Abstract: A computer system includes physical memory devices of different types that store randomly-accessible data in a main memory of the computer system. In one approach, an operating system allocates memory from a namespace for use by an application. The namespace is a logical reference to physical memory devices in which physical addresses are defined. The namespace is bound to a memory type. In response to binding the namespace to the memory type, the operating system adjusts a page table to map a logical memory address in the namespace to a memory device of the memory type.

Abstract: A memory chip having a first set of pins configured to allow the memory chip to be coupled to a first microchip or device via first wiring. The memory chip also having a second set of pins configured to allow the memory chip to be coupled to a second microchip or device via second wiring that is separate from the first wiring. The memory chip also having a data mover configured to facilitate access to the second microchip or device, via the second set of pins, to read data from the second microchip or device and write data to the second microchip or device. Also, a system having the memory chip, the first microchip or device, and the second microchip or device.

Abstract: A system having a string of memory chips that can implement flexible provisioning of a multi-tier memory. In some examples, the system can include a first memory chip in a string of memory chips of a memory, a second memory chip in the string, and a third memory chip in the string. The first memory chip can be directly wired to the second memory chip and can be configured to interact directly with the second memory chip. The second memory chip can be directly wired to the third memory chip and can be configured to interact directly with the third memory chip. As part of implementing the flexible provisioning of a multi-tier memory, the first memory chip can include a cache for the second memory chip, and the second memory chip can include a buffer for the third memory chip.

Abstract: Exemplary methods, apparatuses, and systems include a controller to manage memory proximity disturb. The controller identifies a first memory location in response to an access of a second memory location, the first memory location storing a first value. The controller updates a first disturb value by a first amount, the first disturb value representing a cumulative disturb effect on the first value in the first memory location by accesses to a first plurality of memory locations proximate to the first memory location, the first plurality of memory locations including the second memory location.

Abstract: A system having multiple devices that can host different versions of an artificial neural network (ANN). In the system, inputs for the ANN can be obfuscated for centralized training of a master version of the ANN at a first computing device. A second computing device in the system includes memory that stores a local version of the ANN and user data for inputting into the local version. The second computing device includes a processor that extracts features from the user data and obfuscates the extracted features to generate obfuscated user data. The second device includes a transceiver that transmits the obfuscated user data. The first computing device includes a memory that stores the master version of the ANN, a transceiver that receives obfuscated user data transmitted from the second computing device, and a processor that trains the master version based on the received obfuscated user data using machine learning.

Abstract: A memory device can include three-dimensional memory entities each including a plurality of two-dimensional memory entities. A controller can read data from the memory at a first resolution and collect error rate information from the memory at a second resolution including a portion of a two-dimensional memory entity. The controller can determine a quantity of two-dimensional memory entities that have a greater error rate than a remainder of the two-dimensional memory entities based on the error rate information. The controller can determine a quantity of portions of three-dimensional memory entities that have a greater error rate than a remainder of the portions of three-dimensional memory entities based on the error rate information excluding error rate information for portions of the two-dimensional memory entities associated with the quantity of two-dimensional memory entities.

Abstract: An instance of an event associated with error correcting code operations performed on a data block of the non-volatile memory is identified. An entry for a record is generated. The entry is indicative of the instance of the event. Whether a frequency of the event satisfies a threshold condition based on the record is determined. Responsive to determining that the frequency of the event satisfies the threshold condition, a remix operation on the data block is performed to change a logical to physical association of the data block from a first logical association to a second logical association.

Abstract: A process for wear-leveling in a memory subsystem where references to invalidated chunks and a write count for each of the invalidated chunks of a memory subsystem are received by a wear-leveling manager. The wear-leveling manager orders the received references to the invalidated chunks of the memory subsystem in a tracking structure based on the write count of each of the invalidated chunks, and provides a reference to at least one of the invalidated chunks based on the ordering from the tracking structure to a write scheduler to service a write request, wherein the memory subsystem is wear-leveled by biasing the order of the invalidated chunks to prioritize low write count chunks.

Abstract: A technique of receiving a write transaction directed to a group of memory parcels of a memory device from a client source. The technique determines a state of a first indicator used to indicate which one of two data structures contains a newer mapping of the group of memory parcels, while the other data structure contains an older mapping of the group of memory parcels. The technique determines a state of a second indicator used to indicate which one of the two data structures is in current use for the group of memory parcels and compares the states of the two indicators. When a data structure in current use does not contain the newer mapping, the technique changes the state of the second indicator to the state of the first indicator. The technique writes content of the write transaction to storage locations based on the newer mapping.

Abstract: A computer-implemented method for remediating disruptions to memory cells is described. The method includes writing user data to an aggressor memory cell and determining a write timestamp and an overwrite count associated with the aggressor memory cell. The write timestamp indicates a last write to the aggressor memory cell and the overwrite count indicates the number of writes to the aggressor memory cell during a time period. Based on the write timestamp and the overwrite count, an increment value is determined for use with a disturb counter associated with a neighbor memory cell of the aggressor memory cell. In particular, the determined increment value is used, in response to the write, to increment the disturb counter associated with the neighbor memory cell. When the disturb counter is greater than or equal to a disturb threshold, remediation for the neighbor memory cell is performed.

