DETAILED ACTION
Notice of Pre-AIA or AIA Status
The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
Response to Arguments
Applicants’ arguments filed 07/13/2026 have been fully considered but they are not persuasive.
Argument 1: (Summary of pages 6 and 7, Examiner’s emphasis - bold)
A. The Claims Require That the Hop Immediately Following Receipt on a Terminal Link Exit the Source VRG on a Global Link.
Applicant argues that claims are directed to hop ordering at ingress. When a packet arrives at a switch from a compute node, on a terminal link, and its destination lies in another VRG, the very next hop taken by that packet exits the source VRG over a global link. The packet is not first redistributed across local links within the source VRG; the global hop comes first. Anticipation accordingly requires a disclosure of when the global link is taken, not merely a disclosure that packets in a Dragonfly network eventually traverse global links.
Response:
Examiner respectfully disagrees.
Consistent with applicant’s disclosure, Froese in [0025] similarly discloses a Dragonfly network topology, where the network is subdivided into groups, with each group containing some number of switches and some number of endpoints. Each switch with many interconnected ports. The ports of each switch are divided between edge ports, local ports, and global ports, Edge ports are where traffic enters and leaves the network. The ports connected to processing nodes/terminals are edge ports. Local ports connect to other switches within the same group. Global ports connect to other switches that are in a different group. In a fault-free Dragonfly network, every group is connected to every other group in the network by one or more global links and every switch within a group is connected to every other switch in the same group by one or more local links. Furthermore, similar to the invention, Froese in [0026] discloses that when a packet enters the network at one group (the source group) and leaves the network at a different group the destination group), the packet may be globally routed (routed between its source and destination group) either minimally or non-minimally. The packet takes a global minimal path if it traverses one global link, directly connecting the source group to the destination group. That means, the packet enters the first/source group through a terminal/edge link and the next hop for the packet is a global link that is directly connected to the destination group. That means, the packet is not first redistributed across local links within the source VRG; instead, the packet’s next hop is a global hop/link that directly connects the source group to the destination group. Therefore, when a packet arrives at a switch/enters the network at a compute node through an edge/terminal link, and its destination lies in another VRG, the very next hop taken by that packet exits the source VRG over a global link as required by the claim.
Argument 2: (Summary of pages 7 and 8, Examiner’s emphasis - bold)
B. The Rejection's Mapping of the Claimed "Terminal Link" to Froese's Local Links Is Erroneous.
The rejection maps the receiving limitation as follows: "receiving, on a terminal link (local link) of a switch, (switch in a local group) a packet (a frame) whose destination is in another VRG (global group)." That mapping conflicts with the claim language, with Froese, and with the Office Action itself. The claims define a terminal link as a link "to compute nodes." Froese distinguishes its port types in precisely the same way: "The ports connected to processing nodes are edge ports. Local ports connect to other switches within the same group. Global ports connect to other switches that are in a different group.
Response:
Examiner respectfully disagrees.
See updated rejection of claim 1 and detailed response to argument 1.
Furthermore, similar to the invention, in [0025] Froese discloses Edge ports where traffic enters and leaves a network. The ports connected to processing nodes are edge ports. Local ports connect to other switches within the same group. Global ports connect to other switches that are in a different group. In a fault-free Dragonfly network, every group is connected to every other group in the network by one or more global links and every switch within a group is connected to every other switch in the same group by one or more local links. [0025-0026] packets destined for a second VRG from a first VRG, may enter the first VRG through an edge port of the first VRG and globally routed (routed between its source and destination group) either minimally or non-minimally. The packet takes a global minimal path if it traverses one global link, directly connecting the source group to the destination group. That means, the packet is not first redistributed across local links within the source VRG; instead, the global hop comes first. This routing technique is consistent with the technique claimed by the invention.
