Synchronization method and apparatus for an interconnection network using parallel-headerless TDMA routing
Embodiments of the present disclosure generally relate to a cloud computing network, or datacenter network, and a method of transferring information among processing nodes in a cloud computing network or datacenter. The network may include a hub that is coupled to a plurality of nodes so that data is transferred between nodes through the hub. Data from different nodes may be written into a slot within the hub, read, and then written into a slot within the destination node. Due to the proximity of the nodes to the hub, or even due to the amount of data to be written, the data may be written at different clock phases. The read may occur one or more clock cycles after the data has been written into the hub.
1. A datacenter, comprising:
one or more racks; and
a plurality of servers disposed in each rack, wherein
each server includes a hub and a plurality of processing nodes,
wherein each processing node of a given server is coupled to a corresponding hub of the server,
wherein the plurality of processing nodes includes a first node spaced a first distance from the hub and a second node spaced a second distance from the hub,
wherein the second distance is greater than the first distance,
wherein the server contains logic to delay communication of data by a predetermined amount at the hub from being routed to an intended node, and
wherein reading the data occurs with a delay of at least one clock cycle after the data has been written to the hub.
2. The datacenter of claim 1 , wherein the hub comprises:
circular memory units, where circular memory units further comprise
a writing address pointer;
a read address pointer; and
a data location corresponding to each node, wherein the writing address pointer is coupled to each data location and wherein the read address pointer is connected to each data location.
3. The datacenter of claim 2 , wherein the circular memory further comprises a clock write line that is connected to each data location.
4. The datacenter of claim 3 , wherein the circular memory further comprises a data in line that is connected to each data location.
5. The datacenter of claim 4 , wherein the circular memory further comprises a clock read line that is connected to each data location.
6. The datacenter of claim 5 , wherein the circular memory further comprises a data out line that is connected to each data location.
7. The datacenter of claim 6 , wherein the write address pointer is connected to each data location with a corresponding write enable line.
8. The datacenter of claim 7 , wherein the read address pointer is connected to each data location with a corresponding read enable line.
9. The datacenter of claim 1 , wherein each node includes a plurality of memory locations for data storage, and wherein the number of memory locations for each node is equal to one less than a total number of nodes.
10. A method of transferring data between nodes on a motherboard, wherein the motherboard includes a plurality of nodes and a hub, wherein each node of the plurality of nodes is coupled to the hub, and hub comprising circular memory comprising:
writing first data from a first node into a first memory position in the circular memory, wherein the first data is written at a first write clock phase, wherein the writing occurs for a first write clock time period;
writing second data from a second node into a second memory position in the circular memory, wherein the second data is written at a second write clock phase that is less than the first write phase, wherein the writing occurs for a second write time period;
reading data from the first memory position at a first read clock phase; and
reading data from the second memory position at a second read clock phase, wherein the second read clock phase and the first read clock phase are substantially identical.
11. The method of claim 10 , wherein the first read clock phase is equal to or less than the first write clock phase.
12. The method of claim 11 , wherein reading data from the first memory locations is occurs after a writing the first data into the first memory locations, wherein the reading is delayed from the writing by a first time period, wherein the first time period is equal to or greater than the first write time period.
13. The method of claim 12 , wherein reading data from the second memory locations is occurs after a writing the second data into the second memory locations, wherein the reading is delayed from the writing by a second time period, wherein the second time period is equal to or greater than the second write time period.
14. The method of claim 13 , wherein the second time period and the first time period are different.
15. The method of claim 14 , wherein the first write time period is greater than the second write time period.
16. The method of claim 15 , wherein the first node is spaced a first distance from the hub, wherein the second node is spaced a second distance from the hub and wherein the first distance is greater than the second distance.
17. The method of claim 10 , wherein the first node is spaced a first distance from the hub, wherein the second node is spaced a second distance from the hub and wherein the first distance is greater than the second distance.
18. A method of transferring data between nodes on a motherboard, wherein the motherboard includes a plurality of nodes and a hub, wherein each node of the plurality of nodes is coupled to the hub, comprising:
writing first data from a first node into first memory locations in the circular memory;
writing second data from a second node into second memory locations in the circular memory;
reading data from the first memory locations, wherein the reading occurs simultaneous with writing the second data from the second node into the second memory locations;
writing third data from a third node into a third memory location in the circular memory;
reading data from the second memory location, wherein the reading occurs simultaneous with writing the third data from the third node into the third memory location; and
writing the data from the first slot into a memory location in the second node, wherein the writing occurs simultaneous with writing the third data from the third node into the third memory location.
19. The method of claim 18 , wherein writing the first data from the first node in the first memory location occurs for a time period that is longer than writing the second data from the second node into the second memory location.
20. The method of claim 18 , wherein reading data from the first slot occurs for a time period that is shorter than writing the first data from the first node in the first memory location.