Home
Application Note AN-572 Upstream Port Redundancy and
Contents
1. fundamental reset of the system while dynamic failover does not Details on each of these failover methods are provided in the next two sections Static Failover Static failover is essentially a swapping of the active and standby upstream ports during a fundamental reset of the system During fundamental reset the SWMODE vector is changed to select the other upstream port To be in a mode conducive to static failover the switch can be operating under SWMODE vectors 0x8 0x9 OxA or 0xB To perform a static failover a processor usually the standby processor except in the case of a planned switchover would assert fundamental reset PERSTN to the switch toggle to the complementary upstream port using the appropriate SWMODE vector then deassert reset Because Static failover requires a fundamental reset of the system the dynamic failover method might be preferred since it does not require a fundamental reset of the switch 3 Other modes could be relevant in certain cases but are of limited use as they do not support toggling between upstream ports re failover 3of4 May 24 2007 IDT Application Note AN 572 Notes Dynamic Failover Dynamic failover makes use of the SERDES multiplexing mentioned previously as well as three triggers which can initiate the switching of the SERDES to the other upstream port This operation requires a hot reset during which all of the switch ports are retrained A configuration option exists
2. whereby the down stream ports states are preserved during hot reset If this option is enabled the new upstream port will be trained but the downstream ports and core switch logic will not be reset This feature is disabled by default and can be enabled in the Switch Control SWCTL register As with static switchover upstream port redundancy must be selected during fundamental reset It can be enabled as part of the boot mode configuration by selecting an appropriate SWMODE vector with redun dant upstream ports Either port 0 or port 2 can be configured to be the upstream port By implication the other port would become the standby port Once redundancy is enabled dynamic failover is also enabled This is reflected in the Upstream Port Failover Status USPFSTS register Dynamic Failover Triggers The failover operation is triggered in one of three ways Software Initiated Failover An upstream port failover can be initiated via the Upstream Port Failover Control USPFCTL register A software initiated failover can be sourced by the upstream port as in the case of a planned switchover or the slave SMBus Start with the upstream port value in the USPFCTL register corresponding to the active upstream port To trigger failover write the USPFCTL register with the new upstream port value Hardware Pin Initiated Failover An upstream port failover may be initiated by a change in the state of the Upstream Port Select USPSEL signal pi
3. Upstream Port Redundancy and Application Note i Failover for System Interconnect AN 572 4 Switches By Sean Sweeney Notes Introduction Fault tolerance of the system processor is a critical requirement in embedded and storage systems like the one shown in Figure 1 In this example if the active processor blade were to fail then the standby card would take its place This is called failover Failover enables fault tolerance which can increase overall system RAS reliability availability and serviceability Standby Processor Proc sso r Blade Blade Switch ZIET PES64H16 Figure 1 A Fault Tolerant System with Active and Standby Processors It is desirable and in fact common for failover to happen automatically via software without the inter vention of a technician A popular way to achieve this kind of fault tolerance is to use two upstream ports in the switch One upstream port is connected to the active processor blade while the other is connected to the standby processor blade During failover the standby processor becomes the active root complex and takes over the control and management of the PCI Express PCle switch and the 1 0 blades In order to support failover a switch must implement two features redundant upstream ports and a failover mechanism Note that non transparent bridging NTB functionality is not required for a switch to Support failover The IDT System Interconnect PCle Switches a
4. e AN 572 Notes Standby Processor Processor Blade Control Plane Blade i SMBus amp Switch Contral Signals i Switch Blade PES64H16 Figure 4 Control Topology Typically a heartbeat message Is sent periodically from the active processor to the standby processor in order to continuously verify that the active processor is functioning Due to Its ubiquity ease of implementa tion and the fact that it comes for free on root complex chipsets Ethernet is a natural choice to carry the heartbeat message Should the heartbeat message not be received by the standby processor within a predetermined time frame the active processor would be considered down and the standby processor would commence the failover operation There are signals which are important to failover operation that must connect each processor blade to the switch blade They are A fundamental switch reset PERSTN signal required for static failover only Four switch mode SWMODE signals required for static failover only Two Slave SMBus SSMB signals required for dynamic software or timer failover only An upstream port select USPSEL signal required for dynamic hardware failover only These signals will be discussed further in subsequent sections Failover Methods Static and Dynamic Once redundancy is enabled failover is also supported The system designer can choose between two failover methods static and dynamic Static failover involves a
