Abstract: A degree of fragmentation is determined based on a number of holes present in a storage system layout or a portion of a layout. Edges between the holes and used portions of the storage system are tabulated by scanning a storage space. The occurrences of a pattern of used/available allocation units and/or the occurrences of another pattern available/used allocation units are recognized. A fragmentation value is calculated based on occurrences of the patterns in view of the total storage space. The present fragmentation measurement system utilizes the number of occurrences of the holes in assessing fragmentation.

Abstract: A method for selectively generating suggested default values for I/O configurations is provided. The method identifies a first selection including a first input value for an I/O configuration. The method determines a set of remaining input options based on the first selection. The method accesses a set of decision trees based on the set of remaining input options and selects a decision tree of the set of decision trees based on the first input value. The method generates a suggested value for a subsequent selection for the I/O configuration and causes presentation of the suggested value and a user interface element representing the subsequent selection.

Abstract: Methods that boot a secondary operating system (O/S) kernel with reclaimed primary kernel memory are disclosed herein. One method includes booting, via a processor performing a boot algorithm, a secondary kernel for an O/S in response to a primary kernel for the O/S going offline, in which the secondary kernel is configured to be loaded to a reserved memory area. The method further includes reclaiming memory space from the primary kernel for use in booting the secondary kernel in response to a determination that the reserved memory area includes insufficient memory space for completing the boot algorithm. Also disclosed herein are apparatus, systems, and computer program products that can include, perform, and/or implement the methods for providing a secondary kernel that includes a reserved area in memory.

Abstract: High performance space efficient distributed storage is disclosed. For example, a distributed storage volume (DSV) is deployed on a plurality of hosts, with a first host storing a local cache, and a storage controller executing on a processor of the first host receives a request to store a first file. The first file is stored to the local cache. The DSV is queried to determine whether a second file that is a copy of the first file is stored in the DSV. In response to determining that the DSV lacks the second file, the first file is transferred from the local cache to the DSV and then replicated to a second host of the plurality of hosts. In response to determining that the second file resides in the DSV, a reference to the second file is stored in the DSV and then replicated to the second host.

Abstract: A data storage device and method for enabling endurance re-evaluation are provided. In one embodiment, a data storage device is provided comprising a memory and a controller. The controller is configured to receive data and a first storage requirement for the data from a host; store the data in a first group of memory cells in the memory, wherein the first group of memory cells satisfies the first storage requirement; receive a notification from the host that the first storage requirement has changed to a second storage requirement; and move the data from the first group of memory cells to a second group of memory cells in the memory, wherein the second group of memory cells satisfies the second storage requirement. Other embodiments are provided.

Abstract: Systems and methods are described for a non-disruptive planned failover from a primary copy of data at a primary storage system to a mirror copy of the data at a cross-site secondary storage system. According to an example, a planned failover feature of a multi-site distributed storage system provides an order of operations such that a primary copy of a first data center continues to serve I/O operations until a mirror copy of a second data center is ready. This planned failover feature improves functionality and efficiency of the distributed storage system by providing non-disruptiveness during planned failover—even if various failures occur. The planned failover feature also includes a persistent fence to avoid serving I/O operations during a timing window when both primary data storage and secondary data storage are attempting to have a master role to serve I/O operations and this avoids a split-brain situation.

Abstract: A Software in the Loop (SiL) system and method is disclosed which may include a simulator operable to provide an environment to simulate dynamic systems, enable rapid development, validation of systems, and testing of complex systems. The system and method may include assembling one or more unsecured models operable to simulate the real-world system. The system and method may then encrypt and generate at least one secured model from the one or more unsecured models using a first cryptographic key. The at least one secured model may be decrypted using a sealed decryption key. The decrypted secured model may then be executed within the one or more TEEs. The at least one secured model may be operable to process incoming data and outgoing data.

Abstract: A method for adjustable error correction in a storage cluster is provided. The method includes determining health of a non-volatile memory of a non-volatile solid-state storage unit of each of a plurality of storage nodes in a storage cluster on a basis of per flash package, per flash die, per flash plane, per flash block, or per flash page. The determining is performed by the storage cluster. The plurality of storage nodes is housed within a chassis that couples the storage nodes as the storage cluster. The method includes adjusting erasure coding across the plurality of storage nodes based on the health of the non-volatile memory and distributing user data throughout the plurality of storage nodes through the erasure coding. The user data is accessible via the erasure coding from a remainder of the plurality of storage nodes if any of the plurality of storage nodes are unreachable.