Argument 3: (Summary of pages 8 and 9, Examiner’s emphasis - bold)
C. Froese's "Intermediate Group" Is Not the Claimed "Next Hop"
The Response to Arguments equates Froese's randomly chosen intermediate group with the claimed next hop: "This other group is chosen at random and is termed the intermediate group (next-hop)." The rejection's claim mapping likewise glosses "the next hop" as "(intermediary/intermediate group)." Office Action at 6. Respectfully, a hop and a group are not the same thing. A hop is a single link traversal from one switch (or compute node) to the next switch.
Response:
Examiner respectfully disagrees.
See updated rejection of claim 1 and detailed response to arguments 1 & 2.
Furthermore, Froese already disclose an embodiment in which a source group directly connected to a destination group via a terminal/edge link of a source group receives a packet to enter a network. The packet is then transmitted through a next hope global link to the destination group. Accordingly, Froese disclosure above is not intended to limit the implementation of the invention. Therefore, in another embodiment, an intermediate group is disclosed that relays packets from a source group to a destination group via global links without traversing local link in the source and intermediate groups. The routing technique is still consistent with applicant’s claimed invention where each next hop is a global link and the packet never traverses a local link between the source and destination groups. Therefore, Froese discloses the routing technique of the invention as claimed.
Argument 4 : (Summary of page 9, Examiner’s emphasis - bold)
D. Froese Paragraph [0173] Describes Intra-Group Routing and Confirms the Absence of the Claimed Behavior
The Response to Arguments characterizes paragraph [0173] as disclosing "a frame received at local group that is transmitted to a global group using local non-minimal paths," such that "the frame was transmitted to a next-hop/switch in an intermediary group." Paragraph [0173] discloses no such thing.
Response:
Examiner respectfully disagrees.
See updated rejection of claim 1 and detailed responses to arguments 1-3.
Argument 5 : (Summary of pages 9 & 10, Examiner’s emphasis - bold)
E. The Remaining Citations Do Not Supply the Missing Disclosure
The rejection's remaining citations are no closer. Paragraph [0033] describes flow-ID assignment and per-hop flow-ID updates; how flow identifiers are updated says nothing about which link is selected for the hop following terminal ingress. The citation of I [0064] for the proposition that links "may be virtual channels" is likewise immaterial: implementing links as virtual channels discloses no rule for choosing between a local link and a global link at ingress. Figures 1, 2, and 9 depict the Dragonfly topology itself; topology is not routing behavior.
For claim 19, the rejection additionally cites [0184]. Claim 19 requires "routing, by a source switch on the first hop, a packet to a switch in another VRG on a global link if the packet is received on the terminal link of the source switch."
Response:
Examiner respectfully disagrees.
See updated rejection of claim 19 and detailed responses to arguments 1-3.
Argument 6 : (Summary of pages 9 & 10, Examiner’s emphasis - bold)
F. Inherency Cannot Rescue the Rejection
To the extent the rejection reasons that packets in Froese must eventually cross a global link to reach another group, that observation cannot establish anticipation. The claims recite when the global link is taken: on the next hop following receipt on a terminal link (claims 1, 7, and 13), and on the first hop by the source switch (claim 19).
Response:
Examiner respectfully disagrees.
See updated rejection of claim 19 and detailed responses to arguments 1-3.
Argument 7 : (Summary of pages 9 & 10, Examiner’s emphasis - bold)
G. The Dependent Claims Are Separately Allowable
Dependent claims 2, 4-6, 8, 10-12, 14, and 16-18 are allowable at least by virtue of their dependence from allowable base claims. As to claims 2, 8, and 14, the rejection's own citation underscores the deficiency: it relies on 1 [0028]'s statement that "a packet arriving at an edge port might be routed across a local link to another switch in the same group." That is an express description of a local, intra-VRG next hop, not of "transmitting the packet on a port on the switch connected to a global link based on local routing decisions that include use of the global link on the next hop," as those claims require.
For at least the foregoing reasons, Froese does not disclose each and every limitation of any pending claim, arranged as claimed. Applicant respectfully requests withdrawal of the rejection of claims 1-2, 4-8, 10-14, and 16-19 under 35 U.S.C. §102(a)(1).
See updated rejection of claim(s) 1 & 19 and detailed responses to arguments 1-3.
