X hits on this document

PDF document

February 10, 2006 - page 16 / 238

844 views

0 shares

0 downloads

0 comments

16 / 238

4

CHAPTER 1. MANAGEMENT, SYSTEMS, AND IPMI

current BIOS state, whether a device is present or not, or other things like that.

FRU data is often stored in I2C EEPROMs on the I2C bus. FRU data is information about a Field Replacable Unit. This includes things like the manufacturer, the serial number, date of manufacture, etc. A system generally has information about the chassis and information about each field replaceable unit it has. Field replaceable units may include power supplies, DIMMs (memory devices), plug-in-boards, or a host of other things.

Connections to other MCs may be done through an IPMB. On an IPMB, each MC is a peer and they communicate directly through messages.

In addition to IPMB, IPMI systems can be interconnected through an Intelligent Chassis Management Bus. This is a serial bus that runs between chassis.

A management controller may be able to control various aspects of the chassis, such as power and reset. It may also have a watchdog timer for the main processor.

The Sensor Device Record (SDR) repositories store information about the sensors and components of the system. The BMC must have a main SDR repository; this repository is writable. There may only be one main SDR repository in the system. Any MC may have a device SDR repository; these are a read-only repositories.

When a problem or change in the system is detected, the MC handling that sensor may issue an event. This allows management software to detect these problems or changes without having to poll every sensor constantly. The events are stored in an event log. Events may be forwarded through the system interface or other interfaces, but they are always stored in the event log. The BMC must have an event log; generally the other management controllers forward their events to the BMC.

1.2

System Types

Although any arbitrary type of system may use IPMI for platform management, systems generally fall into two categories: server systems and bus systems.

Figure 1.3 shows a typical server system. It is a single stand-alone box that is a single computer. It has a BMC that is the main management controller in the system. It controls a number of sensors. In this example, the power supply also has a MC with it’s own sensors.

A BMC can have several connections to managing systems. It may have a system interface connection to the main processor. It may share an interface to the ethernet chip so the system may be managed through the LAN when the main processor is not working. Systems can have serial port connections. They can even have connections to modems where they can dial up a management system or page an operator when they detect a problem, or be dialed into by a management system.

Figure 1.4 shows a typical bus system. The word “bus” is perhaps a bit misleading; these types of systems used to have busses (like CPCI and VME) but recently have tended to not have big busses and use networking for interconnect (like PICMG 2.16 and ATCA). These systems generally contain a number of processors on pluggable boards often called Single Board Computers (SBCs) or blades. One or more power supplies power the whole system. The boards and power supplies can be hot-pluggable.

These systems generally have one or two boards that manage the system; this can be on a standard SBC, on another special purpose blade (like a blade used as a network switch), or on a standalone board with this purpose. The shelf management controller(s) generally act as the BMC in the system; they will have the event log and the main SDRs in the system. A system with two shelf controllers will generally allow the system to be managed even if one of the shelf controllers fails.

Document info
Document views844
Page views844
Page last viewedSat Dec 10 04:34:37 UTC 2016
Pages238
Paragraphs8090
Words85809

Comments