Abstract: Systems, apparatuses, and methods related to memory tracing in an emulated computing system are described. Static tracepoints can be inserted into a particular function as part of operating the emulated computing system. By executing the function including the static tracepoints as part of a memory access request, the emulated computing system can receive information corresponding to both a virtual address and a physical address in a real computing system in which data corresponding to the memory access request is stored.

Abstract: This invention provides an image capturing apparatus having an interface for connecting with a device including a programmable integrated circuit, and a storage unit for storing configuration data of the programmable integrated circuit, wherein the apparatus comprises an updating unit which updates the configuration data of the storage unit of the device via interface; and a determining unit which determines whether the updating is completed by the updating unit, wherein the updating unit writes, via the interface, the configuration data to the storage unit by issuing a updating request to the device and transferring the configuration data, and receives, via the interface, the configuration data by issuing a readout request to the device, wherein the determining unit determines whether the updating is completed by determining whether the received configuration data matches the written configuration data.

Abstract: An apparatus comprises at least one processing device comprising a processor coupled to a memory. The at least one processing device is configured to identify one or more network shares of a filesystem. The at least one processing device is also configured to store, in the filesystem, at least one network share metadata file comprising metadata characterizing the identified one or more network shares of the filesystem. The at least one processing device is further configured to generate a snapshot of the filesystem, the generated snapshot comprising the at least one network share metadata file. The generated snapshot is utilizable for performing a recovery of the filesystem and the identified one or more network shares using at least a portion of the metadata from the at least one network share metadata file.

Abstract: A method of operating a multi-bank storage device includes transmitting a write command including stream identification information to the multi-bank storage device, and allocating at least one bank, in which data associated with the write command is to be stored, from among a plurality of banks in the multi-bank storage device, based on striping size information included within the stream identification information. Upon allocation, the data is written into the allocated at least one bank.

Abstract: Memory usage can be optimized during an imaging process to enable larger-sized deployments and client servicing. A variety of components can be employed on a computing device to present a hybrid virtual volume to the operating system during an imaging process. The hybrid virtual volume will appear to the operating system as having a large size while consuming minimal amounts of memory and disk. As a result, the risk of the imaging process failing for lack of resources will be minimized even when the image is large.

Abstract: A storage system comprises a plurality of storage devices that are configured to store data pages. The data pages are distributed across the plurality of storage devices in a plurality of data stripes. A processing device of a storage controller is configured to identify a data stripe that corresponds to a given data page that was stored on a failed storage device and to provide an indication of the identified data stripe to a first storage device. A first processing device of the first storage device is configured to obtain data pages corresponding to the identified data stripe from at least one storage device other than the failed storage device and to perform a parity calculation based at least in part on the obtained data pages to rebuild the given data page. The rebuilt given data page is then stored on a third storage device.

Abstract: The disclosure describes techniques that enable detection of memory leaks of software executing on devices within a computer network. An example network device includes memory and processing circuitry. The processing circuitry monitors a usage of the memory by a software component operating within the network device. The processing circuitry periodically determines a memory growth pattern score for the software component based on the usage of the memory. The processing circuitry also predicts whether the user-level process is experiencing a memory leak based on the memory growth pattern score. The processing circuitry applies confirmation criteria to current memory usage of the software component to confirm that the software component is experiencing the memory leak. When the software component is experiencing the memory leak, the processing circuitry generates an alert.

Abstract: Analytics-based optimized maintenance operations for a hyper-converged infrastructure are described. An example includes instructions to establish an order for a sequence of maintenance operations including collecting data points that relate to high availability of multiple nodes in a storage infrastructure; performing analysis of the collected data, including discovery of groups of nodes that don't have mutual relations with other paired nodes; receiving a request to perform a maintenance operation for the plurality of nodes; generating an ordered sequence of groups of nodes for the maintenance operation based at least in part on the analysis of the collected data, each group including one or more nodes; and performing the maintenance operation for the plurality of nodes according to the sequence of groups of nodes, wherein the maintenance operation includes a power cycle for each of the plurality of nodes.

Abstract: A system and method for dynamically and automatically sharing resources of a coprocessor AI accelerator based on workload changes during training and inference of a plurality of neural networks. The method comprising steps of receiving a plurality of requests from each neural network and high-performance computing applications (HPCs) through a dynamic adaptive scheduler module. The dynamic adaptive scheduler module morphs the received requests into threads, dimensions and memory sizes. The method then receives the morphed requests from the dynamic adaptive scheduler module through client units. Each of the neural network applications is mapped with at least one of the client units on a graphics processing unit (GPU) hosts. The method then receives the morphed requests from the plurality of client units through a plurality of server units. Further, the method receives the morphed request from the plurality of server units through one or more coprocessors.