Claim Rejections - 35 USC § 102
In the event the determination of the status of the application as subject to AIA 35 U.S.C. 102 and 103 (or as subject to pre-AIA 35 U.S.C. 102 and 103) is incorrect, any correction of the statutory basis (i.e., changing from AIA to pre-AIA ) for the rejection will not be considered a new ground of rejection if the prior art relied upon, and the rationale supporting the rejection, would be the same under either status.
The following is a quotation of the appropriate paragraphs of 35 U.S.C. 102 that form the basis for the rejections under this section made in this Office action:
A person shall be entitled to a patent unless –
(a)(1) the claimed invention was patented, described in a printed publication, or in public use, on sale, or otherwise available to the public before the effective filing date of the claimed invention.
Claim(s) 1-2, 4-8,10-14, and 16 - 19 is/are rejected under 35 U.S.C. 102(a)(1) as being anticipated by Froese (US 2022/0166705 A1).
Regarding claim 1, Froese discloses a method of global first, non-minimal routing in a Dragonfly topology (Froese, fig. 9, [0022; 0025-0026] discloses Dragonfly network topology, the network is subdivided into groups, with each group containing some number of switches and some number of endpoints. In a preferred system, each switch is an ASIC with many interconnected ports. The ports of each switch are divided between edge ports, local ports, and global ports, Edge ports are where traffic enters and leaves the network. When a packet enters the network at one group (the source group) and leaves the network at a different group the destination group), the packet may be globally routed (routed between its source and destination group) either minimally or non-minimally),
the Dragonfly topology comprising a plurality of interconnected virtual routing groups (VRGs) (Froese, fig. 9, [0025] With the Dragonfly network topology, the network is subdivided into groups, with each group containing some number of switches and some number of endpoints. Each switch is an ASIC with many interconnected ports. The ports of each switch are divided between edge ports, local ports, and global ports. The ports connected to processing nodes are edge ports. Local ports connect to other switches within the same group. Global ports connect to other switches that are in a different group. In a fault-free Dragonfly network, every group is connected to every other group in the network by one or more global links and every switch within a group is connected to every other switch in the same group by one or more local links. [0064] each link in this network may be virtual channels(VC)),
wherein each VRG includes a plurality of interconnected switches (switches within the same group) (Froese, figs. 1&9, [0025] With the Dragonfly network topology, the network is subdivided into groups, with each group containing some number of switches and some number of endpoints. Each switch is an ASIC with many interconnected ports. The ports of each switch are divided between edge ports, local ports, and global ports. The ports connected to processing nodes are edge ports. Local ports connect to other switches within the same group. Global ports connect to other switches that are in a different group. In a fault-free Dragonfly network, every group is connected to every other group in the network by one or more global links and every switch within a group is connected to every other switch in the same group by one or more local links),
wherein the switches include terminal links to compute nodes (endpoints), local links to other switches in the same VRG (Froese, figs. 2&9, [0025] With the Dragonfly network topology, the network is subdivided into groups, with each group containing some number of switches and some number of endpoints. Each switch is an ASIC with many interconnected ports. The ports of each switch are divided between edge ports, local ports, and global ports. The ports connected to processing nodes are edge ports. Local ports connect to other switches within the same group. Global ports connect to other switches that are in a different group. In a fault-free Dragonfly network, every group is connected to every other group in the network by one or more global links and every switch within a group is connected to every other switch in the same group by one or more local links), and
global links to switches in other VRGs (Froese, fig. 9, [0025] With the Dragonfly network topology, the network is subdivided into groups, with each group containing some number of switches and some number of endpoints. Each switch is an ASIC with many interconnected ports. The ports of each switch are divided between edge ports, local ports, and global ports. The ports connected to processing nodes are edge ports. Local ports connect to other switches within the same group. Global In a fault-free Dragonfly network, every group is connected to every other group in the network by one or more global links and every switch within a group is connected to every other switch in the same group by one or more local links); and
the method comprises:
