Browse Prior Art Database

A Method for Determining the Reliability of a RIO Link Before Configuration Through the Use of Ping Tests

IP.com Disclosure Number: IPCOM000022406D
Original Publication Date: 2004-Mar-12
Included in the Prior Art Database: 2004-Mar-12
Document File: 2 page(s) / 39K

Publishing Venue

IBM

Abstract

When configuring or hotplugging a RIO (Remote I/O) cable to a server system, there exists a possibility that the server may crash due to a faulty cable or a previously working cable left in a bad state. The current algorithm relies on an input ready signal in the RIO macro which does not necessarily insure that the link is good. The only way to determine if a cable is faulty is to let the cable failover to the redundant port if End-to-End is enabled or to issue a RIO request with the risk that the system may crash due to system timeouts. Disclosed is a way to keep faulty cables from bringing the system down by performing a series of ping tests prior to configuring the cable. This may also be used as a quick runtime test unit.

This text was extracted from a PDF file.
At least one non-text object (such as an image or picture) has been suppressed.
This is the abbreviated version, containing approximately 52% of the total text.

Page 1 of 2

A Method for Determining the Reliability of a RIO Link Before Configuration Through the Use of Ping Tests

High system availability is one of the main goals of server design. Preventing crashes or fails on a running system will avoid down-time for repairs on a production system. When configuring or hotplugging a RIO (Remote I/O) cable to a server, there exists a possibility that the server may crash due to a faulty cable or a previously working cable left in a bad state. The current algorithm relies on an input ready signal in the RIO macro which does not necessarily insure that the link is good. At present time the only way to determine if a cable is faulty is to let the cable failover to the redundant port if End-to-End is enabled on the link or to issue a RIO request with the risk that the system may crash due to system timeouts. The intention of this disclosure is to describe away to keep faulty cables from bringing the server down by performing a series of ping tests prior to configuring the cable. This could also be used as a quick runtime test unit for diagnostic purposes. A method for implementing this will be presented below.

There have been previous scenarios in which RIO links have been enabled, even though there was still a faulty link due to a bad port on the RIO hub chip. The cable hotplug operation was successful, however the server crashed as soon as traffic was sent through the link. A ping test would give more reliability to replacement of cables in the field. The ping test could also be included in diagnostics test for customer engineers or customers in the field to run after a cable replacement. In the future this test could also become a part of RIO hub hotplug operations.

Before configuring a RIO cable either during initial IPL or during a runtime hotplug operation, implement the RIO macros' RIO performance monitor to keep statistics on link acknowledges (acks) sent/received along with CRC errors received. Firmware will enable the RIO performance counters in the originating outbound RIO port. In particular, enabling t...