Router parking in power-efficient interconnect architectures

US9619006B2 · US · B2

Patent metadata
FieldValue
Publication numberUS-9619006-B2
Application numberUS-201213994780-A
CountryUS
Kind codeB2
Filing dateJan 10, 2012
Priority dateJan 10, 2012
Publication dateApr 11, 2017
Grant dateApr 11, 2017

How to read this patent

A practical reading order for non-experts. Skip the full description unless you need deep technical detail.

  1. Title

    What the patent document calls the invention.

  2. Abstract

    A short plain-language summary of the technical disclosure.

  3. Assignees and inventors

    Who owns or filed the patent and who is credited as inventor.

  4. Key dates

    Filing, priority, publication, and grant dates set the timeline.

  5. First independent claim

    The legal scope of protection — read this for what is actually claimed.

  6. CPC / IPC classifications

    Technology tags used to group this patent with similar filings.

  7. Citations and related patents

    Prior art links and similar publications in this corpus.

Abstract

Official abstract text for this publication.

A method and apparatus for selectively parking routers used for routing traffic in mesh interconnects. Various router parking (RP) algorithms are disclosed, including an aggressive RP algorithm where a minimum number of routers are kept active to ensure adequate network connectivity between active nodes and/or intercommunicating nodes, leading to a maximum reduction in static power consumption, and a conservative RP algorithm that favors network latency considerations over static power consumption while also reducing power. An adaptive RP algorithm is also disclosed that implements aspects of the aggressive and conservative RP algorithms to balance power consumption and latency considerations in response to ongoing node utilization and associated traffic. The techniques may be implemented in internal network structures, such as for single chip computers, as well as external network structures, such as computing clusters and massively parallel computer architectures. Performance modeling has demonstrated substantial power reduction may be obtained using the router parking techniques while maintaining Quality of Service performance objectives.

First claim

Opening claim text (preview).

What is claimed is: 1. A method, comprising: employing routers to facilitate transmission of data between a plurality of nodes connected via a plurality of interconnect links, each router associated with a respective node; detecting nodes among the plurality of nodes that are inactive nodes; selectively parking routers associated with a portion of the inactive nodes while operating other routers associated with the inactive nodes in an active state by, determining an operational state of each of the plurality of nodes; for each node determined to be in an inactive state, identifying its associated router as a potential candidate for parking; determining if there are any network partitions corresponding to a router configuration under which each potential candidate for parking would be parked; and, if one or more network partitions exist in the configuration; and selecting routers from among the potential candidates to activate or keep active to remove the one or more network partitions. 2. The method of claim 1 , further comprising: determining intercommunicating groups of nodes for which communication between nodes in an intercommunicating group is to be maintained; and selecting routers to be parked such that there are no network partitions within an intercommunicating group. 3. The method of claim 1 , wherein the plurality of nodes comprise processing elements on a semiconductor die. 4. The method of claim 1 , further comprising selecting routers to be parked based, in part, on power consumption considerations. 5. The method of claim 1 , further comprising selecting routers to be parked based, in part, on Quality of Service considerations. 6. The method of claim 1 , further comprising dynamically reconfiguring an activation state of the plurality of routers on an ongoing basis in response to run time interconnect utilization. 7. The method of claim 1 , further comprising selecting addition routers from among the potential candidates to activate or keep active based on latency considerations. 8. The method of claim 1 , further comprising employing a router parking algorithm on an ongoing basis to reconfigure the operating status of selected routers in response to input parameters including power consumption and latency inputs. 9. The method of claim 1 , further comprising parking a router by changing the state of the router from an active state to a parked state by power gating ports of the router, an arbiter, and crossbar circuitry associated with the router. 10. The method of claim 1 , further comprising implementing changes in router configuration for a network reconfiguration over multiple cycles to avoid undelivered packets. 11. A semiconductor apparatus, comprising: a plurality of processing nodes, each operatively coupled to a respective co-located router; an interconnect, comprising a plurality of links interconnecting the plurality of routers; and embedded logic, configured to perform operations comprising, acquire node status information indicating an active or inactive operating status of each of the plurality of processing nodes; for each processing node in an inactive state, identify its associated router as a potential candidate for parking; determine if there are any network partitions corresponding to a router configuration under which each potential candidate for parking would be parked; and, if one or more network partitions exist in the configuration; and select routers from among the potential candidates to activate or keep active to remove the one or more network partitions. 12. The apparatus of claim 11 , wherein the configuration of the mesh interconnect after the routers are parked ensures there are no network partitions between processing nodes having an active status. 13. The apparatus of claim 11 , further comprising a fabric manager comprising embedded logic for implementing management of the plurality of routers, wherein the logic is configured to enable the fabric manager to perform operations when the apparatus is operating comprising: determining a current network configuration including an operational state of each of the plurality of processing nodes; for each processing node in an inactive state, identifying its co-located router as a potential candidate for parking; determining a reconfiguration of the network to be implemented; and sending configuration information to effect the reconfiguration of the network to the plurality of routers. 14. The apparatus of claim 13 , wherein each of the routers is configured to at least one of forward or return operational status information to the fabric manager. 15. The apparatus of claim 13 , wherein parking a router is effected by changing the state of the router from an active state to a parked state by power gating ports of the router, an arbiter, and crossbar circuitry associated with the router. 16. A system, comprising: a plurality of processing nodes, each operatively coupled to a respective co-located router; an interconnect, comprising a plurality of links interconnecting the plurality of routers; and embedded logic comprising a central router management entity, communicatively coupled to each of the processing nodes via the plurality of links, and configured to perform operations comprising, acquire node status information indicating an active or inactive operating status of each of the plurality of processing nodes; for each processing node in an inactive state, identify its associated router as a potential candidate for parking; determine if there are any network partitions corresponding to a router configuration under which each potential candidate for parking would be parked; and, if one or more network partitions exist in the configuration; and select routers from among the potential candidates to activate or keep active to remove the one or more network partitions. 17. The system of claim 16 , wherein the processing nodes comprise a plurality of processors interconnect via a plurality of wired and/or optical links. 18. The system of claim 16 , wherein the processing nodes are configured in three dimensions, and the mesh interconnect comprises a 3D mesh interconnect. 19. The system of claim 16 , wherein the central router management entity is configured to reconfigure the router configuration on any ongoing basis in view of node utilization and associated interconnect traffic.

Assignees

Inventors

Classifications

  • G06F1/325Primary

    Power saving in peripheral device · CPC title

  • H04L45/06Primary

    Deflection routing, e.g. hot-potato routing · CPC title

  • based on intermediate node capabilities · CPC title

Patent family

Related publications grouped by family.

External sources

Frequently asked questions

Answers are generated from the same data shown on this page.

What does patent US9619006B2 cover?
A method and apparatus for selectively parking routers used for routing traffic in mesh interconnects. Various router parking (RP) algorithms are disclosed, including an aggressive RP algorithm where a minimum number of routers are kept active to ensure adequate network connectivity between active nodes and/or intercommunicating nodes, leading to a maximum reduction in static power consumption,…
Who is the assignee on this patent?
Samih Ahmad, Wang Ren, Maciocco Christian, and 2 more
What technology area does this patent fall under?
Primary CPC classification G06F1/325. Mapped technology areas include Physics.
When was this patent published?
Publication date Tue Apr 11 2017 00:00:00 GMT+0000 (Coordinated Universal Time) (B2). Legal status and post-grant events are not shown on this page.
What related patents are in patentsdb?
We list 8 related publications on this page (citations in our corpus or others sharing the same primary CPC).