Abstract: Techniques for efficiently purging non-active blocks in an NVM region of an NVM device using virtblocks are provided. In one set of embodiments, a host system can maintain, in the NVM device, a pointer entry (i.e., virtblock entry) for each allocated data block of the NVM region, where page table entries of the NVM region that refer to the allocated data block include pointers to the pointer entry, and where the pointer entry includes a pointer to the allocated data block. The host system can further determine that a subset of the allocated data blocks of the NVM region are non-active blocks and can purge the non-active blocks from the NVM device to a mass storage device, where the purging comprises updating the pointer entry for each non-active block to point to a storage location of the non-active block on the mass storage device.

Abstract: Techniques for implementing RDMA-based recovery of dirty data in remote memory are provided. In one set of embodiments, upon occurrence of a failure at a first (i.e., source) host system, a second (i.e., failover) host system can allocate a new memory region corresponding to a memory region of the source host system and retrieve a baseline copy of the memory region from a storage backend shared by the source and failover host systems. The failover host system can further populate the new memory region with the baseline copy and retrieve one or more dirty page lists for the memory region from the source host system via RDMA, where the one or more dirty page lists identify memory pages in the memory region that include data updates not present in the baseline copy. For each memory page identified in the one or more dirty page lists, the failover host system can then copy the content of that memory page from the memory region of the source host system to the new memory region via RDMA.

Abstract: Techniques for implementing high availability for persistent memory are provided. In one embodiment, a first computer system can detect an alternating current (AC) power loss/cycle event and, in response to the event, can save data in a persistent memory of the first computer system to a memory or storage device that is remote from the first computer system and is accessible by a second computer system. The first computer system can then generate a signal for the second computer system subsequently to initiating or completing the save process, thereby allowing the second computer system to restore the saved data from the memory or storage device into its own persistent memory.

Abstract: Techniques for implementing RDMA-based recovery of dirty data in remote memory are provided. In one set of embodiments, upon occurrence of a failure at a first (i.e., source) host system, a second (i.e., failover) host system can allocate a new memory region corresponding to a memory region of the source host system and retrieve a baseline copy of the memory region from a storage backend shared by the source and failover host systems. The failover host system can further populate the new memory region with the baseline copy and retrieve one or more dirty page lists for the memory region from the source host system via RDMA, where the one or more dirty page lists identify memory pages in the memory region that include data updates not present in the baseline copy. For each memory page identified in the one or more dirty page lists, the failover host system can then copy the content of that memory page from the memory region of the source host system to the new memory region via RDMA.

Abstract: Techniques for achieving application high availability via application-transparent battery-backed replication of persistent data are provided. In one set of embodiments, a computer system can detect a failure that causes an application of the computer system to stop running. In response to detecting the failure, the computer system can copy persistent data written by the application and maintained locally at the computer system to one or more remote destinations, where the copying is performed in a manner that is transparent to the application and while the computer system runs on battery power. The application can then be restarted on another computer system using the copied data.

Abstract: Techniques that enable a hypervisor to (1) maintain shared memory pages and (2) handle memory accounting for VMs that are suspended to and resumed from the volatile memory of a host system are provided. Regarding (1), the hypervisor can maintain shared memory pages in volatile memory across the suspend-to-memory and resume-from-memory operations, without having to save their reference counts. Regarding (2), the hypervisor can keep track of the volatile memory reserved and consumed by VMs as they are suspended and resumed, without erroneously double counting that memory.

Abstract: Techniques for implementing RDMA-based recovery of dirty data in remote memory are provided. In one set of embodiments, upon occurrence of a failure at a first (i.e., source) host system, a second (i.e., failover) host system can allocate a new memory region corresponding to a memory region of the source host system and retrieve a baseline copy of the memory region from a storage backend shared by the source and failover host systems. The failover host system can further populate the new memory region with the baseline copy and retrieve one or more dirty page lists for the memory region from the source host system via RDMA, where the one or more dirty page lists identify memory pages in the memory region that include data updates not present in the baseline copy. For each memory page identified in the one or more dirty page lists, the failover host system can then copy the content of that memory page from the memory region of the source host system to the new memory region via RDMA.

Abstract: Techniques for implementing RDMA-based recovery of dirty data in remote memory are provided. In one set of embodiments, upon occurrence of a failure at a first (i.e., source) host system, a second (i.e., failover) host system can allocate a new memory region corresponding to a memory region of the source host system and retrieve a baseline copy of the memory region from a storage backend shared by the source and failover host systems. The failover host system can further populate the new memory region with the baseline copy and retrieve one or more dirty page lists for the memory region from the source host system via RDMA, where the one or more dirty page lists identify memory pages in the memory region that include data updates not present in the baseline copy. For each memory page identified in the one or more dirty page lists, the failover host system can then copy the content of that memory page from the memory region of the source host system to the new memory region via RDMA.

Abstract: In one embodiment, an operating system (OS) or hypervisor running on a computer system can allocate a portion of the volatile memory of the computer system as a persistent memory allocation. The OS/hypervisor can further receive a signal from the computer system's Basic Input/Output System (BIOS) indicating an alternating current (AC) power loss or cycle event and, in response to the signal, can save data in the persistent memory allocation to a nonvolatile backing store. Then, upon restoration of AC power to the computer system, the OS/hypervisor can restore the saved data from the nonvolatile backing store to the persistent memory allocation.

