Samsung Electronics Co., Ltd.

Zhengyu Yang

T David Evans

Clay Mayers

Thomas David Evans

Jiayin Wang

Nithya Ramakrishnan

Allen Russell Andrews

Manu Awasthi

Sudheendra Grama Sampath

Morteza Hoseinzadeh

Ping Terence Wong

Vijay Balakrishnan

Mrinmoy Ghosh

Venkatraman Balasubramanian

Olufogorehan Adetayo Tunde-Onadele

Thomas B Bolt

Thomas R Bolt

Sina Hassani

Daniel Song Kim

Thomas Edward Rainey, Iii

Bridget Molly Davis

Adnan Maruf

John Artoux

Michael Kurt Gehlen

Jeffrey Chun Hung Wong

Terence Ping Wong

David Evans

A method for dispatching input-output in a system. The system may include a centralized processing circuit, a plurality of persistent storage targets, a first input-output processor, and a second input-output processor. The method may include determining whether the first input-output processor is connected to a first target of the plurality of persistent storage targets; determining whether the second input-output processor is connected to the first target; and in response to determining that both the first input-output processor is connected to the first target, and the second input-output processor is connected to the first target, dispatching a first plurality of input-output requests, each to either the first input-output processor or the second input-output processor, the dispatching being in proportion to a service rate of the first input-output processor to the first target and a service rate of the second input-output processor to the first target, respectively.

Intelligent path selection and load balancing

A method of data storage includes determining a latency distance from a primary node to each of two or more replica nodes, choosing a preferred replica node of the two or more replica nodes based on the determined latency distances, and write-caching data into the preferred replica node.

Automatic data replica manager in distributed caching and data processing systems

A method of operating a storage system may include allocating a first partition of a tier of storage resources to a first client, wherein the tier operates at least partially as a storage cache, allocating a second partition of the tier of the storage resources to a second client, monitoring a workload of the first client, monitoring a workload of the second client, and reallocating the first partition of the tier of the storage resources to the first client based on the monitored workload of the first client and the monitored workload of the second client. The method may further include reallocating the second partition of the tier of the storage resources to the second client based on the monitored workload of the first client and the monitored workload of the second client.

Worflow-based partition allocation

A method of adapting a first key-value store to a second key-value store may include determining a conversion strategy based on one or more characteristics of the first key-value store and one or more characteristics of the second key-value store, converting the second key-value store to a converted key-value store based on the conversion strategy, and mapping the first key-value store to the converted key-value store based on a mapping function. The converted key-value store may be accessed on-the-fly. A data storage system may include a key-value interface configured to provide access to a lower key-value store, and a key-value adapter coupled to the key-value interface and configured to adapt an upper key-value store to the lower key-value store, wherein the key-value adapter may be configured to adapt at least two different types of the upper key-value store to the lower key-value store.

Methods and systems for adapting multiple key-value stores

An apparatus, a method, a method of manufacturing an apparatus, and a method of constructing an integrated circuit are provided. A processor of an application server layer detects a degree of a change in a workload in an input/output stream received through a network from one or more user devices. The processor determines a degree range, from a plurality of preset degree ranges, that the degree of the change in the workload is within. The processor determines a distribution strategy, from among a plurality of distribution strategies, to distribute the workload across one or more of a plurality of solid state devices (SSDs) in a performance cache tier of a centralized multi-tier storage pool, based on the determined degree range. The processor distributes the workload across the one or more of the plurality of solid state devices based on the determined distribution strategy.

Method and apparatus for adaptive cache load balancing for SSD-based cloud computing storage system

A load balancing system includes: a centralized queue; a pool of resource nodes connected to the centralized queue; one or more processors; and memory coupled to the one or more processors and storing instructions that, when executed by the one or more processors, cause the one or more processors to: monitor a queue status of the centralized queue to identify a bursty traffic period; calculate an index value for a load associated with the bursty traffic period; select a load balancing strategy based on the index value; distribute the load to the pool of resource nodes based on the load balancing strategy; observe a state of the pool of resource nodes in response to the load balancing strategy; calculate a reward based on the observed state; and adjust the load balancing strategy based on the reward.