Abstract: A storage system determines whether its memory is fragmented (e.g., based on a host read pattern that indicates that a sequential file is being read, but the number of data sense operations required to perform this read indicates that file is stored non-sequentially in the memory). If the storage system determines that its memory is fragmented, the storage system can perform a defragmentation operation on the memory. This defragmentation operation can be done invisibly to the host (i.e., without receiving any hint or instruction from the host to perform the defragmentation operation).

Abstract: A method is used for efficient storage allocation in a storage system. An infrastructure manager receives a request for storage from a first storage operating entity. The infrastructure manager identifies, from entries in a pagebin, a pagelist with sufficient available storage to satisfy the request. The infrastructure manager locks the pagebin while assigning the identified pagelist to the first storage operating entity. Upon a commitment of the identified pagelist by the first storage operating entity, the infrastructure manager unlocks the pagebin and identifies, from the entries in the pagebin, a second pagelist with sufficient available storage to satisfy a request from a second storage operating entity.

Abstract: Techniques involve determining a first slice distribution used to build first storage units in a first pool; in response to a determination that the first pool is expanded to a second pool, determining, at least based on a sum of the slices having been used to build the first storage units, a second slice distribution of updated slices used to build the first storage units in the second pool; determining, based on the first distribution and the second distribution, a first available number of slices and a second available number of slices available for building second storage units in the second pool, the second storage units being different from the first storage units; and determining, at least based on the first available number and the second available number, the number of the second storage units allowed to be built. Accordingly, available capacity allowed for building can be accurately estimated.

Abstract: An information handling system may include a processor and a storage subsystem. The storage subsystem may include a non-expander backplane, a first plurality of storage resources coupled to the processor via the non-expander backplane, and a second plurality of storage resources coupled to the processor via a communication path that does not include the non-expander backplane. The information handling system may be configured to provide slot numbers for the storage resources according to a numbering scheme in which a storage resource from the first plurality of storage resources and a storage resource from the second plurality of storage resources have the same slot number.

Abstract: A neural network method, comprising: modeling an environment; implementing a policy based on the modeled environment, to perform an action by an agent within the environment, having at least one estimated dynamic parameter; receiving an observation and a temporally-associated cost or reward based on operation of the agent in the environment controlled according to the policy; and updating the policy, dependent on the received observation and the temporally-associated cost or reward, to improve the policy to optimize an expected future cumulative cost or reward. The policy may represent a set of parameters defining an artificial neural network having a plurality of hierarchical layers and having at least one layer which receives inputs representing aspects of the received observation indirectly from other neurons, and produce outputs to other neurons which indirectly implement the policy, the plurality of hierarchical layers being trained according to respectfully distinct training criteria.

Abstract: Data migration from a legacy data storage system into an ordered event stream (OES) data storage system is disclosed. In contrast to conventional techniques, the disclosed subject matter can provide for migrating legacy data into a first portion of a vintage OES (VOES) and new data, intended for the legacy data storage system during the migration process, into a second portion of the VOES. The first and second portions of the VOES can then be linked at completion of the migration process. Avoiding writing new data to the legacy data storage system prior to migration enables a corresponding reduction in data access disruption. Moreover, event sequencing by separating migrated legacy events from new events can preserving event order. Scaling of vintage streams can be supported. Vintage streams can support function calls to a single VOES in contrast to the more conventional use of multiple streams.

Abstract: A method for reducing a number of copies required to send a data sample with a Data Distribution Service (DDS) type in a system using an Object Management Group (OMG) Data Distribution Service (DDS) and a Real-Time Publish Subscribe (RTPS) protocol is provided. Key to the invention is the definition/creation of a memory representation of the data samples for the DDS type that is equal to the network representation of the data samples for the DDS type. Sending of data samples to the DataReader is accomplished without making a serialization copy of the data samples, and for the receiving the data samples from the DataWriter is accomplished without making a deserialization copy of the data samples. Further, a method is provided for accessing to a network representation of data samples with a DDS type in a system using an OMG DDS and a RTPS protocol.

