Amazon Technologies, Inc.

Filippo Sironi

Jan Hendrik Schoenherr

Karimallah Ahmed Mohammed Raslan

Anthony Nicholas Liguori

Konrad Jan Miller

Amit Shah

Ioannis Aslanidis

Marius Hillenbrand

Benjamin Charles Serebrin

Robert Charles Swanson

Martin Thomas Pohlack

Said Bshara

Uwe Dannowski

Evgeny Schmeilin

Sebastian Martin Biemueller

Peter Barry

Tsung Ho Wu

Rolf Peter Neugebauer

Pankaj Kumar Rai

Jan Hendrik Schönherr

Nandagopal Sathyanarayanan

Maulik Kapuria

Yuhui Zheng

Sang Phill Park

Milan Pandurov

Barak Nirenberg

An input/output (I/O) device is coupled to a processor via a first peripheral bus and a second peripheral bus. The I/O device is configured to expose a first physical function (PF) and a second PF to a virtual machine (VM) executing on the processor. A first virtual function (VF) associated with the first PF is assigned to the first peripheral bus, and a second VF associated with the second PF is assigned to the second peripheral bus. The first VF and the second VF are exposed as a single VF to the VM. The VM can perform transactions with the I/O device via the single VF using the first peripheral bus and the second peripheral bus.

Bonding multiple peripheral ports to a single virtual machine

An isolated accelerator management intermediary running at a virtualization host uses a management interface exposed by a hardware accelerator to perform a set of operations to enable access to the hardware accelerator from a guest virtual machine. The operations include configuring a guest-accessible interface of the accelerator. A virtualization management component of the host launches the guest virtual machine after the set of operations is completed.

Isolated accelerator management intermediaries for virtualization hosts

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

A first set of instructions, which is provided access to a first address space, is scheduled for execution at a first hardware thread of a processor. Prior to executing an instruction of a second set of instructions, which accesses a second address space, at the first hardware thread, a determination is made that the second address space is accessible from a second hardware thread of the processor.

Security vulnerability mitigation using address space co-execution

A computing device including executable processes may determine that a future likelihood of access for virtual memory pages of an executable process are below a threshold likelihood of access based on an execution status of the executable process or a tracking of memory accesses to the virtual memory pages of the executable process. Responsive to this determination, memory pages found to store contents matching that of memory pages mapped to other processes may be unmapped from the process and released for reuse by the computing device. The virtual memory pages may then be marked as being shared with the similar memory pages mapped to the other processes. At a later time, the memory pages of the process may be configured to be non-shared, the configuring including either copying respective shared pages to non-shared pages or enabling a processor exception on access to the memory pages.

Safe sharing of hot and cold memory pages

As part of a compute instance migration, a compute instance which was executing at a first server begins execution at a second server before at least some state information of the compute instance has reached the second server. In response to a determination that a particular page of state information is not present at the second server, a migration manager running at one or more offload cards of the second server causes the particular page to be transferred to the second server via a network channel set up between the offload cards of both servers, and stores the page into main memory of the second server.

Accelerated migration of compute instances using offload cards

An opportunistic hypervisor determines that a guest virtual machine of a virtualization host has voluntarily released control of a physical processor. The hypervisor uses the released processor to identify and initiate a virtualization management task which has not been completed. In response to determining that at least a portion of the task has been performed, the hypervisor enters a quiescent state, releasing the physical processor to enable resumption of the guest virtual machine.

Virtual machine management using partially offloaded virtualization managers

An interrupt recovery manager detects that a compute instance has exited a state in which interrupts from one or more interrupt sources were not processed at the compute instance. The recovery manager identifies a source from which interrupts may have been missed by the compute instance, and causes an indication of a potential missed interrupt from that source to be delivered to the compute instance.

Interrupt recovery management

An offloaded virtualization management component of a virtualization host receives an indication from a hypervisor of a portion of main memory of the host for which memory allocation decisions are not to be performed by the hypervisor. The offloaded virtualization management component assigns a subset of the portion to a particular guest virtual machine and provides an indication of the subset to the hypervisor.

Memory allocation techniques at partially-offloaded virtualization managers

Performance variability reduction using an opportunistic hypervisor

Growing community of inventors

Dresden, Germany

Filippo Sironi