Emc Corporation

Emc IP Holding Company LLC

James L Davidson

Arieh Don

Patrick Brian Riordan

John F Madden, Jr

Chris Bunting

Owen Martin

Alexandr Veprinsky

Hui Wang

Yechiel Yochai

Jeffrey Wilson

William Z Zahavi

Jaeyoo Jung

Scott Rowlands

Igor Fradkin

Alexander Veprinsky

Alan Rajapa

Wolfgang Klinger

Alexander V Dunfey

M Michael Hadavi

Bandwidth consumption for IO paths between a storage system and host may be managed. It may be determined whether there is congestion on a front-end port (FEP) link. For example, the storage system may monitor for a notification from the switch in accordance with a Fibre Channel (FC) protocol. If a notification is received indicating congestion on an FEP link, the bandwidth thresholds (BWTs) for one or more IO paths between the storage system and one or more hosts that include the FEP link may be reduced. The host port BWTs may continue to be reduced until a congestion notification communication has not been received for a predetermined amount of time, in response to which the host port BWTs for one or more host port links on IO paths that include the FEP link may be increased. Similar techniques may be employed for an FEP link determined to be faulty.

Managing IO path bandwidth

Described are techniques for controlling I/O response time. It is determined whether a first observed I/O response time for a storage group exceeds I/O performance specified by a service level objective. If so, first processing may be performed to decrease I/O performance for the storage group. The first processing may include: for each I/O directed to the storage group that is received within a first subsequent time period, determining a first response time denoting an amount of time taken to service each I/O prior to returning a response for each I/O; determining whether the first response time for each I/O is less than a response time delay threshold; and if the first response time for each I/O is less than the response time delay threshold, applying a response time delay to each I/O prior to returning a response for each I/O.

Controlling I/O response time to meet service levels

Controlling I/O operations with a storage device includes establishing a quota that corresponds to a maximum amount of data to store on the storage device in a given amount of time, determining if processing an I/O operation would cause the quota to be exceeded, and performing the I/O operation if the quota is not exceeded. The quota may be provided in I/O operations per second or as I/O throughput. Controlling I/O operations with a storage device may also include accumulating credit in response to a rate of I/O operations being less than the quota and performing I/O operations when the quota is exceeded in response to the credit being greater than zero. The credit may be decreased if an I/O operation is performed when the quota is exceeded.

I/O performance quotas

Described are techniques for redistributing workload. A first port, from which a first workload is to be redistributed, is selected. A second port, having available capacity to perform additional workload, is selected. A device, accessible over both the first port and the second port, is selected. Each of one or more paths including the first port is set to a passive state for the device. Each of the one or more paths having the passive state for the device indicates to a client that sends I/O operations to the device that said each path is not to be used for sending I/O operations to said device and that said each path is available for use in performing control operations directed to the device.

Techniques for workload redistibution

Described are techniques for servicing data requests in a data storage system. One or more data requests from an application executing on the host are received. The data storage system allocates a first amount of resources for use in servicing data requests from the application. Behavior of the host is monitored for occurrences of I/O failure. In accordance with criteria and the occurrences of I/O failure, it is determined whether at least one of the host or the application is in a state of pending failure. If it is determined that at least one of the host or the application is in a state of pending failure, a second amount of resources larger than the first amount are allocated for a time period when servicing data requests from the application. After the time period, data requests from the application are serviced in accordance with the first amount of resources.

Techniques for adjusting priorities associated with servicing requests

Described are techniques for managing a wait queue in a system. A plurality of buckets associated with the wait queue are defined. Each of the plurality of buckets is associated with one of more queue depth values and one or more counters. For each received request for service, a current depth of the wait queue indicating a number of other requests included in the wait queue waiting to be serviced is determined, a bucket in accordance with the current depth of the wait queue is selected and information is recorded by updating said one or more counters of the bucket selected. The received request is placed in the wait queue if there is another request currently being serviced or if there is at least one other request currently in the wait queue.

Techniques for monitoring and managing wait queues

Controlling communication through a communication resource includes allocating a percentage of a maximum bandwidth of the communication resource to a subset of processes that communicate using the resource, determining a delay amount between I/O operations, where the delay amount corresponds to the percentage of bandwidth allocated to the subset of processes, and waiting an amount of time corresponding to the delay amount between I/O operations for the subset of processes. The delay amount may be increased in response to the actual I/O throughput being greater than an amount corresponding to the percentage of the maximum bandwidth allocated to the subset of processes. The delay amount may be decreased in response to the actual I/O throughput being less than an amount corresponding to the percentage of the maximum bandwidth allocated to the subset of processes.

Setting a ceiling for bandwidth used by background tasks in a shared port environment

Growing community of inventors

Tyngsboro, MA, United States of America

James L Davidson