Abstract: An information handling system includes a processor that detects a cache flush request of a memory device within the processor, and identifies multiple blocks of data within an address space associated with the cache flush request. The processor groups the multiple blocks of data into a single composite block of data, and compresses the composite block of data. The processor stores the compressed composite block of data, and stores metadata for the compressed composite block of data. The metadata includes information for both the composite block of data and information for each of the multiple blocks of data.

Abstract: Techniques are described herein that are capable of dynamically re-allocating computing resources while maintaining network connection(s). Applications of users are run in a computing unit. Computing resources are allocated among the applications based at least in part on dynamic demands of the applications for the computing resources and resource limits associated with the respective customers. In a first example, the computing resources are dynamically re-allocated among the applications, as a result of changing the resource limit of at least one customer, while maintaining at least one network connection between a client device of each customer and at least one respective application. In a second example, the computing resources are dynamically re-allocated among the applications, as a result of changing the resource limit of at least one customer, while maintaining at least one network connection between an interface and a client device of each customer.

Abstract: Techniques for managing storage blocks involve: according to a first request for allocating a target number of consecutive storage blocks from a storage device, querying an index table that is used to record index information of a plurality of groups of consecutive idle storage blocks in the storage device to determine a first group of consecutive idle storage blocks from the plurality of groups of consecutive idle storage blocks; allocating the target number of consecutive storage blocks from the first group of consecutive idle storage blocks; updating a bitmap that is used to record states of a plurality of storage blocks in the storage device; and updating first index information of the first group of consecutive idle storage blocks recorded in the index table. Accordingly, such techniques may reduce storage complexity and improve system performance.

Abstract: A database management system is described that can encode data to generate a plurality of data vectors. The database management system can perform the encoding by using a dictionary. The database management system can adaptively reorder the plurality of data vectors to prepare for compression of the plurality of data vectors. During a forward pass of the adaptive reordering, most frequent values of a data vector of the plurality of data vectors can be moved-up in the data vector. During a backward pass of the adaptive reordering, content within a rest range of a plurality of rest ranges can be rearranged within the plurality of data vectors according to frequencies of the content. The reordering according to frequency can further sort the rest range by value. Related apparatuses, systems, methods, techniques, computer programmable products, computer readable media, and articles are also described.

Abstract: Systems, methods, and apparatus related to a memory system that manages an interface for a volatile memory device and a non-volatile memory device to control memory system power. In one approach, a controller evaluates a demand on memory performance. If the demand of a current computation task needed by the host is high, a DRAM device is powered-up to meet the demand. Otherwise, if the non-volatile memory device is adequate to meet the demand, the DRAM memory is partially or fully-powered down to save power. In another approach, a task performed for a host device uses one or more resources of a first memory device (e.g., DRAM). A performance capability of a second memory device (e.g., NVRAM) is determined. A controller of the memory system determines whether the performance capability of the second memory device is adequate to service the task.

Abstract: Disclosed is a network management system that provides an interface to enable diagnostics and troubleshoot of a remotely managed multi-site network. Some embodiments provide a natural language interface, while other embodiments provide a chatbot type interface that communicates with a technician via traditional text information on a display screen. The diagnostic and troubleshooting capabilities search a central data store that receives device property information from each site of the multi-site network. Based on devices or users that match portions of the entity, queries to the data store are initiated to obtain additional data on the devices. A response to the query is then provided based on the properties of the devices.

Abstract: A computer-implemented method for duplexing data from multiple files systems onto a shared tape is disclosed. The computer-implemented method includes receiving a unique identifier of a first tape image written to a primary storage device of a first client system. The computer-implemented method further includes determining that a position of a first end-of-data marker on the first tape image differs from a position of a second end-of-data marker on a tape corresponding to the unique identifier of the first tape image. Responsive to determining that the position of the first end-of-data marker on the first tape image differs from the position of the second end-of-data marker on the tape corresponding to the unique identifier of the first tape image, the computer-implemented method further includes copying a difference in data between the second end-of-data marker on the tape and be first end-of-data marker on the first tape image to the tape.

Abstract: A user definition of a memory shape can be received and a multidimensional, contiguous, physical portion of a memory array can be allocated according to the memory shape. The user definition of the memory shape can include a quantity of contiguous columns of the memory array, a quantity of contiguous rows of the memory array, and a major dimension of the memory shape. The major dimension can correspond to a dimension by which to initially stride data stored in the memory shape.

