Emc IP Holding Company LLC

Dell Products Inc.

Leonid Ravich

Eldad Zinger

Elad Grupi

Alexander Barabash

Xiangping Chen

Ran Anner

Amit Engel

Eyal Ben Ner

Vladimir Shveidel

Anton Kucherov

Amitai Alkalay

Lior Kamran

Svetlana Kronrod

Vitaly Zharkov

Yuri Chernyavsky

Eran Banin

Omer Margolin

Or Idgar

An apparatus includes at least one processing device comprising a processor coupled to a memory, with the processing device being configured to maintain a synchronous replication input-output (IO) request list having a plurality of entries corresponding to respective synchronous replication IO requests, a given such entry identifying at least a sender component and one or more associated component resources to be released responsive to a failure of the sender component. The processing device is further configured to detect a failure of a particular one of a plurality of sender components, to access the synchronous replication IO request list to determine one or more associated component resources to be released, to release the one or more associated component resources, and to update the synchronous replication IO request list by marking the one or more associated component resources as released. Other embodiments include methods and computer program products.

Storage system with efficient release of failed component resources during synchronous replication

A method, computer program product, and computer system for exposing, by a user block layer of a computing device, a storage device as a block device to an application associated with an IO request. A Non-Volatile Memory Express (NVMe) protocol layer of the computing device may expose a NVMe storage device associated with the IO request to the user block layer. A NVMe-over-RDMA layer of the computing device may create a NVMe RDMA queue for the IO request to be processed.

System and method for user space NVMe-over-RDMA host library

A method, computer program product, and computing system for associating a plurality of NVMe IO submission queues with an NVMe IO completion queue. An NVMe IO completion queue polling frequency may be defined for a plurality of threads configured to provide one or more IO requests to the plurality of NVMe IO submission queues. The NVMe IO completion queue may be polled based upon, at least in part, the NVMe IO completion queue polling frequency defined for the plurality of threads.

System and method for processing requests in a multithreaded system

A method, computer program product, and computer system for directing, by a computing device, an incoming Non-Volatile Memory express (NVMe) command to a kernel driver. The kernel driver may enqueue the incoming NVMe command until fetched by a user space. The NVMe command may be fetched from the kernel driver for processing. The NVMe command may be pushed to a user space block device of the user space.

System and method for efficient scalable software-defined NVMEoF front-end connectivity

Techniques for avoiding race conditions in high availability (HA) storage systems that employ disk array enclosures (DAEs). The techniques include receiving a notification of an input/output (IO) error from a storage node, in which the IO error occurs subsequent to the storage node issuing, to a first DAE controller, an IO operation to access data at a logical address of a storage device. The techniques further include, in response to the notification of the IO error, issuing, over an out-of-band channel, a flush command to the first DAE controller, causing data stored in cache memory of the first DAE controller to be written to the storage device, and receiving, over the out-of-band channel, an acknowledgement of completion of the flush operation from the first DAE controller. The techniques further include sending a notification of the completion of the flush operation to the storage node.

Smart disk array enclosure race avoidance in high availability storage systems

A method, computer program product, and computer system for using, by a computing device, free central processing unit (CPU) cycles for processing Non-Volatile Memory express Over Fiber (NVMeoF)/Transmission Control Protocol (TCP) sockets. Using the free CPU cycles for processing the NVMeoF/TCP sockets may include exposing a character device with a callback that iterates over the NVMeoF/TCP sockets for processing.

System and method for boosting up performance of NVMe-TCP transport

A method, computer program product, and computing system for assigning a first set of interrupts for exclusive processing by a first set of central processing units (CPU) cores. A second set of interrupts may be assigned for processing by a second set of CPU cores. The first set of interrupts may be processed using the first set of CPU cores. The second set of interrupts may be converted to a set of polling operations, thus defining a set of converted polling operations. The set of converted polling operations may be processed using the second set of CPU cores.