Systems and methods for spike detection and load balancing resource management

A system and method for managing input output queue pairs. In some embodiments, the method includes calculating a system utilization ratio, the system utilization ratio being a ratio of: an arrival rate of input output requests, to a service rate; determining whether: the system utilization ratio has exceeded a first threshold utilization during a time period exceeding a first threshold length, and adding a new queue pair is expected to improve system performance; and in response to determining: that the system utilization ratio has exceeded the first threshold utilization during a time period exceeding the first threshold length, and that adding a new queue pair is expected to improve system performance: adding a new queue pair.

Systems and methods for I/O transmissions in queue pair-based NVMeoF initiator-target system

An a method for simulating a data center is provided and a non-transitory computer-readable storage medium having recorded thereon a computer program for executing the method of simulating a data center. The method includes storing at least one hardware configuration file and at least one functional description file of a data center to be simulated in a configuration file application; generating a simulation program of the data center using the at least one hardware configuration file and the at least one functional description file by a data center storage evaluation framework (DCEF) application; and executing a flow-based simulation on the simulation program generated by the DCEF application by a simulator.

Method for data center storage evaluation framework simulation

A non-volatile memory (NVM) express (NVMe) system includes at least one user application, an NVMe controller and a hypervisor. Each user application runs in a respective virtual machine environment and including a user input/output (I/O) queue. The NVMe controller is coupled to at least one NVM storage device, and the NVMe controller includes a driver that includes at least one device queue. The hypervisor is coupled to the user I/O queue of each user application and to the NVMe controller, and selectively forces each user I/O queue to empty to a corresponding device queue in the driver of the NVMe controller or enables a private I/O channel between the user I/O queue and a corresponding device queue in the driver of the NVMe controller.

Hybrid framework of NVMe-based storage system in cloud computing environment

A system for performing auto-tiering is disclosed. The system may include a plurality of storage devices offering a plurality of resources and organized into storage tiers. The storage devices may store data for virtual machines. A receiver may receive I/O commands and performance data for the virtual machines. A transmitter may transmit responses to the I/O commands. An auto-tiering controller may select storage tiers to store the data for the virtual machines and may migrate data between storage tiers responsive to the performance data. The selection of the storage tiers may optimize the performance of all virtual machines across all storage tiers, factoring the change in performance of the virtual machines and a migration cost to migrate data between storage tiers.

Automatic data placement manager in multi-tier all-flash datacenter

A method for allocating workloads based on a total cost of ownership (TCO) model includes receiving a workload; estimating a cost for allocating the workload to each disk of disks in a disk pool based on a TCO model; determining a disk among the disks in the disk pool that minimizes a TCO; and allocating the workload to the disk. The TCO model incorporates a plurality of cost factors for estimating costs for each disk in the disk pool for allocating the workload.

Online flash resource allocation manager based on a TCO model

A system is disclosed. The system may include a virtual machine server, which may include a processor, a memory, and at least two virtual machines that may be stored in the memory and executed by the processor. The virtual machine server may also include a virtual machine hypervisor to manage the operations of the virtual machine. The virtual machine server may also include a cache that may include at least one storage device. A Dynamic Cache Partition Manager (DCPM) may manage the partition of the cache into a performance guarantee zone, which may be partitioned into regions, and a spike buffer zone. The DCPM may include a reusability statistic calculator to calculate a reusability statistic, a comparator to compare the reusability statistic with a spike threshold, an access statistic calculator to calculate an access statistic, a spike data identifier to identify data in the spike buffer zone that satisfies the access statistic, and a cache partitioner to repartition the cache into a new performance guarantee zone and a second spike buffer zone for a new epoch, the second spike buffer zone is large enough to store the identified data.

Growing community of inventors

Malden, MA, United States of America

Zhengyu Yang