Abstract: A method for data processing, comprising updating intermediate storage information according to data to be processed and address information of the data to be processed in a first storage space, until the intermediate storage information has reached a preset size; and performing, in the first storage space, an operation corresponding to the data to be processed using the intermediate storage information, when the intermediate storage information reaches the preset size. By the above method, the computing cost for performing an operation corresponding to the data to be processed in the first storage space can be reduced, the efficiency in performing the corresponding operation can be improved, and with intermediate storage information adapted to the first storage spaces of different sizes, the number of operations on the first storage spaces can be reduced.

Abstract: Examples relate to preventing scheduling or executing a resource on an inconsistent host node in a networked system. Some examples track a taint status of the host node and identify whether the host node is inconsistent based on the taint status of the host node over a predefined period of time. Upon identifying that the host node is inconsistent, a master taint is applied on the inconsistent host node, which prevents scheduling or executing a resource on the identified inconsistent host node.

Abstract: A memory device includes a non-volatile (NV) memory including a plurality of NV memory elements. A method for performing programming management of the NV memory includes: setting a programming sequence of the NV memory elements; determining a selection interval between each of the NV memory elements according to the programming sequence and a serial number of each of the NV memory elements; for a target NV memory element of the plurality of NV memory elements in the programming sequence, determining a serial number of an immediately previous NV memory element in the programming sequence according to the selection interval and a serial number of the target NV memory element; determining whether the immediately previous NV memory element is in a busy state; and only when the immediately previous NV memory element is not in the busy state, programming the target NV memory element.

Abstract: Embodiments for providing global inline name space verification for a distributed file system in a network of a metadata server coupled to a plurality of data servers by taking a global dataless snapshot of a namespace of the distributed file system; walking all of the files in the namespace for each data server and the metadata server to generate parsed information; combining, by an XOR operation, the parsed information into data blocks for each server; obtaining a checksum of each data block of the data blocks; comparing actual and expected checksums from the metadata server and all of the data servers; and generating an alert if a comparison of any actual and expected checksums do not match.

Abstract: Disclosed is an evaluation system for evaluating a resource operation of an information system by a user. The evaluation system includes one or more arithmetic units, and one or more storage devices. The one or more storage devices are configured to store risk management information for managing an erroneous operation risk evaluation index relating to the resource operation of the information system. The one or more arithmetic units are configured to: receive operation information indicating a first resource operation designated by a first user, and evaluate an erroneous operation risk of the first resource operation based on the first resource operation and the risk management information.

Abstract: A data processing system and method, and a corresponding apparatus, where the data processing system includes a controller and at least two storage nodes. The controller is configured to receive, using a first coupling between the controller and a host, an operation request received from the host, where the operation request includes an identity of target data and an operation type, determine at least one target storage node from the at least two storage nodes according to the identity of the target data, and send an instruction message to the at least one target storage node using a second coupling to the at least one target storage node, where the at least one target storage node is configured to send the target data to the host or obtain the target data from the host according to the instruction message.

Abstract: A storage client needs to store to-be-written data into a distributed storage system, and storage nodes corresponding to a first data unit assigned for the to-be-written data by a management server are only some nodes in a storage node group. When receiving a status of the first data unit returned by the management server, the storage client may determine quantities of data blocks and parity blocks needing to be generated during EC coding on the to-be-written data. The storage client stores the generated data blocks and parity blocks into some storage nodes designated by the management server in a partition where the first data unit is located. Accordingly, dynamic adjustment of an EC redundancy ratio is implemented, and the management server may exclude some nodes in the partition from a storage range of the to-be-written data based on a requirement, thereby reducing a data storage IO amount.

Abstract: The present disclosure provides a management method and an apparatus for coexisting multi-storage mediums. The method includes: scanning a plurality of storage mediums to identify the storage medium operating in an online mode; configuring the storage medium operating in the online mode as a first storage medium; establishing a doubly linked list by a controller, wherein the doubly linked list records storage mediums operating in the online mode; determining whether the first storage medium is in a register table; turning on a switching mode of the first storage medium upon determining the first storage medium is in the register table; and retrieving a switching command via a storage layer, wherein the switching command is generated by the controller in response to user request.

Abstract: A performance manager (400, 500) and a method (200) performed thereby are provided, for managing the performance of a logical server of a data center. The data center comprises at least one memory pool in which a memory block has been allocated to the logical server. The method (200) comprises determining (230) performance characteristics associated with a first portion of the memory block, comprised in a first memory unit of the at least one memory pool; and identifying (240) a second portion of the memory block, comprised in a second memory unit of the at least one memory pool, to which data of the first portion of the memory block may be migrated to apply performance characteristics associated with the second portion. The method (200) further comprises initiating migration (250) of the data to the second portion of the memory block.

