Collective operation management in a parallel computer
US-9807132-B2 · Oct 31, 2017 · US
US10778738B2 · US · B2
| Field | Value |
|---|---|
| Publication number | US-10778738-B2 |
| Application number | US-201515319775-A |
| Country | US |
| Kind code | B2 |
| Filing date | Sep 7, 2015 |
| Priority date | Sep 9, 2014 |
| Publication date | Sep 15, 2020 |
| Grant date | Sep 15, 2020 |
A practical reading order for non-experts. Skip the full description unless you need deep technical detail.
What the patent document calls the invention.
A short plain-language summary of the technical disclosure.
Who owns or filed the patent and who is credited as inventor.
Filing, priority, publication, and grant dates set the timeline.
The legal scope of protection — read this for what is actually claimed.
Technology tags used to group this patent with similar filings.
Prior art links and similar publications in this corpus.
Official abstract text for this publication.
A hybrid network system, a communication method and a network node are provided. The hybrid network system includes a combined network system architecture of a point-to-point network and a broadcast network, the point-to-point network is provided with a plurality of point-to-point channels, and each network node can send a message to a corresponding network node or receive the message from the corresponding network node through the point-to-point channel; the broadcast network is provided with a plurality of broadcast channels, and each broadcast channel can only have one sender at most at the same time, but can have a plurality of receivers; at least a part of network nodes have the capability of sending messages to be broadcast, at least a part of network nodes have the capability of receiving the broadcast messages, and at least some network nodes are configured to select the broadcast network or the point-to-point network at least according to communication tasks to be executed per se to complete the communication tasks to be executed. Preferably, each network node has the capability of sending and receiving the broadcast message. The hybrid network system of the embodiment of the present invention has high flexibility and high communication performance.
Opening claim text (preview).
The invention claimed is: 1. A hybrid network system, comprising: a combined network system architecture including a point-to-point network, a wireless broadcast network, and network nodes in the point-to-point network and the wireless broadcast network, wherein each of the network nodes includes one or more processors, and the wireless broadcast network is realized by one of WIFI, a millimeter wave technology, and a visible light network, wherein the point-to-point network is provided with a plurality of point-to-point channels, each of the network nodes being operable to send a message to a corresponding network node or receive a message from the corresponding network node through a point-to-point channel, wherein the wireless broadcast network is provided with a plurality of broadcast channels, and a broadcast channel has one single sender at a time and has a plurality of receivers, and wherein each of the network nodes has capability of sending messages to be broadcast through the broadcast channels, and each of the network nodes has capability of receiving the broadcast messages, wherein each of the network nodes is configured to select one of the broadcast channels at least according to a first communication task to be executed and broadcast data associated with the first communication task through the selected broadcast channel, and to select one of the point-to-point channels at least according to a second communication task to be executed and send data associated with the second communication task through the selected point-to-point channel, wherein the hybrid network system is applied to high performance parallel computation, wherein each of the network nodes is used as a computing node in a parallel computer system, and the computing node executes a collective operation that includes identifying a type of the collective operation called by the parallel computer system, and selecting the point-to-point network and/or the wireless broadcast network according to the type of the collective operation to execute the collective operation, and wherein when the type of the collective operation is MPI_Barrier, MPI_Gather, or MPI_Reduce, the point-to-point network is selected to execute the collective operation, and when the type of the collective operation is MPI_Bcast or MPI_Scatter, the wireless broadcast network is selected to execute the collective operation. 2. The hybrid network system of claim 1 , wherein each of the network nodes is configured to select a channel between the broadcast channels and the point-to-point channels at least according to a communication task to be executed and complete the communication task to be executed through the selected channel. 3. The hybrid network system of claim 1 , wherein the point-to-point network is realized by Infiniband or the Ethernet. 4. The hybrid network system of claim 1 , wherein the point-to-point network is used as a channel for the wireless broadcast network to coordinate and allocate information. 