Browse Prior Art Database

Method For Synchronizing Start of Multi-node Systems Operation

IP.com Disclosure Number: IPCOM000011755D
Original Publication Date: 2003-Mar-12
Included in the Prior Art Database: 2003-Mar-12
Document File: 1 page(s) / 40K

Publishing Venue

IBM

Abstract

In a multi-node system environment where each node has different attachments and resources, rebooting into synchronous multi-node operation may often not be successful. This is due to the differences in delays of various attachments initialization, diagnostics, and application loading. For example, in a multi-node system, one system has RAID storage devices attached while others typically do not. Starting the RAID controller and its attached devices, say 50 attached drives, can take more than 30 minutes because typically one to two devices are spun up in a staged and predetermined sequence. When such a condition occurs, the system having been delayed for 30 minutes after other nodes have already been brought up, jeopardizes the start up of synchronous multi-node operation and as such, often causes synchronous multi-node operation to fail to start properly. This invention discloses a method that will prevent or avoid such a condition from occurring and at the same time maximizes the efficiency in multi-node bring up time.

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 65% of the total text.

Page 1 of 1

Method For Synchronizing Start of Multi-node Systems Operation

         Disclosed is a method and system that utilizes resident service processors (SP) and its out-o- band maintenance network to manage the optimized bring up of all nodes in order to achieve successful initialization of synchronous multi-node operation. The SP on each physical node communicates to the other SP's when the system is first powered-on. Each SP reports its system's initialization status to the designated primary service processor on the network. When all SPs have reported that they are ready to the primary SP, the primary SP will instruct each SP to send a signal to its managed system to start multi-node operation. By doing so, the systems will be reliably brought up in synchronous multi-node operation in the shortest amount of time during which all systems' attachment and resources will be fully operational.

      The primary SP will have been provided configuration information on the synchronous multi-node staging requirements. Certain nodes will be designated at critical to operation of the multi-node system. When the critical nodes become available, the primary SP can instruct the critical nodes to start establishing multi-node operation prior to the rest of the nodes becoming available, for example, if any are not engaged in multi-node initialization or not completely booted. The configuration information can also stage the start up of multi-node operation based upon a priority scheme. It is als...