Abstract: Disclosed are a general machine learning model generation method and apparatus, and a computer device and a storage medium. The method comprises: acquiring task parameters of a machine learning task (S1201); performing classification processing on the task parameters to obtain task instructions and model parameters (S1202); aggregating the task instructions and the model parameters according to a data type to obtain stack data and heap data (S1203); and integrating the stack data and the heap data to obtain a general machine learning model (S1204). By means of the method, compiled results of a corresponding general model in the running of an algorithm can be directly executed, which avoids repetitive compilation, thus greatly improving the efficiency of machine learning algorithm implementation and shortening the time from compilation to obtaining execution results.

Abstract: The presently disclosed subject matter includes various inventive aspects, which are directed to direct access of a host computer device to a share storage space in a storage system, as well as secured access control of the direct access of the host computer device by a control computer device in the storage system, the direct access including direct read access and direct write access.

Abstract: A set of device maintenance related data is received from each of a plurality of non-volatile memory modules. Each of the plurality of non-volatile memory modules comprise a plurality of non-volatile memory devices. Based at least in part on said set of device maintenance related data a maintenance operation to be performed is determined. The determined device maintenance related operation is performed.

Abstract: Method and system are provided for storage handling guidance for host input/output (I/O) operations. The method includes: providing a guidance array of indications of storage operations handling instructions, the guidance array having multiple dimensions of performance characteristics with each dimension having multiple levels; and associating a reference vector with one or more I/O operations, wherein the reference vector points to a level for each dimension of the array to obtain an indication of a storage operations handling instruction at an intercept of the dimension levels for application by a storage system controller for the one or more I/O operations.

Abstract: A memory system includes a memory blade including a plurality of operation memories; and a controller coupled to the memory blade, wherein the controller includes: a data base (DB) memory suitable for storing a data base (DB) having first information, which is an average usage amount of an operation memory used for processing a generated target workload, as a field and workloads as an entry; an allocation unit suitable for allocating an usage amount of an operation memory for processing a current target workload based on a value of the first information stored in the DB, corresponding to the current target workload requested to be processed; and a monitor suitable for updating the value of the first information to the average usage amount of the operation memory allocated for processing the current target workload after the processing of the current target workload is completed.

Abstract: A solid state drive includes: a plurality of non-volatile memories, each of the non-volatile memories connected to a channel, the channel connected to at least one way connected to a die; a host interface which receives stream data and stream information from a host; and a resource allocator which allocates the stream data to super blocks of the plurality of non-volatile memories on the basis of the stream information. A super block includes a unit super block, and the unit super block includes a block of a first die corresponding to a first channel and connected to a plurality of the ways connected to the first channel. The stream data may include stream groups, and the stream information may include the number of streams included in a stream group. A performance factor of a stream or stream group an extent size of a stream, and an allocation position of the stream, may also be included in the stream information.

Abstract: Provides an efficient method for aligned heap allocations based upon spare space management, provides an efficient method for MPU region allocations based upon aligned subregion allocations, and provides integrated block pools for small data blocks with heap backup if size or alignment requirements cannot be met from pools or pools are empty. The operation of malloc( ), calloc( ), realloc( ) and free( ) on pool blocks versus heap blocks is transparent to application code. These methods are suitable for limited-memory, real-time systems as well as for general-purpose systems.

Abstract: A system, method, and machine-readable storage medium for providing a recommendation to a client to modify minimum IOPS settings are provided. In some embodiments, a client may assign a minimum input/output operations per second (IOPS) setting to each volume of a plurality of volumes. The plurality of volumes may reside in a common cluster and include a first number of volumes. A set of volumes of the plurality of volumes may be determined, where each volume of the set of volumes has a relative disparity that satisfies a relative threshold. The set of volumes includes a second number of volumes. A minimum IOPS recommendation may be transmitted to the client to modify (e.g., increase or decrease) the minimum IOPS settings of the set of volumes in response to determining that a ratio of the second number of volumes to the first number of volumes is greater than a performance threshold.

