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: Disclosed are various embodiments for improving resiliency and performance of clustered memory. A computing device can acquire a chunk of byte-addressable memory from a cluster memory host. The computing device can then identify an active set of allocated memory pages and an inactive set of allocated memory pages for a process executing on the computing device. Next, the computing device can store the active set of allocated memory pages for the process in the memory of the computing device. Finally, the computing device can store the inactive set of allocated memory pages for the process in the chunk of byte-addressable memory of the cluster memory host.

Abstract: Disclosed are various embodiments for improving the resiliency and performance of cluster memory. First, a computing device can submit a write request to a byte-addressable chunk of memory stored by a memory host, wherein the byte-addressable chunk of memory is read-only. Then, the computing device can determine that a page-fault occurred in response to the write request. Next, the computing device can copy a page associated with the write request from the byte-addressable chunk of memory to the memory of the computing device. Subsequently, the computing device can free the page from the memory host. Then, the computing device can update a page table entry for the page to refer to a location of the page in the memory of the computing device.

Abstract: Disclosed are various embodiments for improving the resiliency and performance for clustered memory. A computing device can mark a page of the memory as being reclaimed. The computing device can then set the page of the memory as read-only. Next, the computing device can submit a write request for the contents of the page to individual ones of a plurality of memory hosts. Subsequently, the computing device can receive individual confirmations of a successful write of the page from the individual ones of the plurality of memory hosts. Then, the computing device can mark the page as free in response to receipt of the individual confirmations of the successful write from the individual ones of the plurality of memory hosts.

Abstract: Techniques for migrating virtual machines (VMs) in the presence of uncorrectable memory errors are provided. According to one set of embodiments, a source host hypervisor of a source host system can determine, for each guest memory page of a VM to be migrated from the source host system to a destination host system, whether the guest memory page is impacted by an uncorrectable memory error in a byte-addressable memory of the source host system. If the source host hypervisor determines that the guest memory page is impacted, the source host hypervisor can transmit a data packet to a destination host hypervisor of the destination host system that includes error metadata identifying the guest memory page as being corrupted. Alternatively, if the source host hypervisor determines that the guest memory page is not impacted, the source host hypervisor can attempt to read the guest memory page from the byte-addressable memory in a memory exception-safe manner.

Abstract: Various examples are disclosed for cluster resource management using adaptive memory demands. In some examples, a local memory estimate is determined for a workload. The local memory estimate is determined using a memory reclamation parameter for the workload executed by a current host of the workload. A destination memory estimate is also determined for the workload. The destination memory estimate is determined using a full memory estimate unreduced by memory reclamation parameters. The workload is executed using a host that is selected in view of an analysis that uses the local memory estimate for the current host and the destination memory estimate for at least one destination host.

Abstract: Disclosed are various embodiments for improving the resiliency and performance of clustered memory. A computing device can generate at least one parity page from at least a first local page and a second local page. The computing device can then submit a first write request for the first local page to a first one of a plurality of memory hosts. The computing device can also submit a second write request for the second local page to a second one of the plurality of memory hosts. Additionally, the computing device can submit a third write request for the parity page to a third one of the plurality of memory hosts.

Abstract: Disclosed are various embodiments for high throughput reclamation of pages in memory. A first plurality of pages in a memory of the computing device are identified to reclaim. In addition, a second plurality of pages in the memory of the computing device are identified to reclaim. The first plurality of pages are prepared for storage on a swap device of the computing device. Then, a write request is submitted to a swap device to store the first plurality of pages. After submission of the write request, the second plurality of pages are prepared for storage on the swap device while the swap device completes the write request.

Abstract: Methods, apparatus, systems, and articles of manufacture are disclosed for high availability (HA) application migration in a virtualized environment. An example apparatus includes at least one memory, instructions in the apparatus, and processor circuitry to at least one of execute or instantiate the instructions to identify an HA slot in a virtual server rack, the HA slot to facilitate a failover of an application executing on a first virtual machine (VM) in the virtual server rack, the first VM identified as a protected VM, deploy a second VM in the HA slot, transfer data from the first VM to the second VM, and, in response to not identifying a failure of at least one of the first or second VMs during the transfer, trigger a shutdown of the first VM, and synchronize migration data associated with the virtual server rack to identify the second VM as the protected VM.