5. n This failover mode must be enabled in the USPFCTL register and the upstream port selected by the USPSEL signal must differ from the current upstream port The USPSEL signal should initially correspond to the active upstream port Enable hardware based failover in the USPFCTL register To initiate failover change the state of the USPSEL signal Watchdog Timer Initiated Failover An upstream port failover may be initiated as the result of an expiration of a watchdog timer Such a failover must be enabled in the USPFCTL register If software does not reset the timer thereby allowing the Watchdog Timer Count COUNT field in the Upstream Port Failover Watchdog Timer USPFTIMER register to transition from a one to a zero then failover will occur The timer can be written via the slave SMBus or the upstream port The upstream port value in the USPFCTL register should initially correspond to the active upstream port Enable timer based failover in the USPFCTL register and set the timer in the USPFTIMER register to the desired value Failover is initiated if software does not reset the timer before it reaches zero After a failover has occurred the switch will be in a state wherein another failover can be performed Simply supply another trigger to perform another failover Summary The IDT System Interconnect PCI Express Switches are ideal for building fault tolerant high RAS embedded and storage systems Their key advantages are the ability to su
6. pport failover the highest port densities in the industry and wire speed simultaneous switching between any and all ports References PES64H16 User Manual Revision History May 24 2007 Initial publication 4of4 May 24 2007
7. re useful in building fault tolerant systems because they satisfy both of the requirements In addition to being fault tolerant these systems can achieve high port densities up to 16 ports with full peer to peer simultaneous switching between all ports at wire Speed This document describes the redundancy and failover features of the IDT System Interconnect PCI Express Switches and also discusses how these switches can be used to build fault tolerant systems Upstream Port Redundancy The redundancy feature can be enabled during fundamental reset Pin strapping of the SWMODE Signals is used to turn on redundancy and select which upstream port is active and which is standby This section provides an overview of the PES64H16 ports with respect to number and widths and also describes the upstream redundancy implementation l While this application note refers to the PES64H16 device the information herein applies to any device in the System Interconnect Switch family 1 of 4 May 24 2007 2007 Integrated Device Technology Inc IDT Application Note AN 572 Notes The PES64H16 has sixteen physical x4 ports In a 16x4 configuration there are also sixteen logical ports 0 through 15 Even odd and not odd even x4 ports can be merged to form a x8 port In such a case the two merged physical ports take on the even port number This rule is valid for all ports including the upstream port s Port 0 is always an upstream port regardless of por
8. t width Port 2 is an upstream port if redundancy is enabled also regardless of port width Figure 2 illustrates a switch configuration using redundant x4 upstream ports 0 and 2 Note that ports 1 and 3 are downstream ports This configuration has sixteen logical x4 ports two redundant upstream ports and fourteen downstream ports xd xd wd xd wd wd xd x4 xd wd wd xd wd xd Downstream Ports Figure 2 Upstream x4 Ports Figure 3 illustrates a switch configuration using redundant x8 upstream ports 0 and 2 There are four teen logical ports two redundant upstream ports and twelve downstream ports x4 x4 wd xd xd xd xd xd xd xd xd xd Downstream Ports Figure 3 Upstream x8 Ports The SERDES for ports 0 and 2 are connected via a multiplexer This is an important feature used in dynamic failover mode wherein certain triggers can cause the multiplexer to switch This causes the Standby port to become active that is to fail over If the redundancy feature is not enabled then the SERDES multiplexer is not active and the dynamic failover feature is not available Note that this is not a hot standby scheme the standby processor port must undergo link training before it can become active System Considerations A typical system control topology for implementing failover with the PES64H16 Is shown in Figure 4 2 Dynamic failover is explained in subsequent sections of this application note 2 of 4 May 24 2007 IDT Application Not
Download Pdf Manuals
Related Search
Related Contents
Aficionado V2.04 リリースノート - 160.17 KB Operators' Manual Electronic Theodolite FET 420K Sun Microsystems 20 Webcam User Manual FCTVA - Mode d`emploi Concepts User Manual.. - Guillevin Industrial Automation Group Copyright © All rights reserved.
Failed to retrieve file