Abstract: A method for preventing a dirty virtual machine from executing on an undesirable host server includes receiving, by a caching module provided for a first host server that hosts a virtual machine, a write data that is of the virtual machine and that is to be cached for the first host server. The virtual machine uses a virtual hard disk supporting hyper-V Virtual hard disk (VHDX) and virtual hard disk (VHD) file formats or any virtual file format with uniquely identifiable metadata. In response to the receipt of write data, the caching module provided for the first host server changes metadata of virtual hard disk files to a custom format before the virtual machine migrates from the first host server to a second host server, and the virtual machine becomes dirty as a result. When the dirty virtual machine sends a migration request to the second host server, a caching module provided for the second host server checks whether the custom format of the virtual hard disk files is identifiable.

Abstract: A computing device can monitor a set of memory usage metrics of the computing device. Based on historical memory usage data and the set of memory usage metrics, the computing device can determine whether memory usage will exceed a critical memory threshold at a future instance in time. In response to determining that the memory usage will exceed the critical memory threshold at the future instance in time, the computing device can degrade one or more application features of an application executing on the computing device.

Abstract: Systems and methods for analysis of execution patterns for applications executing on remote devices. In some implementations of the system, a knowledge base stores successful traces from a plurality of instances of an application and one or more computing processors in the system receive, via a network interface, call-stack information from an instance of the application executing on a remote device, call-stack information including periodic captures of an execution status for the instance of the application, and determine whether there is a similarity between the call-stack information received from the instance of the application and the stored plurality of successful traces. Responsive to determining a similarity, the computing processors add the remote device to a population of devices likely to execute the object and facilitate further actions specific to the device population.

Abstract: A computer system (e.g. a file system) for managing blocks, comprises a memory unit including a memory and a processing unit including a processor and coupled to the memory unit. The memory unit is configured to store metadata of data including objects. The metadata points to blocks storing the objects in a storage unit including multiple different sized blocks. The processing unit is configured to, for at least one of the objects, determine different sized blocks for storing the object, from among free blocks in the multiple different sized blocks of the storage unit.

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: Graphics processing units (GPUs) and other compute units are allowed to enqueue tasks for themselves by themselves, without needing a host processor to queue the work for the GPU. Built-in functions enable kernels to enqueue kernels for execution on a device. In some embodiments, ndrange kernels execute over an N-dimensional range to provide data-parallel operations. Task kernels provide task-parallel operations. In some embodiments, kernels may be defined using clang block syntax. The order of execution of commands on a compute unit may be constrained or allow execution of commands out-of-order. Compute units may control when kernels enqueued by the compute unit begins execution.

Abstract: A multi-processor system includes a plurality of processors, a scheduler, and a selector. Each of the plurality of processors is included in at least one of a plurality of domains. The plurality of domains include three or more domains and are defined based on a hierarchical domain structure. The scheduler receives and manages at least one task performed by the plurality of processors. The selector selects one of the plurality of domains based on states of the plurality of domains to assign the at least one task to the selected domain, and selects at least one processor among processors in the selected domain based on states of the processors in the selected domain to assign the at least one task to the selected processor, when the at least one task is received.

Abstract: Systems and methods are provided herein for automatically configuring newly installed secondary storage computing devices and managing secondary storage computing devices when one or more become unavailable. For example, a storage manager can then detect the computing resources available to the newly installed secondary storage computing device, assign a role to the newly installed secondary storage computing device based on the detected computing resources, configure the newly installed secondary storage computing device with deduplication and storage policies used by the other secondary storage computing devices, re-partition secondary storage devices to allocate memory for the newly installed secondary storage computing device, and instruct other secondary storage computing devices to replicate their managed data such that the newly installed secondary storage computing device has access to the replicated data.

Abstract: A network device, such as a Network Interface Card (NIC), can have a receive queue (RxQ) that changes size based on whether the network device is in a normal operating mode or in a maintenance mode. In a normal operating mode, it is desirable that the receive queue has a smaller number of free buffers, to increase cache locality in a processor subsystem. However, there can be known periods when the receive queue can be overloaded. During a maintenance period, it is desirable that the receive queue absorbs a large burst of network packets while the processor subsystem is not processing the packets. A solution is to maintain a receive queue at a smaller percentage of its maximum during the normal operation mode, but then before or upon entering the maintenance mode, expand the receive queue to a larger size.

Abstract: A network controller, including: a processor; and a resource permission engine to: provision a composite node including a processor and a first disaggregated compute resource (DCR) remote from the processor, the first DCR to access a target resource; determine that the first DCR has failed; provision a second DCR for the composite node, the second DCR to access the target resource; and instruct the target resource to revoke a permission for the first DCR and grant the permission to the second DCR.

