具 说明 c#源码

基于树莓派的安保巡逻机器人-（一、快速人脸录入与精准人脸识别）

王涛-EULITHA_CIOE Chengdu_share version.pdf

最新资源

施耐德控制专家 Control Expert DFB Password 解密工具说明 c#源码