receiving, on a terminal link (local link) of a switch,(switch in a local group) a packet (a frame) whose destination is in another VRG (global group), routing, by the switch, the packet on the next hop (intermediary/intermediate group) to a switch in another VRG (global group) on a global link (Global link) (Froese [0026; 0173] when a packet enters the network at one group (the source group) and leaves the network at a different group the destination group), the packet may be globally routed (routed between its source and destination group) either minimally or non-minimally. The packet takes a global minimal path if it traverses one global link, directly connecting the source group to the destination group. With global non-minimal routing, some packets leaving the source group are routed to a group other than the destination group. This other group is chosen at random and is termed the intermediate group (next-hop). On reaching an intermediate group, the packet is then routed directly to the destination group using a global link that directly connects the intermediate group to the destination group. [0026] when a packet enters the network at one group (the source group) and leaves the network at a different group the destination group), the packet may be globally routed (routed between its source and destination group) either minimally or non-minimally. The packet takes a global minimal path if it traverses one global link, directly connecting the source group to the destination group. That means, the packet enters the first/source group through a terminal/edge link and the next hop for the packet is a global link that is directly connected to the destination group. That means, the packet is not first redistributed across local links within the source VRG; instead, the packet’s next hop is a global hop/link that directly connects the source group to the destination group. [0173] further discloses a frame received at local group that is transmitted to a global group using local non-minimal paths by taking two switch-to-switch hops, from the source switch to an intermediate switch, known as the Root switch, and from there to the destination switch. That is, the frame/packet is received at a switch in a local group, using a local link, the frame is transmitted to a next-hop/switch in an intermediary group, and from there to a switch in a destination group).
Regarding claim 2, Froese discloses the method of claim 1 wherein routing, by the switch, the packet to a switch in another VRG on a global link further comprises transmitting the packet on a port on the switch connected to a global link based on local routing decisions that include use of the global link on the next hop (Froese [0026;0028;0033] a packet arriving at an edge port might be routed across a local link to another switch in the same group which has some global ports the packet can use. When a packet enters the network at one group (the source group) and leaves the network at a different group the destination group), the packet may be globally routed (routed between its source and destination group) either minimally or non-minimally. The packet takes a global minimal path if it traverses one global link (next hop), directly connecting the source group to the destination group).
Regarding claim 4, Froese discloses the method of claim 1 wherein routing, by the switch, the packet to a switch in another VRG on a global link further comprises transmitting the packet on a port on the switch connected to a global link (Froese [0026;0028] a packet arriving at an edge port might be routed across a local link to another switch in the same group which has some global ports the packet can use. The packet takes a global minimal path if it traverses one global link, directly connecting the source group to the destination group. With global non-minimal routing, some packets leaving the source group are routed to a group other than the destination group. This other group is chosen at random and is termed the intermediate group).
Regarding claim 5, Froese discloses the method of claim | wherein routing, by the switch, the packet to a switch in another VRG on a global link further comprises routing the packet to the destination VRG (Froese [0026;0028] a packet arriving at an edge port might be routed across a local link to another switch in the same group which has some global ports the packet can use. The packet takes a global minimal path if it traverses one global link, directly connecting the source group to the destination group. With global non-minimal routing, some packets leaving the source group are routed to a group other than the destination group. This other group is chosen at random and is termed the intermediate group).
Regarding claim 6, Froese discloses the method of claim | wherein routing, by the switch, the packet to a switch in another VRG on a global link further comprises routing the packet to a pass-through VRG (intermediate group) (Froese [0026;0028] a packet arriving at an edge port might be routed across a local link to another switch in the same group which has some global ports the packet can use. With global non-minimal routing, some packets leaving the source group are routed to a group other than the destination group. This other group is chosen at random and is termed the intermediate group).
Regarding claim(s) 7-8 and 10 – 12, the claim(s) are rejected with rational similar to that of claim(s) 1-2 and 4-6, respectively.
Regarding claim(s) 13-14 and 16-18, the claim(s) are rejected with rational similar to that of claim(s) 1-2 and 4-6, respectively.