Abstract: The disclosure herein describes reconfiguring a data object in a data store using batch portions of components of the data object. Instructions to reconfigure the data object to a target object layout are received, the data object including a concatenation layer that references a plurality of components. The components are grouped into batch portions and the batch portions are reconfigured in a series based on the received instructions. The reconfiguring for each batch portion includes allocating free storage space according to the target object layout as target components, copying the data stored in the batch portion to the target components, redirecting references in the concatenation layer from the components of the batch portion to the target components, and releasing storage space occupied by the components of the batch portion. As a result, storage space used as transient capacity during the reconfiguring is reduced to a size of the batch portions.

Abstract: The embodiments herein disclose a system and method for implementing reinforcement learning agents using a reinforcement learning processor. An application-domain specific instruction set (ASI) for implementing reinforcement learning agents and reward functions is created. Further, instructions are created by including at least one of the reinforcement learning agent ID vectors, the reinforcement learning environment ID vectors, and length of vector as an operand. The reinforcement learning agent ID vectors and the reinforcement learning environment ID vectors are pointers to a base address of an operations memory. Further, at least one of said reinforcement learning agent ID vector and reinforcement learning environment ID vector is embedded into operations associated with the decoded instruction. The instructions retrieved by agent ID vector indexed operation are executed using a second processor, and applied onto a group of reinforcement learning agents.

Abstract: A computer-implemented method for calibrating DRAM is provided. A non-limiting example of the computer-implemented method includes reading, by a processor, system configuration information and disabling, by the processor, one or more steps in a list of calibration steps to be performed based on the system configuration information to leave a list of remaining calibration steps. Based on a determination that two or more remaining calibration steps are co-dependent, the method configures, by the processor, a single calibration step that encapsulates the co-dependent algorithm and places, by the processor, the single calibration step in a list of steps to be called. The method then provides, by the processor, the list of steps to be called.

Abstract: Disclosed herein are methods, systems, and apparatus, including computer programs encoded on computer storage media, for storing blockchain data. One method includes receiving a request from an application component of a blockchain node to execute one or more software instructions in a trusted execution environment (TEE); determining one or more blockchain node blocks for executing the one or more software instructions; performing error correction coding of the one or more blocks in the TEE to generate one or more encoded blocks; dividing each of the one or more encoded blocks into a plurality of datasets; selecting one or more datasets from each of the one or more encoded blocks; and hashing the one or more datasets to generate one or more hash values corresponding to the one or more datasets for use in replacing the one or more datasets to save storage space of the blockchain node.

Abstract: A catalog of services defines which storage-related services are applied to each service level of storage where storage capacity is managed and allocated in terms of chunks of capacity at different service levels. Each service level is defined based on one or more of tiered storage policy settings, drive size, drive speed, drive count, RAID protection, engine fractions, bandwidth and availability and characterized by one or more performance capabilities. Examples of services in the catalog include but are not limited to mobility, availability, continuity, and recovery services. Any combination of services, or no services at all, can be associated with a given service level of storage. Services may be performed at different performance levels. Further, services may be provided on-demand and billed on a usage basis.

Abstract: Examples of techniques for memory tagging for sensitive data redaction in a memory dump are described herein. An aspect includes receiving a first call to a memory tagging application programming interface (API) from an application, wherein the first call designates a virtual memory page belonging to the application as containing sensitive data. Another aspect includes, based on the first call to the memory tagging API, tagging a physical memory page corresponding to the virtual memory page as sensitive.

Abstract: Garbage collection is performed for a virtualized storage system whose virtual address space is addressed in extents. Valid data in source extents is copied via a cache into destination extents. Once all valid data in a source extent is copied into one or more destination extents, the source extent may be reused. A source extent is released for reuse only after the one or more destination extents that received the valid data copied from the source extent are determined to be full, and the valid data copied from the source extent to the destination extent via the cache is flushed out of the cache.

Abstract: In a data storage system, latency optimization can be practiced by logging a plurality of data accesses to a memory in a register with each data access of the plurality of data accesses corresponding with a command generated by a host connected to the memory. The register may be analyzed with a system module to predict a command execution latency value for the plurality of data accesses that can be used to generate a deterministic data access sequence with the system module. A queue of data accesses can then be reorganized from a first sequence to the deterministic data access sequence to reduce command execution latency variability during a deterministic window selected by the host.

