IBM System Z上设置Linux生产环境指南

需积分: 10 139 浏览量更新于2024-07-20 收藏 3.96MB PDF 举报

"Setup Linux on System Z for Production是IBM红皮书中的一本，主要面向服务器和Linux工程师，详细介绍了如何在IBM的大型主机System Z上部署Linux系统用于生产环境。内容涵盖从测试和开发到生产环境迁移的考虑因素、容量规划、性能优化，以及样例检查列表等。此书适用于z/VM Version 6, Release 2以及Red Hat Enterprise Linux 6.1和SUSE Linux Enterprise Server 11 SP1版本。" 在IBM的System Z上设置Linux操作系统对于生产环境至关重要，因为它涉及到系统的稳定性和效率。以下是对该资源中涉及的知识点的详细说明： 1. **系统迁移**：从测试和开发环境迁移到生产环境时，需要考虑诸多因素，包括但不限于安全性、性能、可用性、数据迁移和兼容性。这些因素都需要在迁移前进行详尽的评估和规划。 2. **容量规划**：在System Z上部署Linux时，需要进行详细的容量规划，这包括对CPU、内存、存储和网络带宽的需求预测，以确保系统在高负载下仍能正常运行。同时，容量规划还涉及到未来扩展性的考量。 3. **性能优化**：针对System Z的特性，书中可能提供了针对Linux性能调优的策略，如内核参数调整、I/O管理和资源调度等，以确保系统运行高效。 4. **样例检查列表**：这些列表通常包含了一系列步骤和检查点，帮助工程师在部署和配置过程中避免常见错误，确保过程的标准化和一致性。 5. **IBM z/VM**：IBM的z/VM是一个虚拟化平台，能够支持多个Linux实例在同一硬件上并发运行。了解其工作原理和管理方式是成功部署的关键。 6. **Red Hat Enterprise Linux 和 SUSE Linux Enterprise Server**：作为流行的Linux发行版，它们在System Z上的兼容性和性能表现是关注的重点。书中可能会提供这两个发行版在System Z上的最佳实践。 7. **安全与合规**：在政府用户方面，根据GSA ADP Schedule Contract with IBM Corp，系统可能需要满足特定的安全和使用限制。 8. **持续学习与反馈**：IBM Redbooks鼓励读者参与评论和分享经验，同时也提供了保持与IBM最新技术资讯同步的方式。 "Setup Linux on System Z for Production"是一本全面的指南，对于那些希望在IBM System Z上实现Linux稳定运行的工程师来说，是不可多得的参考资料。通过深入理解和应用书中的知识，可以有效地提升System Z上的Linux生产环境的性能和可靠性。

2 Set up Linux on IBM System z for Production

1.1 Hardware

There are a number of reasons to use Linux on System z. System z hardware was originally

delivered in the early 1960s and has evolved over the decades. In 2000, Linux on System z

became a commercial reality and development investments were made to improve the Linux

usability on the System z hardware. Resources can be shared among multiple Linux images

running on the same z/VM system. Resource improvements include improvements to the

CPU clock speed, size of the memory allocated per logical partition (LPAR), and the use of

Fibre Channel Protocol (FCP) with IBM FICON® devices. For example, where once the

System z 990 had 64 GB of memory, the zEC12 now has 3 TB of memory.

Additionally, there has been much investment in the IBM z/VM virtualization operating system

that benefits Linux on System z guests. An example of this would be the faster speed of the

Integrated Facility for Linux (IFL) on the new IBM zEnterprise® EC12 (zEC12), which enables

you to host more virtual servers per processor core than other server platforms. Combined

with the up to 101 user-configurable IFLs, the total Linux processing capacity has enormously

increased with the zEC12.

Consolidation of server hardware resources allows for the running of tens or hundreds of

Linux instances on a single System z server, offering savings in space and personnel that are

required to manage real hardware.

1.2 Reliability, availability, and serviceability

Reliability, availability, and serviceability (RAS) is a concept that describes computers that can

work for a long time without a single stop. The IBM System z is designed with redundancy of

components to provide RAS. Batteries, power connections, cooling systems, processors, as

well as memory, have been developed with full redundancy as the objective so that any

device failure does not become an unplanned outage.

In 2012, the IBM System zEC12 machine was released and designed with a number of

processor cores that can reach up to 101 active specialized processors that are called

Integrated Facilities for Linux (IFL). There are 120 active cores in the zEC12. The reason for

this is that there are up to 16 cores that are used as system assist processors (SAPs). Two

cores are always used as spare cores and one core is for reserved use. The two spare cores

can assume all software requests without stopping the application if one of the active

specialized cores fails. Also, physical RAM has an algorithm that prevents the applications

from stopping due to a failure of one single RAM.

Some other features that are related to high availability characteristics are Enhanced Book

Availability (EBA), concurrent memory replacement using Licensed Internal Code

Configuration Control (LICCC), redundant I/O interconnection, and more. For information

about System z processors, memory, and other devices, see the IBM zEnterprise EC12

Technical Guide, SG24-8049.

1.3 System z virtualization

The virtual machine environment is highly flexible and adaptable. New Linux guests can be

added to a z/VM system quickly and easily without requiring dedicated resources.

Chapter 1. Introduction 3

The System z hardware has two levels of virtualization. The first level is managed by the IBM

Processor Resource/Systems Manager™ (PR/SM™) microcode. The PR/SM provides the

ability to split single system hardware into up to 60 LPARs that are able to share the

processors and I/O channel devices including storage devices and network device access. In

an LPAR, it is possible to install and run a z/VM hypervisor or a Linux operating system as

well as IBM operating systems, such as z/OS, z/TPF, or IBM z/VSE®.