Regarding claim 19, Froese discloses a routing algorithm for non-minimal routing in a Dragonfly topology (Froese [0184] discloses Dragonfly routing algorithms allow a non-minimal path in both the intermediate and destination groups),
The Dragonfly topology comprising a plurality of interconnected virtual routing groups (VRGs) (i.e., the Dragonfly network topology, the network is subdivided into groups, with each group containing some number of switches and some number of endpoints [0025]),
wherein each VRG includes a plurality of interconnected switches (fig. 1),
wherein the switches include terminal links to compute nodes (fig. 2), local links to other switches in the same VRG, and global links to switches in other VRGs (Froese, fig. 9, [0025-0026] With the Dragonfly network topology, the network is subdivided into groups, with each group containing some number of switches and some number of endpoints. Each switch is an ASIC with many interconnected ports. The ports of each switch are divided between edge ports, local ports, and global ports. The ports connected to processing nodes are edge ports. Local ports connect to other switches within the same group. Global ports connect to other switches that are in a different group. In a fault-free Dragonfly network, every group is connected to every other group in the network by one or more global links and every switch within a group is connected to every other switch in the same group by one or more local links. [0064] each link in this network may be virtual channels(VC)),
the routing algorithm comprising routing, by a source switch on the first hop, a packet to a switch in another VRG on a global link if the packet is received on the terminal link of the source switch (Froese [0026-0028] When a packet enters the network at one group (the source group) and leaves the network at a different group the destination group), the packet may be globally routed (routed between its source and destination group) either minimally or non-minimally. The packet takes a global minimal path if it traverses one global link (global next hop), directly connecting the source group to the destination group. When a packet enters the network at one group (the source group) and leaves the network at a different group the destination group), the packet may be globally routed (routed between its source and destination group) either minimally or non-minimally. The packet takes a global minimal path if it traverses one global link, directly connecting the source group to the destination group. That means, the packet enters the first/source group through a terminal/edge link and the next hop for the packet is a global link that is directly connected to the destination group. That means, the packet is not first redistributed across local links within the source VRG; instead, the packet’s next hop is a global hop/link that directly connects the source group to the destination group).
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure. The following publications show the state of the art related to non-minimal routing in dragonfly topologies.
Kim et al. (US 2010/0049942 A1)
Parker et al. (US 2012/0144065 A1)
THIS ACTION IS MADE FINAL. Applicant is reminded of the extension of time policy as set forth in 37 CFR 1.136(a).
A shortened statutory period for reply to this final action is set to expire THREE MONTHS from the mailing date of this action. In the event a first reply is filed within TWO MONTHS of the mailing date of this final action and the advisory action is not mailed until after the end of the THREE-MONTH shortened statutory period, then the shortened statutory period will expire on the date the advisory action is mailed, and any nonprovisional extension fee (37 CFR 1.17(a)) pursuant to 37 CFR 1.136(a) will be calculated from the mailing date of the advisory action. In no event, however, will the statutory period for reply expire later than SIX MONTHS from the mailing date of this final action.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to DIXON F DABIPI whose telephone number is (571)270-3673. The examiner can normally be reached on Monday - Friday from 9:00 am – 5:00 pm.
If attempts to reach the examiner by telephone are unsuccessful, the examiner’s supervisor, Christopher L Parry, can be reached at telephone number 571-272-8328. The fax phone number for the organization where this application or proceeding is assigned is 571-273-8300.
Information regarding the status of an application may be obtained from Patent Center. Status information for published applications may be obtained from Patent Center. Status information for unpublished applications is available through Patent Center to authorized users only. Should you have questions about access to the USPTO patent electronic filing system, contact the Electronic Business Center (EBC) at 866-217-9197 (toll-free).
Examiner interviews are available via a variety of formats. See MPEP § 713.01. To schedule an interview, applicant is encouraged to use the USPTO Automated Interview Request (AIR) Form at https://www.uspto.gov/InterviewPractice.
/D.F.D/Examiner, Art Unit 2451
/Chris Parry/Supervisory Patent Examiner, Art Unit 2451
Prosecution Insights