Loading patent…
AssigneeDell Products L.P.
InventorsDmitry Vladimirovich Krivenok, Eldad Zinger, Amitai Alkalay
Estimating the term…
A method for distributing a number K network connections from a host computer (host) to a smaller number M of core-affined receive queues (RXQs) of a data storage system (DSS) includes, by the host, allocating source network port identifiers (port IDs) from M identifier-defined groups, and using the allocated port IDs for respective network connections to the DSS. The DSS performs receive flow steering of host network traffic among the RXQs based on rules associating the M groups of port IDs with corresponding ones of the RXQs. In an example, port IDs are allocated and assigned to groups based on a modular arithmetic technique, and the DSS uses a modular operation on the port IDs of received network traffic to direct the traffic to respective cores, promoting balanced core utilization as well as alignment of requests to cores via the RXQs, with system performance benefit.
A method of distributing a first number K of network connections from a host computer to a smaller number M of core-affined receive queues of a data storage system, comprising:
A host computer having a plurality of processing cores and corresponding storage-access queues to access a data storage system using a first number K of network connections to a smaller number M of core-affined receive queues of the data storage system, the host computer being configured and operative, in connection with establishing the network connections, to (1) allocate source network port identifiers from M identifier-defined groups, and (2) use the allocated source network port identifiers for respective distinct network connections to the data storage system for data transfer therewith, wherein allocating source network port identifiers uses a modular arithmetic scheme based on M, defining the groups as source network port identifiers having a corresponding same remainder value.
A data storage system having a number M of processing cores and respective core-affined receive queues for handling host storage requests received on a larger number K of network connections to a host computer, the data storage system being configured and operative to steer received host network traffic among the core-affined receive queues based on a set of rules associating M identifier-defined groups of source port identifiers with corresponding ones of the core-affined receive queues, wherein the source network port identifiers are allocated using a modular arithmetic scheme based on M, defining the groups as source network port identifiers having a corresponding same remainder value.
Dependencies are inferred from the claim text.