Abstract: Systems, methods, and apparatus including computer-readable mediums for managing block arrangement of super blocks in a memory such as NAND flash memory are provided. In one aspect, a memory controller for managing block arrangement of super blocks in a memory includes control circuitry coupled to the memory having at least two planes of physical blocks and configured to maintain block information of each individual physical block in the planes and select one or more physical blocks from the planes for a super block based on the block information of the physical blocks in the planes.

Abstract: Method and apparatus for managing data in a data storage device configured as a storage compute appliance. In some embodiments, the data storage device has a controller circuit and a non-volatile memory (NVM) with an overall data storage capacity. A processor authenticates each of a plurality of authorized users of the NVM via data exchanges between a host device and the processor without reference to an external authentication authority device. Upon authentication, each authorized user is allocated a separate portion of the overall data storage capacity of the NVM using an allocation model. The size of at least one of the separate portions is subsequently adjusted based on an access history of the NVM. The storage device may be a key-value storage device so that a separate set of key values is provided to each authorized user to identify data blocks stored to and retrieved from the NVM.

Abstract: A method for managing a storage medium in a storage device is provided. The method includes detecting whether there is a problematic storage block in a data storage area of the storage medium, wherein the problematic storage block is a storage block whose health degree is less than a preset value, wherein the health degree is used to measure performance of a storage block in the storage medium and is in direct proportion to the performance of the storage block; and replacing the problematic storage block with a replacement storage block in a replacement area of the storage medium when the problematic storage block is detected, wherein a health degree of a storage block in the data storage area is greater than a health degree of a storage block in the replacement area.

Abstract: A mechanism is provided for destaging one or more data files in a volatile memory using a set of heat registers associated with each data file. Responsive to receiving a notification indicating that free space within the volatile memory has fallen below a predetermined threshold, a rule is implemented, based on values associated with the set of heat registers assigned to each data file, to identify a data file to move to a non-volatile memory, where the set of heat registers comprises a content heat register, an access heat register, and a metadata heat register. The data file is moved to the nonvolatile memory. Responsive to determining that the free space fails to have risen above the predetermined threshold, the operations are repeated until the free space rises above the predetermined threshold at which time the set of heat registers assigned to the data files that were moved are deassigned.

Abstract: The present invention provides systems and methods for data storage. A hierarchical storage management architecture is presented to facilitate data management. The disclosed system provides methods for evaluating the state of stored data relative to enterprise needs by using weighted parameters that may be user defined. Also disclosed are systems and methods evaluating costing and risk management associated with stored data.

Abstract: A memory system includes a controller suitable for managing, first mapping information for mapping a number of first logical addresses among a logical addresses received from a host to a number of first physical addresses and second mapping information for mapping a number of second logical addresses among the logical addresses to a number of second physical addresses, the controller may compress the first mapping information of at least two first logical addresses, which are sequential, and at least two first physical addresses into first sequential mapping information, and may compress the second mapping information of at least two second logical addresses, which are sequential, and at least two second physical addresses into second sequential mapping information, the first physical addresses may be represented by a first number of bits, and the second physical addresses may be represented by a second number of bits.

Abstract: A system including a first storage drive and a superior device superior to the first storage drive, wherein the superior device specifies a first allowable environmental temperature that makes the remaining lifetime of the first storage drive longer than the remaining operation schedule period of the first storage drive, and controls an environmental temperature adjusting device that adjusts the environmental temperature of the first storage drive on the basis of the first allowable environmental temperature.

Abstract: Disclosed herein are methods, systems, and apparatus, including computer programs encoded on computer storage media, for storing blockchain data. One method includes receiving a request from an application component of a blockchain node to execute one or more software instructions in a trusted execution environment (TEE); determining one or more blockchain node blocks for executing the one or more software instructions; performing error correction coding of the one or more blocks in the TEE to generate one or more encoded blocks; dividing each of the one or more encoded blocks into a plurality of datasets; selecting one or more datasets from each of the one or more encoded blocks; and hashing the one or more datasets to generate one or more hash values corresponding to the one or more datasets for use in replacing the one or more datasets to save storage space of the blockchain node.

