Multi-threaded transmit transport engine for storage devices

• US 8,149,854 B2
• Filed: 06/30/2005
• Issued: 04/03/2012
• Est. Priority Date: 06/30/2005
• Status: Expired due to Fees

• Current Assignee: Intel Corporation
• Original Assignee: Intel Corporation
• Inventors: Lau, Victor, Seto, Pak-lung
• Primary Examiner(s): Moe, Aung S
• Assistant Examiner(s): PASIA, REDENTOR M
• Application Number:
  US11/174,197
• Publication Number:
  US 20070019636A1
• Time in Patent Office: 2,469 Days
• Field of Search: 370/389, 370/392, 370/401, 370/412, 370/471, 370/474, 709212-219, 710/22, 710/29, 710/30, 710/52
• US Class Current: 370/412
• CPC Class Codes: G06F 13/385 for adaptation of a particular data processing system to different peripheral devices

Abstract

An embodiment of the present invention is a technique to process a plurality of I/O sequences associated with a storage device. A task context pre-fetch engine pre-fetches a task context from a task context memory based on a pre-fetch request. At least a multi-threaded transmit transport layer (T×TL) processes the plurality of I/O sequences from an I/O pool simultaneously. The multi-threaded T×TL generates the pre-fetch request and one or more frames from the plurality of I/O sequences. A switch fabric and controller routes the frame to a link layer associated with the storage device.

Claims

1. An apparatus comprising:

a task context pre-fetch engine to pre-fetch a first and a second task context from a task context memory based on a first and a second pre-fetch request, respectively, wherein the first and second task contexts each include context information for accessing one of a plurality of storage devices remotely coupled via a network external to the apparatus;

a multi-threaded transmit transport layer (T×

TL) coupled to the task context pre-fetch engine, including a first transport layer to process a first input/output (I/O) sequence from a pool of I/O sequences based on the first task context to generate a first transmit data sequence, and a second transport layer to process a second I/O sequence based on the second task context from the pool of I/O sequences to generate a second transmit data sequence, wherein processing the second I/O sequence is associated with a latency period and each I/O sequence of the pool of I/O sequences includes I/O operations for reading and writing data to a different storage device of the plurality of storage devices; and

a switch fabric/controller coupled to the multi-threaded T×

TL to route the first transmit data sequence to a single link layer interface during the latency period and the second transmit data sequence to the single link layer interface using the switch fabric/controller.

2. The apparatus of claim 1 wherein the multi-threaded T×

TL comprises;

a transmit transport protocol engine (PE) to perform I/O operations for reading and writing data;

a frame queue coupled to the transmit transport PE to store frame information;

a direct memory access (DMA) engine coupled to the transmit transport PE to transfer data from a memory to a transmit data buffer based on a DMA descriptor; and

a framer coupled to the frame queue and the transmit data buffer to create the frame based on the frame information and to generate a frame transmit request to the switch fabric and controller.

3. The apparatus of claim 2 wherein the transmit transport PE comprises:

an I/O requester to request for an I/O sequence from an I/O sequence scheduler;

a pre-fetch requester to generate the pre-fetch request to the task context pre-fetch engine;

an I/O decision generator to generate an I/O handling decision based on the first and second task context;

a DMA descriptor generator to create the DMA descriptor; and

a task context updater to update the first and second task context.

4. The apparatus of claim 3 wherein the I/O decision generator generates a decision to transmit a data frame and a decision to transmit a command frame.

5. The apparatus of claim 2 wherein the frame queue comprises:

a frame header queue to store a frame header for the respective transmit data; and

a frame data queue to store the respective transmit data.

6. The apparatus of claim 2 wherein the switch fabric/controller comprises:

a switch to connect the single link layer interface and the multi-threaded T×

TL;

a link layer monitor to check link layer status to a remote node having a remote node address; and

a controller to control the switch using the flame transmit request and to transmit the flame to the remote node via the switch according to the link layer status and the remote node address.

7. A method comprising:

pre-fetching a first and a second task context from a task context memory based on a first and a second pre-fetch request, respectively, using a task context pre-fetch engine, wherein the first and second task contexts each include context information for accessing one of a plurality of storage devices remotely coupled via a network;

processing a first input/output (I/O) sequence from a pool of I/O sequences using a first transport layer of a multi-threaded transmit transport layer (T×

TL) based on the first task context to generate a first transmit data sequence, wherein each I/O sequence of the pool of I/O sequences includes I/O operations for reading and writing data to a different remote storage device of the plurality of storage devices;

processing a second I/O sequence from the pool of I/O sequences via a second transport layer thread of the multi-threaded T×

TL based on the second task context to generate a second transmit data sequence, wherein processing the second I/O sequence is associated with a latency period;

routing the first transmit data sequence to a single link layer interface using a switch fabric/controller during the latency period; and

routing the second transmit data sequence to the single link layer interface using the switch fabric/controller.

