METHOD AND APPARATUS FOR ESTABLISHING A QUALITY OF SERVICE MODEL

• US 20100211935A1
• Filed: 02/16/2010
• Published: 08/19/2010
• Est. Priority Date: 10/31/2003
• Status: Active Grant

• Current Assignee: META Platforms Technologies LLC (Meta Platforms, Inc. (f/k/a Facebook, Inc.))
• Original Assignee: Sonics Incorporated (Meta Platforms, Inc. (f/k/a Facebook, Inc.))
• Inventors: Weber, Wolf-Dietrich, Wingard, Drew E., Chou, Chien-Chun
• Granted Patent: US 8,504,992 B2
• Time in Patent Office: Days
• US Class Current: 717/127
• CPC Class Codes: H04L 47/10 Flow control; Congestion control
• CPC Class Codes: H04L 47/2425 for supporting services specification, e.g. SLA
• CPC Class Codes: H04L 47/70 Admission control; Resource allocation
• CPC Class Codes: H04L 47/805 QOS or priority aware
• CPC Class Codes: H04L 47/821 Prioritising resource allocation or reservation requests
• CPC Class Codes: H04L 47/827 Aggregation of resource allocation or reservation requests

Abstract

In general, methods and apparatus for implementing a Quality of Service (QoS) model are disclosed. A Quality of Service (QoS) contract with an initiating network device may be satisfied. A request may be received from the initiating network device in a first time less than or equal to an ordinal number times an arrival interval. The ordinal number signifies a position of the request among a group of requests. The request that has been serviced may be returned to the initiator in a second time less than or equal to a constant term plus the ordinal number times a service interval.

Claims

1. A method for satisfying a Quality of Service (QoS) contract with an initiator, comprising:

receiving a request from an initiator in a first time less than or equal to an ordinal number times an arrival interval, wherein the ordinal number signifies a position of the request among a group of requests; and

returning the request that has been serviced to the initiator in a second time less than or equal to a constant term plus the ordinal number times a service interval.

2. The method of claim 1, further comprising:

measuring the first and second time at a boundary between the initiator and an interconnect.

3. The method of claim 2, further comprising:

satisfying the QoS contract using the initiator, the target, the interconnect, and a QoS unit.

4. The method of claim 1, wherein the service interval is greater than the arrival interval.

5. The method of claim 1, wherein the service interval is equal to the arrival interval.

6. The method of claim 1, further comprising:

determining the constant term specifically for the group of requests.

7. A system, comprising:

an interconnect coupled between an initiator and a target; and

a logic coupled to the interconnect configured to satisfy a service model if the initiator satisfies an arrival model, wherein the arrival model is satisfied by the initiator if a request arrives before a first time less than or equal to an ordinal number times an arrival interval, wherein the ordinal number signifies a position of the request among a group of requests; and

wherein the logic satisfies the service model if the request is serviced before a second time less than or equal to a constant term plus the ordinal number times a service interval.

8. The system of claim 7, wherein the service interval is greater than the arrival interval.

9. The system of claim 7, wherein the constant term is determined specifically for the group of requests.

10. The system of claim 7, wherein the logic includes a Quality of Service (QoS) unit to guide the interconnect.

11. The system of claim 10, wherein the QoS unit is coupled between the target and the interconnect.

12. The system of claim 7, wherein the arrival model and the service model comprise a QoS contract.

13. The system of claim 12, wherein the QoS contract is satisfied at a boundary between the initiator and the interconnect.

14. The system of claim 10, wherein the QoS unit is part of the interconnect.

15. The system of claim 10, wherein the QoS unit is part of the target.

16. A method for tracking service of components in an Integrated Circuit, comprising:

maintaining a first allocation count for a first transaction to track whether the first transaction is being serviced by the target in the Integrated Circuit;

establishing an adjustable positive limit having a first value for the first allocation count that establishes an initial maximum amount the first allocation count may go up to;

counting a number of cycles that a given transaction has been waiting when the first transaction has yet to have a request fulfilled from the target during a regular interval of time to guarantee a maximum latency is at least met as required by a Quality of Service contract for a component initiating that transaction; and

using spare capacity in components of the Integrated Circuit system when all QoS contracts are currently being met for the components in the Integrated Circuit system to let some components in the system use this spare capacity to exceed their target rate, where allocation count is correlateable to establish priority for servicing specific transaction threads.

17. The method of claim 16, further comprising:

raising the adjustable positive limit for the first allocation count to a second value when the first value for the first allocation count has been reached and the first transaction has yet to have a request fulfilled from the target during a regular interval of time;

reducing the positive limit from the second value to a third value when the second allocation count receives a credit and is positive, wherein the positive limit is not reduced below the first value;

crediting the first and second allocation count at a regular interval;

debiting the first allocation count when the first thread is serviced;

and debiting the second allocation count when the second thread is serviced.

18. The method of claim 16, further comprising:

counting the number of cycles that a given transaction has been waiting to guarantee the maximum latency is at least met as required by their QoS contract, when this number of counted cycles exceeds a limit, the priority of the transaction is in essence promoted to a higher level in the system.

19. The method of claim 17, wherein an allocation count associated with each transaction is initially set to a low adjustable positive limit for the first allocation count to achieve the performance defined by the component initiating the transaction'"'"'s QoS contract and an upper maximum limit is set to the allocation count values for each component in the Integrated Circuit to enforce the QoS contract priority throughout the system.

20. The method of claim 17, further comprising:

determining whether the first thread will be serviced according to the first allocation count; and

servicing the first thread instead of the second thread when the second allocation count is less than the first allocation count, wherein the second thread is a high-priority thread and the first thread is a bandwidth-allocation thread.