Abstract: Techniques for efficiently purging non-active blocks in an NVM region of an NVM device using virtblocks are provided. In one set of embodiments, a host system can maintain, in the NVM device, a pointer entry (i.e., virtblock entry) for each allocated data block of the NVM region, where page table entries of the NVM region that refer to the allocated data block include pointers to the pointer entry, and where the pointer entry includes a pointer to the allocated data block. The host system can further determine that a subset of the allocated data blocks of the NVM region are non-active blocks and can purge the non-active blocks from the NVM device to a mass storage device, where the purging comprises updating the pointer entry for each non-active block to point to a storage location of the non-active block on the mass storage device.

Abstract: Techniques for efficiently purging non-active blocks in an NVM region of an NVM device using virtblocks are provided. In one set of embodiments, a host system can maintain, in the NVM device, a pointer entry (i.e., virtblock entry) for each allocated data block of the NVM region, where page table entries of the NVM region that refer to the allocated data block include pointers to the pointer entry, and where the pointer entry includes a pointer to the allocated data block. The host system can further determine that a subset of the allocated data blocks of the NVM region are non-active blocks and can purge the non-active blocks from the NVM device to a mass storage device, where the purging comprises updating the pointer entry for each non-active block to point to a storage location of the non-active block on the mass storage device.

Abstract: Techniques for implementing high availability for persistent memory are provided. In one embodiment, a first computer system can detect an alternating current (AC) power loss/cycle event and, in response to the event, can save data in a persistent memory of the first computer system to a memory or storage device that is remote from the first computer system and is accessible by a second computer system. The first computer system can then generate a signal for the second computer system subsequently to initiating or completing the save process, thereby allowing the second computer system to restore the saved data from the memory or storage device into its own persistent memory.

Abstract: Techniques for implementing high availability for persistent memory are provided. In one embodiment, a first computer system can detect an alternating current (AC) power loss/cycle event and, in response to the event, can save data in a persistent memory of the first computer system to a memory or storage device that is remote from the first computer system and is accessible by a second computer system. The first computer system can then generate a signal for the second computer system subsequently to initiating or completing the save process, thereby allowing the second computer system to restore the saved data from the memory or storage device into its own persistent memory.

Abstract: Techniques for using non-volatile random access memory (NVM) as volatile random access memory (RAM) are provided. In one set of embodiments, a computer system can detect that an amount of free space in a volatile RAM of the computer system has become low and, in response, can add one or more memory pages from an unused portion of an NVM of the computer system to the system's volatile RAM pool. Conversely, the computer system can detect that an amount of free space in the NVM has become low and, in response, can return the one or more memory pages from the volatile RAM pool back to the NVM.

Abstract: Techniques for achieving application high availability via application-transparent battery-backed replication of persistent data are provided. In one set of embodiments, a computer system can detect a failure that causes an application of the computer system to stop running. In response to detecting the failure, the computer system can copy persistent data written by the application and maintained locally at the computer system to one or more remote destinations, where the copying is performed in a manner that is transparent to the application and while the computer system runs on battery power. The application can then be restarted on another computer system using the copied data.

Abstract: A journal-based process to achieve atomicity in a device driver write operation includes committing a transaction associated with the operation to a journal that include a status indicating the target block is corrupted. Subsequent to committing the transaction, the data is written to the target block. If the write operation is successfully committed, the transaction can be deleted from the journal. If a system crash occurs (e.g., power failure) before the write operation is successfully committed, the transaction remains in the journal and can be used to update block metadata associated with the target block when the system reboots to indicate that it is corrupted; e.g., the target block is a torn write.

Abstract: Techniques for achieving application high availability via application-transparent battery-backed replication of persistent data are provided. In one set of embodiments, a computer system can detect a failure that causes an application of the computer system to stop running. In response to detecting the failure, the computer system can copy persistent data written by the application and maintained locally at the computer system to one or more remote destinations, where the copying is performed in a manner that is transparent to the application and while the computer system runs on battery power. The application can then be restarted on another computer system using the copied data.

Abstract: Techniques for implementing application fault tolerance via battery-backed replication of volatile state are provided. In one set of embodiments, a primary host system can detect a failure that causes an application of the primary host system to stop running. In response to detecting the failure, the primary host system can replicate volatile state that is used by the application to a secondary host system, where the secondary host system maintains a copy of the application, and where execution of the application is failed over to the copy on the secondary host system using the replicated volatile state.

Abstract: Techniques for achieving application high availability via crash-consistent asynchronous replication of persistent data are provided. In one set of embodiments, an application running on a computer system can, during runtime of the application: write persistent data to a local nonvolatile data store of the computer system, write one or more log entries comprising the persistent data to a local log region of the computer system, and asynchronously copy the one or more log entries to one or more remote destinations. Then, upon detecting a failure that prevents the application from continuing execution, the computer system can copy the local log region or a remaining portion thereof to the one or more remote destinations, where the copying is performed while the computer system runs on battery power and where the application is restarted on another computer system using a persistent state derived from the copied log entries.

Abstract: Exemplary methods, apparatuses, and systems include a distributed memory agent within a first node intercepting an operating system request to open a file from an application running on the first node. The request includes a file identifier, which the distributed memory agent transmits to a remote memory manager. The distributed memory agent receives, from the remote memory manager, a memory location within a second node for the file identifier and information to establish a remote direct memory access channel between the first node and the second node. In response to the request to open the file, the distributed memory agent establishes the remote direct memory access channel between the first node and the second node. The remote direct memory access channel allows the first node to read directly from or write directly to the memory location within the second node while bypassing an operating system of the second node.