Abstract: A method of preserving the contiguity of large pages of a workload during migration of the workload from a source host to a destination host includes the steps of: detecting at the destination host, receipt of a small page of zeros from the source host, wherein, at the source host, the small page is part of one of the large pages of the workload; and upon detecting the receipt of the small page of zeros, storing, at the destination host, all zeros in a small page that is part of one of the large pages of the workload.

Abstract: Various aspects are disclosed for unified resource management of containers and virtual machines. A podVM resource configuration for a pod virtual machine (podVM) is determined using container configurations. The podVM comprising a virtual machine (VM) that provides resource isolation for a pod based on the podVM resource configuration. A host selection for the podVM is received from a VM scheduler. The host selection identifies hardware resources for the podVM. A container scheduler is limited to bind the podVM to a node corresponding to the hardware resources of the host selection from the VM scheduler. The podVM is created in a host corresponding to the host selection. Containers are started within the podVM. The containers correspond to the container configurations.

Abstract: Examples described herein include systems and methods for retaining information about bad memory pages across an operating system reboot. An example method includes detecting, by a first instance of an operating system, an error in a memory page of a non-transitory storage medium of a computing device executing the operating system. The operating system can tag the memory page as a bad memory page, indicating that the memory page should not be used by the operating system. The operating system can also store tag information indicating memory pages of the storage medium that are tagged as bad memory pages. The example method can also include receiving an instruction to reboot the operating system, booting a second instance of the operating system, and providing the tag information to the second instance of the operating system. The operating system can use the tag information to avoid using the bad memory pages.

Abstract: Disclosed are various embodiments for optimized memory tiering. A first page can be allocated in a first memory for a process, the first memory being associated with a first memory tier. Accesses of the first page by the process during execution of the process can be monitored. Then, accesses of the first page by the process during execution of the process can be compared to an allocation policy to make a first determination to move the contents of the first page from the first memory to a second memory associated with a second memory tier. Next, the contents of the first page can be copied from the first memory to a second page in the second memory in response to the first determination.

Abstract: Disclosed are various embodiments for optimized memory tiering. An ideal tier size for a first memory and an ideal tier size for a second memory can be determined for a process. Then, a host computing device can be identified that can accommodate the ideal tier size for the first memory and the second memory. Subsequently, the process can be assigned to the host computing device.

Abstract: Disclosed are various embodiments for optimizing the migration of pages of memory servers in cluster memory systems. To begin, a computing device can mark in a page table of the computing device that a page stored on a first memory host is not present. Then, the computing device can flush a translation lookaside buffer of the computing device. Next, the computing device can copy the page from the first memory host to a second memory host. Moving on, the computing device can update a page mapping table to reflect that the page is stored in the second memory host. Then, the computing device can mark in the page table of the computing device that the page stored in the second memory host is present. Subsequently, the computing device can discard the page stored on the first memory host.

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: Disclosed are various embodiments for optimizing the migration of processes or virtual machines in cluster memory systems. To begin, a first computing device can identify a set of pages allocated to a process or virtual machine hosted by the first computing device. Then, the first computing device can identify a subset of the allocated pages that have been accessed with a least a predefined frequency. Next, the first computing device can copy the subset of the allocated pages to a second computing device. Subsequently, the first computing device can copy a page mapping table to the second computing device, the page mapping table specifying which pages in the set of pages allocated to the process or virtual machine are stored by a memory host. Finally, the first computing device can copy remaining ones of the allocated pages to the second computing device.

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: A method of preserving the contiguity of large pages of a workload during migration of the workload from a source host to a destination host includes the steps of: detecting at the destination host, receipt of a small page of zeros from the source host, wherein, at the source host, the small page is part of one of the large pages of the workload; and upon detecting the receipt of the small page of zeros, storing, at the destination host, all zeros in a small page that is part of one of the large pages of the workload.

Abstract: Disclosed are various embodiments for improving resiliency and performance of clustered memory. A computing device can acquire a chunk of byte-addressable memory from a cluster memory host. The computing device can then identify an active set of allocated memory pages and an inactive set of allocated memory pages for a process executing on the computing device. Next, the computing device can store the active set of allocated memory pages for the process in the memory of the computing device. Finally, the computing device can store the inactive set of allocated memory pages for the process in the chunk of byte-addressable memory of the cluster memory host.