Address resource mapping in a shared memory computer system
View Patent ↗An algorithm for mapping memory and a method for using a high performance computing (“HPC”) system are disclosed. The algorithm takes into account the number of physical nodes in the HPC system, and the amount of memory in each node. Some of the nodes in the HPC system also include input/output (“I/O”) devices like graphics cards and non-volatile storage interfaces that have on-board memory; the algorithm also accounts for the number of such nodes and the amount of I/O memory they each contain. The algorithm maximizes certain parameters in priority order, including the number of mapped nodes, the number of mapped I/O nodes, the amount of mapped I/O memory, and the total amount of mapped memory.
1. A computer system having hardware-based non-uniform memory access, the computer system comprising a plurality of nodes, each node having a volatile memory and at least one computing processor, each computing processor being capable of addressing a maximum amount of volatile memory that is less than the collective amount of volatile memory in the plurality of nodes, wherein at least one node also includes an I/O device having I/O memory, each node being configured to calculate a memory layout that includes a portion of the volatile memory of each of a plurality of nodes in the computer system, the memory layout being calculated to minimize an amount of unmapped memory as a function of 1) the number of nodes in the computer system, 2) the amount of volatile memory in each node, 3) the number of nodes that include an I/O device having I/O memory, and 4) the amount of I/O memory in each such I/O device, wherein the computing processor of each node is further configured to access the memory of another node using the calculated memory layout.
2. The computer system of claim 1 , wherein the maximum amount of volatile memory is at least 16 terabytes.
3. The computer system of claim 1 , wherein the I/O device includes a network card, a graphics card, a non-volatile storage device, a coprocessor, or an accelerator card.
4. The computer system of claim 1 , wherein calculating the memory layout comprises minimizing the number of unmapped nodes.
5. The computer system of claim 4 , wherein calculating the memory layout further comprises minimizing the number of unmapped nodes that include an I/O device after minimizing the number of unmapped nodes.
6. The computer system of claim 5 , wherein calculating the memory layout further comprises maximizing the amount of mapped I/O memory after minimizing the number of unmapped nodes that include an I/O device.
7. The computer system of claim 6 , wherein calculating the memory layout further comprises maximizing the amount of mapped volatile memory after maximizing the amount of mapped I/O memory.
8. A method of operating a computer system having hardware-based non-uniform memory access, the computer system having a plurality of nodes, each node having a volatile memory and at least one computing processor, each computing processor being capable of addressing a maximum amount of volatile memory that is less than the collective amount of volatile memory in the plurality of nodes, wherein at least one node also includes an I/O device having I/O memory, the method comprising:
in each node of the computer system, calculating a memory layout that includes a portion of the volatile memory of each of a plurality of nodes in the computer system, the memory layout being calculated to minimize an amount of unmapped memory as a function of 1) the number of nodes in the computer system, 2) the amount of volatile memory in each node, 3) the number of nodes that include an I/O device having I/O memory, and 4) the amount of I/O memory in each such I/O device; and
accessing, by a computing processor in a first node of the computer system, the memory of a second node of the computer system using the calculated memory layout.
9. The method of claim 8 , wherein the maximum amount of volatile memory is at least 16 terabytes.
10. The method of claim 8 , wherein the I/O device includes a network card, a graphics card, a non-volatile storage device, a coprocessor, or an accelerator card.
11. The method of claim 8 , wherein calculating the memory layout comprises minimizing the number of unmapped nodes.
12. The method of claim 10 , wherein calculating the memory layout further comprises minimizing the number of unmapped nodes that include an I/O device after minimizing the number of unmapped nodes.
13. The method of claim 12 , wherein calculating the memory layout further comprises maximizing the amount of mapped I/O memory after minimizing the number of unmapped nodes that include an I/O device.
14. The method of claim 13 , wherein calculating the memory layout further comprises maximizing the amount of mapped volatile memory after maximizing the amount of mapped I/O memory.
15. A computer program product for operating a computer system having hardware-based non-uniform memory access, the computer system having a plurality of nodes, each node having a volatile memory and at least one computing processor, each computing processor being capable of addressing a maximum amount of volatile memory that is less than the collective amount of volatile memory in the plurality of nodes, wherein at least one node also includes an I/O device having I/O memory, the computer program product comprising a non-transitory, tangible computer-readable storage medium on which is stored non-transitory program code for:
instructing each node of the computer system to calculate a memory layout that includes a portion of the volatile memory of each of a plurality of nodes in the computer system, the memory layout being calculated to minimize an amount of unmapped memory as a function of 1) the number of nodes in the computer system, 2) the amount of volatile memory in each node, 3) the number of nodes that include an I/O device having I/O memory, and 4) the amount of I/O memory in each such I/O device; and
instructing a computing processor in a first node of the computer system to access the memory of a second node of the computer system using the calculated memory layout.
16. The product of claim 15 , wherein the maximum amount of volatile memory is at least 16 terabytes.
17. The product of claim 15 , wherein the I/O device includes a network card, a graphics card, a non-volatile storage device, a coprocessor, or an accelerator card.
18. The product of claim 15 , wherein the program code for calculating the memory layout comprises program code for minimizing the number of unmapped nodes.
19. The product of claim 18 , wherein the program code for calculating the memory layout further comprises program code for minimizing the number of unmapped nodes that include an I/O device after minimizing the number of unmapped nodes.
20. The product of claim 19 , wherein the program code for calculating the memory layout further comprises program code for maximizing the amount of mapped I/O memory after minimizing the number of unmapped nodes that include an I/O device.
21. The product of claim 20 , wherein the program code for calculating the memory layout further comprises program code for maximizing the amount of mapped volatile memory after maximizing the amount of mapped I/O memory.