5. The hybrid network system of claim 1 , wherein each of the network nodes is further configured to determine whether to select the wireless broadcast network to complete a part of broadcast or multicast communication according to communication traffic and a network use condition of the wireless broadcast network. 6. The hybrid network system of claim 5 , wherein each of the network nodes is further configured to determine whether to select the wireless broadcast network to complete the part of broadcast or multicast communication according to the communication traffic and the network use condition of the wireless broadcast network by: when carrying out a certain operation capable of adopting a broadcast algorithm, and when the wireless broadcast network is being occupied by other operations, determining whether to wait for the wireless broadcast network to be released or immediately communicate through the point-to-point network according to an occupation condition of the wireless broadcast network and/or an evaluation of benefits being obtained by the certain operation from the wireless broadcast network. 7. The hybrid network system of claim 5 , wherein each of the network nodes is further configured to determine whether to select the wireless broadcast network to complete the part of broadcast or multicast communication according to the communication traffic and the network use condition of the wireless broadcast network by: when carrying out a certain operation capable of adopting the broadcast algorithm, and when a storage space volume of the network node to carry out a broadcast operation does not satisfy requirements of an operation on the storage space volume, adopting the point-to-point network to realize the certain operation. 8. The hybrid network system of claim 1 , wherein the selecting the point-to-point network and/or the wireless broadcast network according to the type of the collective operation to execute the collective operation comprises: for the collective operation only involving point-to-point communication, selecting the point-to-point network, for the collective operation only involving broadcast or multicast communication, selecting the wireless broadcast network, and for the collective operation involving both of point-to-point communication and broadcast or multicast communication, selecting the point-to-point network to realize the point-to-point communication and selecting the wireless broadcast network to realize the broadcast or multicast communication. 9. The hybrid network system of claim 8 , wherein: when the type of the collective operation is MPI_Allgather, MPI_Alltoall, or MPI_Allreduce, the collective operation is decomposed into MPI_Bcast and other collective operations, wherein the wireless broadcast network is selected to execute the collective operation that includes the decomposed MPI_Bcast. 10. A communication method in a hybrid network system, wherein the hybrid network system comprises: a combined network system architecture including a point-to-point network, a wireless broadcast network, and network nodes in the point-to-point network and the wireless broadcast network, wherein each of the network nodes includes one or more processors, and the wireless broadcast network is realized by one of WIFI, a millimeter wave technology, and a visible light network, wherein the point-to-point network is provided with a plurality of point-to-point channels, each of the network nodes being operable to send a message to a corresponding network node or receive a message from the corresponding network node through a point-to-point channel, and wherein the wireless broadcast network is provided with a plurality of broadcast channels, and a broadcast channel has one single sender at a time and has a plurality of receivers, and wherein each of the network nodes has capability of sending messages to be broadcast through the broadcast channels, and each of the network nodes has capability of receiving the broadcast messages, and wherein the communication method comprises: configuring each of the network nodes to select one of the broadcast channels at least according to a first communication task to be executed and to broadcast data associated with the first communication task through the selected broadcast channel; and configuring each of the network nodes to select one of the point-to-point channels at least according to a second communication task to be executed and to send data associated with the second communication task through the selected point-to-point channel, wherein the hybrid network system is applied to high performance parallel computation, wherein each of the network nodes is used as a computing node in a
with heterogeneous network architecture · CPC title
using dynamic host configuration protocol [DHCP] or bootstrap protocol [BOOTP] · CPC title
for multicast or broadcast (systems for broadcast or conference H04L12/18; arrangements for broadcast or distribution combined with broadcast H04H20/00; arrangements for broadcast applications with a direct linkage to broadcast information or to broadcast space-time H04H60/00; selective distribution of broadcast services, e.g. multimedia broadcast multicast service [MBMS], H04W4/06) · CPC title
Replication or mirroring of data, e.g. scheduling or transport for data synchronisation between network nodes · CPC title
based on web technology, e.g. hypertext transfer protocol [HTTP] · CPC title
Related publications grouped by family.
Answers are generated from the same data shown on this page.