Multi-core storage processor assigning other cores to process requests of core-affined streams

• US 9,485,310 B1
• Filed: 12/23/2014
• Issued: 11/01/2016
• Est. Priority Date: 12/23/2014
• Status: Active Grant

• Current Assignee: Emc IP Holding Company LLC (Dell Technologies Inc.)
• Original Assignee: Emc IP Holding Company LLC (Dell Technologies Inc.)
• Inventors: Gupta, Mukesh, Bono, Jean-Pierre, Forecast, John, Corniquet, Frederic, Armangau, Philippe
• Primary Examiner(s): Lim, Krisna
• Application Number:
  US14/580,865
• Time in Patent Office: 679 Days
• US Class Current: 1/1
• CPC Class Codes: G06F 16/1824 implemented using Network-attached Storage [NAS] architecture distributed or networked storage systems G06F3/067; protocols for distributed storage of data in a network H04L67/1097
• CPC Class Codes: H04L 67/1097 for distributed storage of data in networks, e.g. transport arrangements for network file system [NFS], storage area networks [SAN] or network attached storage [NAS]

Abstract

A multi-core processor of a network attached storage system processes requests from host computers for services of a file system service. Each core maintains endpoints of respective connection-layer connections to the hosts to affine respective streams of network traffic with the core, and dynamically and preferentially assigns execution threads of the core to process file system service requests of the streams affined with the core. Each core also co-operates with the other cores to dynamically and non-preferentially (a) assign execution threads of the core to process file system service requests of the streams affined with the other cores, and (b) assign execution threads of the other cores to process file system service requests of the streams affined with the core, promoting efficient use of the cores for the processing workload of the file system service.

Claims

1. A network attached storage system for providing a distributed file system service to host computers via a network, comprising:

network interface circuitry to be coupled to the network;

one or more data storage devices to store file system data; and

a storage processor coupled to the network interface circuitry and the data storage devices to perform file system operations of the distributed file system service using the data storage devices for underlying data storage, the file system operations including file read and file write operations,

the storage processor including a multi-core processing unit and memory, the multi-core processing unit having a set of cores for respective independent instruction execution of instructions stored in the memory, the instructions including;

(1) first instructions of a distributed file system application including respective sets of execution threads for the cores, each execution thread used to perform file system operations based on corresponding requests from the host computers,

(2) second instructions of sets of network stream modules forming respective stream network stacks, sets of two or more of the stream network stacks being executed by respective cores, each stream network stack including a stream head to and from which the execution threads read and write stream data corresponding to the requests from and responses to the host computers via respective distinct host-specific connection-layer connections, and

(3) third instructions of collector modules executed by the respective cores, a collector module of each core being operative on a per-request basis to (a) dynamically and preferentially assign execution threads of the core to the stream heads of the core to process first file system requests and responses in stream data of the respective stream network stacks, and (b) co-operate with collector modules of the other cores to dynamically and non-preferentially assign execution threads of the other cores to the stream heads of the core to process second file system requests and responses in stream data of the respective stream network stacks to promote efficient use of the cores in providing the distributed file system service.

2. A network attached storage system according to claim 1, wherein preferentially assigning execution threads of a core to respective stream heads of the core is performed when at least one of the execution threads of the core is idle, and wherein non-preferentially assigning execution threads of a core to respective stream heads of another core is performed when all execution threads of the core are busy and the other core has at least one execution thread that is idle.

3. A network attached storage system according to claim 1, wherein the collectors maintain respective queues of requests received from respective stream network stacks, entries of the queues identifying respective local stream heads which are to be read from to obtain corresponding file system service requests received by the respective stream network stacks, the entries of each queue having execution threads assigned thereto in sequential order.

4. A network attached storage system according to claim 3, wherein the collectors apply respective thresholds against the number of entries in the queues to determine when to direct a next request to another core rather than assigning a local execution thread to process the request.

5. A network attached storage system according to claim 1, wherein, for a request obtained from the stream head of a given core being processed by an execution thread of another core, the execution thread of the other core returns a corresponding response to the stream head of the given core, and the stream network stack of the given core returns the response to the host issuing the request.

6. A network attached storage system according to claim 1, wherein the stream network stacks of a core are operative to call a function of the collector of the core once well-formed file system requests are identified in received network traffic, and wherein the function of the collector responds by (1) adding an entry to a queue of requests, (2) selecting an execution thread for handling a queued request, and (3) activating the selected execution thread, and wherein an activated execution thread (4) obtains a next entry from the queue which identifies a corresponding stream head, (5) obtains a well-formed request by performing a read stream function on the identified stream head, (6) performs file system processing for the well-formed request, and (7) returns a corresponding response by performing a write stream function to the same stream head.

7. A network attached storage system according to claim 1, wherein the execution threads of the cores form respective pools of execution threads available for processing the requests, the execution threads transitioning from idle to in-use upon having stream heads assigned thereto for processing respective requests, the execution threads transition from in-use to idle upon completing the processing for respective requests including returning respective responses to the stream heads for delivery to respective host computers.

8. A method of operating a multi-core processor of a network attached storage system to process requests from host computers for services of a file system service, the method including, at each of a set of cores of the multi-core processor:

maintaining endpoints of respective connection-layer connections to the hosts to affine respective streams of network traffic with the core;

dynamically and preferentially assign execution threads of the core to process file system service requests of the streams affined with the core; and

co-operate with other cores of the set of cores to dynamically and non-preferentially (a) assign execution threads of the core to process file system service requests of the streams affined with the other cores, and (b) assign execution threads of the other cores to process file system service requests of the streams affined with the core.

9. A method according to claim 8, wherein preferentially assigning execution threads of a core to respective stream heads of the core is performed when at least one of the execution threads of the core is idle, and wherein non-preferentially assigning execution threads of a core to respective stream heads of another core is performed when all execution threads of the core are busy and the other core has at least one execution thread that is idle.

10. A method according to claim 8, wherein the collectors maintain respective queues of requests received from respective stream network stacks, entries of the queues identifying respective local stream heads which are to be read from to obtain corresponding file system service requests received by the respective stream network stacks, the entries of each queue having execution threads assigned thereto in sequential order.

11. A method according to claim 10, wherein the collectors apply respective thresholds against the number of entries in the queues to determine when to direct a next request to another core rather than assigning a local execution thread to process the request.

12. A method according to claim 8, wherein, for a request obtained from the stream head of a given core being processed by an execution thread of another core, the execution thread of the other core returns a corresponding response to the stream head of the given core, and the stream network stack of the given core returns the response to the host issuing the request.

13. A method according to claim 8, wherein the stream network stacks of a core call a function of the collector of the core once well-formed file system requests are identified in received network traffic, and wherein the function of the collector responds by (1) adding an entry to a queue of requests, (2) selecting an execution thread for handling a queued request, and (3) activating the selected execution thread, and wherein an activated execution thread (4) obtains a next entry from the queue which identifies a corresponding stream head, (5) obtains a well-formed request by performing a read stream function on the identified stream head, (6) performs file system processing for the well-formed request, and (7) returns a corresponding response by performing a write stream function to the same stream head.

14. A method according to claim 8, wherein the execution threads of the cores form respective pools of execution threads available for processing the requests, the execution threads transitioning from idle to in-use upon having stream heads assigned thereto for processing respective requests, the execution threads transition from in-use to idle upon completing the processing for respective requests including returning respective responses to the stream heads for delivery to respective host computers.