Abstract: Described herein are techniques for improving transfer of metadata from a metadata database to a database stored in a data system, such as a data warehouse. The metadata may be written into the metadata database with a version stamp, which is monotonic increasing register value, and a partition identifier, which can be generated using attribute values of the metadata. A plurality of readers can scan the metadata database based on version stamp and partition identifier values to export the metadata to a cloud storage location. From the cloud storage location, the exported data can be auto ingested into the database, which includes a journal and snapshot table.

Abstract: An initialization data memory space allocation system includes a memory system having a memory space that includes an initialization data bucket that reserves a contiguous subset of the memory space for initialization data. Each initialization engine that is coupled to the memory system is configured during initialization operations to allocate, for that initialization engine, a portion of the contiguous subset of the memory space reserved by the initialization data bucket, and then store initialization data in that portion of the contiguous subset of the memory space reserved by the initialization data bucket. A runtime engine that is coupled to the memory system is configured, during runtime operations, to claim the contiguous subset of the memory space reserved for initialization data by the initialization data bucket for runtime data, and store runtime data in at least a portion of the contiguous subset of the memory space.

Abstract: In some examples, a system assigns workload fingerprints to each respective storage volume of a plurality of storage volumes, the workload fingerprints assigned to the respective storage volume across a plurality of points. Based on the workload fingerprints assigned to respective storage volumes of the plurality of storage volumes, the system groups the storage volumes into clusters of storage volumes. The system manages an individual cluster of the clusters of storage volumes according to an attribute associated with the individual cluster.

Abstract: Systems, methods, and non-transitory computer readable media are provided for recursively searching a plurality of workspaces of the system for linked data associated with the seed data, initiating an endpoint process for each the seed data and the linked data, and, upon completion of the search, delete the seed data and the linked data identified based at least in part on the endpoint process. The process may be automatically repeated at a predetermined time interval to identify and remove future data that is stored in the plurality of datasets.

Abstract: Methods, systems, and devices for caching identifiers for access commands are described. A memory sub-system can receive an access command to perform an access operation on a transfer unit of the memory sub-system. The memory sub-system can store an identifier associated with the access command in a memory component and can generate an internal command using a first core of the memory sub-system. In some embodiments, the memory sub-system can store the identifier in a shared memory that is accessible by the first core and can issue the internal command to perform the access operation on the memory sub-system.

Abstract: A memory management system includes a physical memory associated with a computing device and a memory manager. The memory manager is configured to manage a shared memory cache as part of a compression of the physical memory using a cache compression algorithm, wherein a compression block size for the compression is a single cache line size. The physical memory includes a sector translation table (STT) region and a sector memory region. The memory manager uses a memory descriptor defined by an STT entry having a cache line map and a plurality of sector pointers to load cache from the physical memory to a level 3 Cache. The cache line map contains cache line metadata including a size of each cache line, a location of the cache line in one of the sectors pointed to by the STT entry, and a plurality of flags.

Abstract: Managing pool memory in a data storage system includes maintaining free lists for corresponding object sizes. For a memory-consuming request (e.g., host write) an allocation operation is performed and the request data is stored. The allocation operation includes (1) selecting a memory object at least as large as the request size and removing all pages of the selected memory object from the corresponding free list, and (2) selecting pages of the selected memory object to store the request data and marking the selected pages as non-free, and leaving any leftover pages as free pages. For a memory-freeing request (e.g., destaging), a deallocation operation is performed that includes (1) marking the request pages free, and (2) based on neighboring pages being free, merging the request pages and neighboring pages into a corresponding memory object and adding the merged pages to the corresponding free list.

Abstract: A gateway for interfacing a host with a subsystem for acting as a work accelerator to the host. The gateway enables the transfer of batches of data to the subsystem at precompiled data exchange synchronisation points. The gateway acts to route data between accelerators which are connected in a scaled system of multiple gateways and accelerators using a global address space set up at compile time of an application to run on the computer system.

Abstract: Techniques for determining a virtual machine configuration based on garbage collection characteristics are disclosed. GC characteristics associated with GC cycles executed on a memory allocated for execution of an application are obtained. A relationship between GC variables is determined based on the GC characteristics collected over a time period of interest. A target GC characteristic is input to the GC variable relationship to determine an expected GC characteristic. The target GC characteristic may be, for example, a target cumulative pause time per a time interval. A virtual machine configuration is determined based on the expected GC characteristic. The virtual machine configuration is applied to a virtual machine executing the application.