By using the System z LPAR configuration, it is possible to increase memory, processor, and

channels dynamically. These characteristics prevent a running application from stopping

because of a possible capacity problem. Because RAM memory is the only resource at the

PR/SM level that is not shared, it is recommended in a production System z environment that

all memory area addresses not be allocated to all LPARS while the logical partition is defined.

It is better to reserve some memory to increase an LPAR dynamically, if necessary.

Because the PR/SM is able to control and define the weight of priority for each LPAR defined

on the system, the system administrator should separate the development environment from

the production environment, whether sharing resources or not, but always prioritizing the

order of execution. The definition of the LPAR weight can be in a relative basis. Suppose that

you have 10 IFL System z machines and you shared those 10 IFLs between two LPARs,

defined as

LPAR A and LPAR B. LPAR A executes the production servers and LPAR B

executes the development or even quality assurance (QA) servers, as described in

Figure 1-1.

Figure 1-1 Production and development environments

In a shared processor environment, if you define the number of 80 to LPAR A and 20 to LPAR

B, it means that LPAR A executes four times more instructions than LPAR B when all systems

are running at 100%. This feature allows you to prepare your system and ensure that the

development server will not interfere with the production servers. System z was designed to

4 Set up Linux on IBM System z for Production

run at full capacity, so if there were free resources in the system and the development LPAR

was requesting more processing, it would be able to use any free IFL resources.

It is also possible to mix dedicated IFLs with shared IFLs in LPARs on the same machine for

highly critical system configurations. In Figure 1-2, we show two production environments and

one development LPAR environment. The production environment, LPAR A, has a dedicated

IFL. LPAR B is also a production LPAR but it is sharing an IFL with development LPAR C.

Figure 1-2 Two production LPARs and one development LPAR

This example can be used to allocate Linux on System z database guests in a high availability

solution where the active nodes are running in LPAR A and the passive production servers in

LPAR B. In this case, all highly critical with high utilization system guests in LPAR A have the

benefit of the reduced overhead that exists when the PR/SM must balance the weight and

guarantee that all LPARs have their own CPU running time.

Another situation that can benefit from this kind of solution is to have very large database

servers running directly on the LPAR. This includes databases that need to be scaled in

numbers not possible in other hardware architectures such as x86 or x64 servers. In this

case, all hardware cycles are dedicated to the Linux guests and the database application, and

also still have the benefits of increasing or decreasing memory and IFLs as needed by

application peak load times.

Second-level virtualization is equally efficient and includes features such as sharing and

overcommitment of memory allocation as well as the number of virtual processors allocated.

The second level of virtualization is managed by the z/VM operating system. z/VM provides a

fully virtualized System z environment to run Linux, z/OS, z/TPF, z/VSE, or another z/VM as a

guest system.

One single z/VM LPAR can run hundreds to thousands of Linux guests. But in a production

environment, it is necessary to understand the peak load times of each server and the

utilization of the resources during those periods. You do not want all of your servers to have

6 Set up Linux on IBM System z for Production

In the processor utilization chart for LPAR A (Figure 1-4), we describe the utilization of z/VM

production LPAR A and how it works on a busy day.

Figure 1-4 Processor utilization chart for LPAR A

This chart demonstrates how well System z works and how well it manages virtualization. For

peak time, indicated by a red circle and labeled “Peak Load”, about 20 minutes elapsed and

the LPAR utilizations ran at almost 100%. The real number was 91 - 97% of the 16 real IFL

processors. The LNX A guest used 73% - 100% of its 11 virtual processors and LNX B used

83 - 99% of its nine virtual processors. Although the total number of virtual processors is 28,

z/VM and PR/SM are able to scale all requests in 16 real IFLs, but only for a short time. If that

interval is extended for a longer period, the Linux guest might have slow response time. Keep

in mind, the system will not stop, but the response time is slower.

System peak load time is an important consideration that you should look at when you are

planning your production environment. For more information about processor considerations

and z/VM scheduling, see Chapter 10, “Performance considerations” on page 123.

Memory allocation is another important topic that is related to virtualization. It is important to

understand that on System z, the memory requirement is analyzed in a different way. There

are several factors such as z/VM memory allocation and z/VM system paging, as well as the

ability to define a certain amount of z/VM memory as virtual disks and use that on the Linux

guest as a swap disk. The high performance I/O system of System z hardware allows you to

use more of the hardware and decrease the need to cache files and databases systems and

should be considered during the development phase of your production environment. It is

also possible to change the size of available Linux guest memory dynamically, which provides

a dynamic scalability for the applications and systems. More about memory considerations

can be found in Chapter 3, “Hardware planning considerations” on page 29.

If you do good planning, setup, and monitoring of your Linux guests and z/VM logical

partitions, considering the peak load utilization aspects, I/O channel utilization, and memory

definitions, it is possible to allocate twice the size of real memory of logical partitions as virtual

memory if you consider the sum of all of the Linux guests’ memory in the LPAR.

剩余189页未读，继续阅读

michael_yzq123

粉丝: 3
资源: 75

IBM System Z上设置Linux生产环境指南

astrill-setup-linux64-20210901.deb

Ecology-setup-forLinux-v2.66-forPythony3.py修改版（支持部署在ubuntu）

IBM Linux红皮书英文资料集

VMware Lab setup - A virtualized lab for testing HA and DRS

running setup.py develop for opencv-contrib-python

Running setup.py install for fasttext did not run successfully.

linux命令setup

Unable to create a virtual device: Missing system image required for an AVD setup

Running setup.py bdist_wheel for opencv-contrib-python ...

最新资源