System and method for polling-based storage command processing

A method of performing a remotely-initiated procedure on a computing device is provided. The method includes (a) receiving, by memory of the computing device, a request from a remote device via remote direct memory access (RDMA); (b) in response to receiving the request, assigning processing of the request to one core of a plurality of processing cores of the computing device, wherein assigning includes the one core receiving a completion signal from a shared completion queue (Shared CQ) of the computing device, the Shared CQ being shared between the plurality of cores; and (c) in response to assigning, performing, by the one core, a procedure described by the request. An apparatus, system, and computer program product for performing a similar method are also provided.

Asymmetric fulfillment of remote procedure calls by multi-core systems

Techniques for providing an RPC-less locking mechanism based on RDMA CAW for a storage cluster with an active-active architecture. The techniques include, in response to receipt at a storage node of a lock request for accessing an LBA of a storage object, synchronizing, between the storage node and a remote node, a state of a lock for the LBA. The synchronizing of the lock state includes locating an ALS object having a field containing the LBA, updating a field of the ALS object to include the storage node, and setting a field of the ALS object to a type of lock specified in the lock request. If the storage node locates the ALS object in an array maintained within visible memory on the remote node, then the storage node updates/sets corresponding fields of the ALS object by performing atomic CAW operations supported by an RDMA channel between the respective nodes.

RPC-less locking mechanism based on RDMA CAW for storage cluster with active-active architecture

A method is disclosed comprising: instantiating, on a first device, a plurality of first connection objects; generating, by the first device, a first set of connection parameters that is associated with the plurality of first connection objects; transmitting the first set of connection parameters from the first device to a second device; receiving from the second device a second set of connection parameters, the second set of connection parameters being associated with a plurality of second connection objects that are instantiated on the second device; updating, by the first device, each of the first connection objects based on the second set of connection parameters; and transmitting, from the first device to the second device, a confirmation that a plurality of communications channels is established.

Efficient establishment on RDMA peer-to-peer connections

Techniques for processing messages may include: establishing a plurality of streams between a source system and a target system; reserving one or more streams of the plurality of streams, wherein the one or more streams are reserved and used in transmitting only management messages, wherein the one or more streams includes a first stream of the plurality of streams; transmitting a first management message over the first stream from the source system to the target system; and transmitting a first data message over a second of the plurality of streams, wherein the second stream is not included in the one or more streams reserved for use in transmitting only management messages.

Techniques for processing management messages using multiple streams

A method is disclosed comprising: instantiating a first plurality of type-1 network sockets on a source computing system, each of the type-1 network sockets having a first queue depth; instantiating a second plurality of type-2 network sockets on the source computing system, each of the type-2 network sockets having a second queue depth that is greater than the first queue depth; transitioning the source computing system into a first state, the first state being one in which data replication messages are transmitted by the source computing system to a target computing system by using both the type-1 and type-2 network sockets; transitioning the source computing system from the first state into a second state, the second state being one in which data replication messages are transmitted by the source computing system to the target computing system by using the type-2 network sockets.

Sockets for shared link applications

Techniques for providing an efficient NTB-based data transport in a highly available storage system. The techniques include performing an ordered transfer of a source data buffer from a primary storage node to a secondary storage node over a PCIe NTB, writing a 'transfer ID value' for the ordered transfer over the NTB to a 'received transfer ID register' of the secondary storage node, performing a remote procedure call (RPC) to send, over a side channel, the transfer ID value as a &#x201c;sent transfer ID value&#x201d; to the secondary storage node, and processing the RPC call to verify the value contained in the received transfer ID register against the sent transfer ID value. Having performed the verification, the secondary storage node determines a successful or unsuccessful status of the ordered transfer, and sends an RPC reply to acknowledge or inform the primary storage node of the successful or unsuccessful status.

Efficient non-transparent bridge (NTB) based data transport

A method, computer program product, and computing system for deploying a kernel driver extension in a kernel of a storage processor communicatively coupled to one or more non-volatile memory express (NVMe) devices. The kernel driver extension may be communicatively coupled with a standard NVMe kernel driver deployed in the kernel of the storage processor. One or more input/output (I/O) operations may be performed on the one or more NVMe devices via the standard NVMe kernel driver and the kernel driver extension.

Growing community of inventors

Yatzitz, Israel

Leonid Ravich