Abstract: A method for managing garbage collection in a memory subsystem, where a stream data manager writes data units from a stream of data into an allocated portion of memory composed of a plurality of blocks. The stream data manager evaluates a behavior of the stream of data to calculate the stream's efficiency, the behavior including amounts of valid data units from the stream of data in the allocated portion of memory. The stream data manager estimates a number of block stripe fills until an amount of valid data units is predicted to be within a predetermined range of a threshold value of valid data units in the block using the evaluated behavior. The stream data manager performs the estimated number of block stripe fills. The stream data manager performs the garbage collection of a first block of the plurality of blocks in response to performance of the estimated number of block stripe fills.

Abstract: Memory management in a computer system is described. Memory is allocated within page-aligned real memory storage to requesting processes. During a normal operation mode, pages of the real memory storage are allocated in accordance with a buddy allocation algorithm. A triggering condition is detected. Responsive to the detection, at least one contiguous page-sized or smaller memory segment that violates the buddy allocation algorithm is allocated to a requesting process.

Abstract: A method and apparatus is disclosed for using supervised learning with closed loop feedback to improvement of output consistency for memory arrangements, such as a solid state drive.

Abstract: Systems and methods for adapting garbage collection (GC) operations in a memory device to a pattern of host accessing the device are discussed. The host access pattern can be represented by how frequent the device is in idle states free of active host access. An exemplary memory device includes a memory controller to track a count of idle periods during a specified time window, and to adjust an amount of memory space to be freed by a GC operation in accordance with the count of idle periods. The memory controller can also dynamically reallocate a portion of the memory cells between a single level cell (SLC) cache and a multi-level cell (MLC) storage according to the count of idle periods during the specified time window.

Abstract: A computer-implemented method for diagnosing workload performance problems in computer servers includes measuring activity metrics. The method further includes aggregating activity metrics for each user. The method further includes generating condensed diagnostic data for identifying workload performance problems on a synchronized, regular interval. Generating the diagnostic data includes grouping users into buckets based on bucket and user attributes. Generating the diagnostic data includes aggregating activity metrics across all users in each bucket. Generating the diagnostic data includes including one or more most significant users and corresponding activity metrics for each activity in each bucket, and recording bucket contents. Diagnostic data can be generated in a human consumable form for human analysis or in a machine consumable form for machine analysis through machine learning.

Abstract: An apparatus is described that includes a memory controller to couple to a multi-level memory characterized by a faster higher level and a slower lower level. The memory controller having early demotion logic circuitry to demote a page from the higher level to the lower level without system software having to instruct the memory controller to demote the page and before the system software promotes another page from the lower level to the higher level.

Abstract: Techniques for data processing may include: receiving a data chunk and an associated digest; and performing data deduplication processing for the data chunk comprising: determining whether there is an existing entry in a deduplication digest cache for the digest; and responsive to determining there is no existing entry in the deduplication digest cache, performing processing including: determining whether there is an existing entry in a mapping structure for the digest, the mapping structure mapping digests to associated pages of related entries in a deduplication data store; and responsive to determining there is an existing entry in the mapping structure, performing second processing including: obtaining, from the existing entry, a page mapped to the digest; and loading the page of entries from the deduplication data store into the deduplication digest cache. At least some entries of the page may be prefetched and loaded into the deduplication digest cache prior to use.

Abstract: A virtual memory system includes a virtual memory engine coupled to a plurality of physical memory devices and a virtual memory database. During an initialization process, virtual memory engine uses a first unique global identifier to create virtual memory in the virtual memory database by mapping a continuous virtual memory address range to non-continuous physical memory device address ranges that are provided across the plurality of physical memory devices. During the initialization process, or subsequently during runtime, the virtual memory engine uses a second global unique identifier to define a virtual memory device namespace in the virtual memory that includes a first continuous subset of the continuous virtual member address range. During runtime, the virtual memory engine then provides read and write block mode access to the plurality of physical memory devices via the virtual memory device namespace defined in the virtual memory database.

Abstract: Systems and methods for analysis of execution patterns for applications executing on remote devices. In some implementations of the system, a knowledge base stores successful traces from a plurality of instances of an application and one or more computing processors in the system receive, via a network interface, call-stack information from an instance of the application executing on a remote device, call-stack information including periodic captures of an execution status for the instance of the application, and determine whether there is a similarity between the call-stack information received from the instance of the application and the stored plurality of successful traces. Responsive to determining a similarity, the computing processors add the remote device to a population of devices likely to execute the object and facilitate further actions specific to the device population.

