Amazon Technologies, Inc.

Tsung Ho Wu

Jiming Sun

Robert Charles Swanson

Filippo Sironi

Pankaj Kumar Rai

Bradshaw Darrow Austin

Yuhui Zheng

Sang Phill Park

Nandagopal Sathyanarayanan

Maulik Kapuria

A host server computer with an uncorrectable memory error can be repaired without a reboot operation. While initially booting a hypervisor, a special software Application Programming Interface (API) can be loaded between a BIOS System Management Mode (SMM) code and the hypervisor. Once the host server computer is booted and a number of virtual machines are executing, a memory error (e.g., uncorrectable error correction code (UECC)) can occur. In response, the hypervisor calls into the special software API identifying the defective memory rows that the BIOS needs to repair. The BIOS starts a soft Post Package Repair (PPR) process on those rows and gives back control to the hypervisor. When the repair is completed, the hypervisor loads a scrubbing virtual machine and validates that the memory is corrected. After the repair is validated, the hypervisor allows the available partition to take a new customer instance.

Runtime memory repair without requiring a reboot of a server computer

An update process for updating components of one or more host devices includes selecting an update path for updating a system component of the one or more host devices and generating an update manifest including a plurality of update payloads for the system component that are included in the update path. The update path may be selected and/or derived based on a multi-stage combination of a hash of an update payload corresponding to a current version of the system component and respective hashes of the one or more update payloads using a reversible function to maintain backward and forward trackability and compatibility for the updates. A plurality of possible update paths may be represented in a directed graph that is managed by a control system. The update payloads may be uploaded to the host devices to cause the host devices to update the system component based on the update manifest.

Upgrading system components with forward and backward compatibility

Various embodiments provide for the live updating of system software or firmware, such as BIOS information, without the need for a reboot. A payload containing BIOS update information can be generated and signed with an encryption key. The information in the payload is able to be updated through a live update, and can be selected in part based on the configuration and BIOS version of the computing device to be updated. Upon an update being triggered, a target device is placed into a management mode wherein normal work is suspended and the device is isolated. The payload is verified and decrypted, and appropriate functions called in order to update the relevant BIOS information. After the update is verified, the device exits management mode and resumes a typical workflow. If the update is not successful, or unable to be performed via live update, a conventional update can be performed that includes a reboot.

Live system updates

Growing community of inventors

Pleasanton, CA, United States of America

Tsung Ho Wu