8. The method of claim 7 wherein processing the first and second of I/O sequences comprises:

storing frame information;

transferring data from a memory to a transmit data buffer based on a DMA descriptor;

creating one or more frames based on the frame information; and

generating a frame transmit request to the switch fabric/controller.

9. The method of claim 8 wherein the I/O operations for reading and writing data to different remote storage devices of the plurality of storage devices comprises:

requesting for an I/O sequence from an I/O sequence scheduler;

generating the pre-fetch request to the task context pre-fetch engine;

generating an I/O handling decision based on the first and second task context;

creating the DMA descriptor; and

updating the first and second task context.

10. The method of claim 9 wherein generating the I/O handling decision comprises generating a decision to transmit a data frame and a decision to transmit a command frame.

11. The method of claim 8 wherein storing the frame information comprises:

storing a frame header for the respective transmit data sequence; and

storing the respective transmit data sequence.

12. The method of claim 8 wherein routing the frame comprises:

connecting the single link layer interface to the multi-threaded T×

TLs using a switch;

checking link layer status to a remote node having a remote node address;

controlling the switch using the frame transmit request; and

transmitting the frame to the remote node via the switch according to the connection status and the remote node address.

13. A system comprising:

a plurality of storage device interfaces to interface to a plurality of storage devices remotely coupled via a network;

an input/output sequence scheduler coupled to the storage device interfaces to schedule processing first input/output (I/O) sequence from a pool of I/O sequences associated with the storage devices and based on a first task context to generate a first transmit data sequence, and a second transport layer to process a second I/O sequence from the pool of I/O sequences based on the second context, wherein each I/O sequence of the I/O pool includes I/O operations for reading and writing data to a different storage device of the plurality of storage devices; and

a multi-threaded transmit transport engine (TTE) coupled to the storage device interfaces and the I/O sequence scheduler, the multi-threaded TTE comprising;

a task context pre-fetch engine to pre-fetch a first and the second task context from a task context memory based on a first and a second pre-fetch request, respectively, wherein the first and second task contexts each include context information for accessing one of a plurality of storage devices remotely coupled via the network;

a multi-threaded transmit transport layer (T×

TL) coupled to the task context pre-fetch engine, including a first transport layer to process the first I/O sequence to generate a first transmit data sequence, and the second transport layer to process the second I/O sequence to generate a second transmit data sequence, wherein processing the second I/O sequence is associated with a latency period; and

a switch fabric/controller coupled to the multi-threaded T×

TLs to route the first transmit data sequence during the latency period and the second transmit data sequence to a single link layer interface.

14. The system of claim 13 wherein each of the multi-threaded T×

TLs comprises;

a transmit transport protocol engine (PE) to perform I/O operations for reading and writing data;

a frame queue coupled to the transmit transport PE to store frame information;

a direct memory access (DMA) engine coupled to the transmit transport PE to transfer data from a memory to a transmit data buffer based on a DMA descriptor; and

a framer coupled to the frame queue and the transmit data buffer to create the frame based on the frame information and to generate a frame transmit request to the switch fabric and controller.

15. The system of claim 14 wherein the transmit transport PE comprises:

an I/O requester to request for an I/O sequence from the I/O sequence scheduler;

a pre-fetch requester to generate at least one of the pre-fetch requests to the task context pre-fetch engine;

an I/O decision generator to generate an I/O handling decision based on the first and second task context;

a DMA descriptor generator to create the DMA descriptor; and

a task context updater to update the first and second task context.

16. The system of claim 15 wherein the I/O decision generator generates a decision to transmit a data frame and a decision to transmit a command frame.

17. The system of claim 14 wherein the frame queue comprises:

a frame header queue to store a frame header for the transmit data; and

a frame data queue to store the respective transmit data.

18. The system of claim 14 wherein the switch fabric controller comprises:

a switch to connect the single link layer interface and the multi-threaded T×

TL;

a link layer monitor to check link layer statuses to remote nodes having remote node addresses; and

a controller to control the switch using the frame transmit requests and to transmit the frames to the remote nodes via the switch according to the link layer status and the remote node address.

19. The system of claim 13 wherein the I/O sequence scheduler comprises:

an assigner to assign one of the I/O sequences to the multi-threaded T×

TL based on a link layer status; and

a counter to keep track of number of threads executing for a remote node.

20. The system of claim 13 wherein the storage devices include at least one of a Serial Attached Small Computer System Interface (SAS) device, a Fiber Channel (FC) device, a Serial Advanced Technology Attachment (SATA) device, an Internet Small Computer System Interface (ISCSI) device, and a Host Bus Adapter (HBA) device.