Abstract: A wear-leveling process for a memory subsystem selects a source chunk to be removed from a usable address space of the memory subsystem to distribute wear across all available chunks in the memory subsystem. The memory subsystem has a plurality of non-volatile memory components. The plurality of non-volatile memory components includes a plurality of chunks including at least one chunk in an unusable address space of the memory subsystem. The wear-leveling process copies valid data of the source chunk to a destination chunk in the unusable address space of the memory subsystem and assigns the destination chunk to a location in the usable address space of the memory subsystem occupied by the source chunk.

Abstract: Exemplary methods, apparatuses, and systems include a controller to manage memory proximity disturb. The controller identifies a first memory location in response to an access of a second memory location, the first memory location storing a first value. The controller updates a first disturb value by a first amount, the first disturb value representing a cumulative disturb effect on the first value in the first memory location by accesses to a first plurality of memory locations proximate to the first memory location, the first plurality of memory locations including the second memory location.

Abstract: Systems, methods and apparatuses to intelligently migrate content involving borrowed memory are described. For example, after the prediction of a time period during which a network connection between computing devices having borrowed memory degrades, the computing devices can make a migration decision for content of a virtual memory address region, based at least in part on a predicted usage of content, a scheduled operation, a predicted operation, a battery level, etc. The migration decision can be made based on a memory usage history, a battery usage history, a location history, etc. using an artificial neural network; and the content migration can be performed by remapping virtual memory regions in the memory maps of the computing devices.

Abstract: Systems, methods and apparatuses to throttle network communications for memory as a service are described. For example, a computing device can borrow an amount of random access memory of the lender device over a communication connection between the lender device and the computing device. The computing device can allocate virtual memory to applications running in the computing device, and configure at least a portion of the virtual memory to be hosted on the amount of memory loaned by the lender device to the computing device. The computing device can throttle data communications used by memory regions in accessing the amount of memory over the communication connection according to the criticality levels of the contents stored in the memory regions.

Abstract: Systems, methods and apparatuses of fine grain data migration in using Memory as a Service (MaaS) are described. For example, a memory status map can be used to identify the cache availability of sub-regions (e.g., cache lines) of a borrowed memory region (e.g., a borrowed remote memory page). Before accessing a virtual memory address in a sub-region, the memory status map is checked. If the sub-region has cache availability in the local memory, the memory management unit uses a physical memory address converted from the virtual memory address to make memory access. Otherwise, the sub-region is cached from the borrowed memory region to the local memory, before the physical memory address is used.

Abstract: Systems, methods and apparatuses to accelerate accessing of borrowed memory over network connection are described. For example, a memory management unit (MMU) of a computing device can be configured to be connected both to the random access memory over a memory bus and to a computer network via a communication device. The computing device can borrow an amount of memory from a remote device over a network connection using the communication device; and applications running in the computing device can use virtual memory addresses mapped to the borrowed memory. When a virtual address mapped to the borrowed memory is used, the MMU translates the virtual address into a physical address and instruct the communication device to access the borrowed memory.

Abstract: Systems, methods and apparatuses of Artificial Neural Network (ANN) applications implemented via Memory as a Service (MaaS) are described. For example, a computing system can include a computing device and a remote device. The computing device can borrow memory from the remote device over a wired or wireless network. Through the borrowed memory, the computing device and the remote device can collaborate with each other in storing an artificial neural network and in processing based on the artificial neural network. Some layers of the artificial neural network can be stored in the memory loaned by the remote device to the computing device. The remote device can perform the computation of the layers stored in the borrowed memory on behalf of the computing device. When the network connection degrades, the computing device can use an alternative module to function as a substitute of the layers stored in the borrowed memory.

Abstract: Systems, methods and apparatuses of distributed computing based on Memory as a Service are described. For example, a set of networked computing devices can each be configured to execute an application that accesses memory using a virtual memory address region. Each respective device can map the virtual memory address region to the local memory for a first period of time during which the application is being executed in the respective device, map the virtual memory address region to a local memory of a remote device in the group for a second period of time after starting the application in the respective device and before terminating the application in the respective device, and request the remote device to process data in the virtual memory address region during at least the second period of time.

Abstract: Systems, methods and apparatuses to provide memory as a service are described. For example, a borrower device is configured to: communicate with a lender device; borrow an amount of memory from the lender device; expand memory capacity of the borrower device for applications running on the borrower device, using at least the local memory of the borrower device and the amount of memory borrowed from the lender device; and service accesses by the applications to memory via communication link between the borrower device and the lender device.

Abstract: Techniques for remapping portions of an array of non-volatile memory (NVM) resident on a die, in which the die is one of a plurality of NVM dice forming a memory device. A processing device partitions the NVM into a plurality of subslice elements comprising respective physical portions of non-volatile memory having proximal disturb relationships. The NVM has a first portion of the subslice elements allocated as user subslice elements and a second portion as spare subslice elements and the processing device performs an error analysis to identify a predetermined number of subslice elements having highest error rates for a memory domain on the die. For the identified subslice elements having the highest error rates, the processing device remaps user subslice elements to spare subslice elements that were not identified as having the highest error rates to remove subslice element or elements having highest error rates from a user space of the NVM.