Abstract: A hardware configuration circuit configured to sequentially read data packets and transmit the data to a configuration data client. The configuration data client configured to receive a first and a second set of configuration data addressed to a respective address. The client is configured to store the first set of configuration data in a register and verify whether further configuration data may be written to the respective register as a function of a type identification signal. In response, the configuration data client is configured to overwrite the first set of configuration data by storing the second set of configuration data in the respective register or maintain the first set of configuration data by inhibiting storage of the second set of configuration data received in the respective register. The configuration corresponding to verifying whether further configuration data may be written to the register.

Abstract: A autonomic method for the comprehensive profiling of near or real time representations of time sequenced, synchronized, characterized business activity by any entity within a group of commercially related subscribing business partners, for the purpose of creating a near or real time universal business activity mosaic. An autonomic method within a group of commercially interrelated businesses to utilize a universal business activity mosaic to render a virtual customer storefront, whereby a business may discretely witness in time and character the sale of its own assets and services either fully or partially comprising those sold to anybody of related intermediate or end customers for the purpose of purchasing, selling, financing, warehousing or physically transporting assets on demand.

Abstract: Devices and systems for distributing data across disaggregated memory resources is disclosed and described. An acceleration controller device can include an adaptive data migration engine (ADME) configured to communicatively couple to a fabric interconnect, and is further configured to monitor application data performance metrics at the plurality of disaggregated memory pools for a plurality of applications executing on the plurality of compute resources, select a current application having a current application data performance metric, determine an alternate memory pool from the plurality of disaggregated memory pools estimated to increase application data performance relative to the current application data performance metric, and migrate the data from the current memory pool to the alternate memory pool.

Abstract: A bitmap-based storage space management system and methods thereof are provided, which realize divisible storage blocks that a small block is distributed when storing a small data object and a large block is distributed when storing a large data object. Through a two layer structure of index bitmaps and state bitmaps, a management of a storage space is realized. The index bitmaps comprise a main resource index bitmap (11), a large block index bitmap (12), a sub-resource index bitmap (13) and a small block index bitmap (14). The state bitmaps comprise a resource state bitmap (21) and a distribution state bitmap (22). Through the resource and distribution state bitmaps, the storage space is described respectively as an unoccupied divisible state, an occupied divisible state and an occupied indivisible state. In an above storage space structure, large block distribution, small block distribution, large block release and small block release are realized.

Abstract: Methods and systems for handling requests for data corresponding to a volume of data are disclosed. A method involves generating meta data corresponding to a volume of data, the meta data including a set of range status registers, wherein each range status register corresponds to a unique range of memory in the volume of data, wherein the state of a range status register indicates whether or not data has been persisted within the range of memory that corresponds to the range status register and processing a request for a retrieve operation from the volume of data or for a persist operation to the volume of data as a function of a range status register in the meta data that corresponds to the request.

Abstract: Disclosed herein are system, method, and various embodiments for migrating a program from a first device to a second device. An embodiment operates by receiving a playbook comprising a source location of source code on the first device and a target location for the source code on the second device. An execution function for transferring the source code from the first device to the second device is determined. A set of classes on the first device to which to apply the playbook are identified. At least one internal reference and at least one external reference in the set of classes are updated. The set of classes are migrated from the source location on the first device to the target location on the second device based on the playbook. The at least one internal reference is updated after the migration, wherein the program is executable using the migrated classes after the updating.

Abstract: Detecting and remediating memory leaks associated with an application environment can include monitoring allocations of memory from a managed memory space to respective operations to produce memory allocation data and monitoring deallocations of memory to at least some of the respective operations to produce memory deallocation data. A trend in memory leakage can be determined based on samples of the memory allocation or deallocation data. A projection of future memory usage by operations associated with the trend can be determined using binned sets of the memory allocation data and the memory deallocation data. A predicted time at which memory usage by the operations associated with the trend is expected to exceed a threshold can be determined using the projection of future memory usage. A remediation action can be performed before the predicted time to prevent a memory constraint from occurring with respect to the application environment.