Network computing systems having shared memory clouds with addresses of disk-read-only memories mapped into processor address spaces
View Patent ↗Network computing systems are disclosed including a shared memory cloud coupled to one or more processor complexes. The shared memory cloud has an interconnect network coupled to disk-read-only-memories (disk-ROMs) each including a memory array that is read/write block accessible to access blocks of consecutive memory locations and random read memory accessible to access random memory locations. The processor complexes read and write blocks of data from/to the disk-ROMs to provide disk-like access to the shared memory cloud. Each processor complex maps the addresses of one or more of the disk-ROMs into processor address spaces, and reads from random memory locations of one or more of the disk-ROMs to provide main memory-like access to the shared memory cloud. The network computing systems may further include a power controller coupled to the processor complexes. The power controller can keep the disk-ROMS powered on while it powers off inactive processor complexes.
1. A network computing system comprising:
a shared memory cloud including a plurality of disk-read-only-memories (disk-ROMs) coupled to an interconnect network suitable for memory read operations and block read/write operations, each of the disk-read-only-memories including a memory array that is read/write block accessible to access blocks of consecutive memory locations and random read memory accessible to access random memory locations;
one or more processor complexes coupled to the shared memory cloud, the one or more processor complexes to read blocks of data from and write blocks of data into consecutive memory locations of one or more of the plurality of disk-ROMS to provide disk-like access to the shared memory cloud; and
wherein each processor complex maps the addresses of one or more of the plurality of disk-read-only-memories into processor address spaces to provide memory read access thereto, the one or more processor complexes to read random locations of data from random memory locations of one or more of the plurality of disk-ROMS to provide main-memory-like access to the shared memory cloud.
2. The network computing system of claim 1 , wherein
a total memory space of the shared memory cloud provided by the plurality of disk-ROMs may be pooled together and mapped for usage by one of the one or more processor complexes.
3. The network computing system of claim 1 , wherein
a total memory space of the shared memory cloud provided by the plurality of disk-ROMs is partitioned into a plurality of memory space partitions, and
one or more of the memory space partitions are mapped for usage by one of the one or more processor complexes.
4. The network computing system of claim 1 , wherein
the interconnect network is a hypertransport (HT) routable point to point mesh network.
5. The network computing system of claim 1 , wherein
the interconnect network is a routable point to point packet-switched mesh network.
6. The network computing system of claim 1 , further comprising:
one or more packet routers coupled to the plurality of disk-read-only-memories (disk-ROMs) and the one or more processor complexes, each of the one or more packet routers to route read and write packets towards one of the disk-ROMs or processor complexes in response to a routing table and each destination address of the read and write packets.
7. The network computing system of claim 1 , wherein
one or more of the plurality of disk-read-only-memories (disk-ROMs) in the shared memory cloud include a packet router coupled to other disk-read-only-memories (disk-ROMs) and the one or more processor complexes, the packet router to route read and write packets towards one of the disk-ROMs or processor complexes in response to a routing table and each destination address of the read and write packets.
8. The network computing system of claim 1 , wherein
software drivers in each processor complex map the addresses of the one or more of the plurality of disk-read-only-memories into processor address spaces to provide memory read access thereto.
9. The network computing system of claim 1 , further comprising:
at least one power controller coupled to the one or more processor complexes; and
wherein the at least one power controller to keep the plurality of disk-ROMS powered on while it powers off one or more inactive processor complexes.
10. The network computing system of claim 9 , wherein
upon request, the at least one power controller to power back on one or more processor complexes to execute a process including one or more instructions.
11. The network computing system of claim 1 , wherein
software drivers in each processor complex initiate the block read/write accesses with the plurality of disk-ROM devices and cooperate with each other to coordinate accesses into each disk-ROM device to avoid disk-ROM resource contentions.
12. The network computing system of claim 11 , wherein
the software drivers cooperate with each other to invalidate a plurality of cache lines in respective cache memories of the one or more processor complexes in response to an accumulated plurality of memory change notifications to provide coherency of data in the plurality of disk ROM devices.
13. The network computing system of claim 1 , wherein
each of the disk-read-only-memories (disk-ROMs) in the shared memory cloud includes a packet router, each packet router to route read and write packets towards one of the disk-ROMs or processor complexes in response to a routing table and each destination address of the read and write packets.
14. The network computing system of claim 13 , wherein
each packet router to couple to one or more packet routers in other disk-ROMs.
15. The network computing system of claim 13 , wherein
each packet router to couple to the one or more processor complexes.
16. A network computing system comprising:
a shared memory cloud including a plurality of disk-read-only-memories (disk-ROMs) coupled to an interconnect network suitable for memory read operations and block read/write operations, each of the disk-ROMS including a memory array that is read/write block accessible to access blocks of consecutive memory locations and random read memory accessible to access random memory locations;
a plurality of processor complexes coupled to the shared memory cloud, the plurality of processor complexes to read blocks of data from and write blocks of data into consecutive memory locations of one or more of the plurality of disk-ROMS to provide disk-like access to the shared memory cloud, wherein each processor complex maps the addresses of one or more of the plurality of disk-read-only-memories into processor address spaces to provide memory read access thereto, the plurality of processor complexes to read random locations of data from random memory locations of one or more of the plurality of disk-ROMS to provide main-memory-like access to the shared memory cloud; and
wherein a total memory space of the shared memory cloud provided by the plurality of disk-ROMs is pooled together and mapped for usage by the plurality of processor complexes.
17. The network computing system of claim 16 , wherein
the interconnect network is a hypertransport (HT) routable point to point mesh network.
18. The network computing system of claim 16 , wherein
the interconnect network is a routable point to point packet-switched mesh network.
19. The network computing system of claim 16 , further comprising:
one or more packet routers coupled to the plurality of disk-read-only-memories (disk-ROMs) and the plurality of processor complexes, each of the one or more packet routers to route read and write packets towards one of the disk-ROMs or processor complexes in response to a routing table and each destination address of the read and write packets.
20. The network computing system of claim 16 , wherein
each of the plurality of processor complexes includes a software driver to map the addresses of the one or more of the plurality of disk-read-only-memories into processor address spaces to provide memory read access thereto.
21. The network computing system of claim 16 , wherein
each of the disk-read-only-memories (disk-ROMs) in the shared memory cloud includes a packet router, each packet router to route read and write packets towards one of the disk-ROMs or processor complexes in response to a routing table and each destination address of the read and write packets.
22. The network computing system of claim 21 , wherein
each packet router to couple to one or more packet routers in other disk-ROMs.
23. The network computing system of claim 22 , wherein
each packet router to couple to the plurality of processor complexes.
24. The network computing system of claim 16 , further comprising:
at least one power controller coupled to the plurality of processor complexes; and
wherein the at least one power controller to keep the plurality of disk-ROMS powered on while it powers off one or more inactive processor complexes.
25. The network computing system of claim 24 , wherein
upon request, the at least one power controller to power back on the plurality of processor complexes to execute a process including one or more instructions.
26. The network computing system of claim 16 , wherein
each of the plurality of processor complexes includes a software driver to initiate the block read/write accesses with the plurality of disk-ROM devices and to cooperate with each other to coordinate accesses into each disk-ROM device to avoid disk-ROM resource contentions.
27. The network computing system of claim 26 , wherein
the software drivers cooperate with each other to invalidate a plurality of cache lines in respective cache memories of the plurality of processor complexes in response to an accumulated plurality of memory change notifications to provide coherency of data in the plurality of disk ROM devices.
28. A network computing system comprising:
a shared memory cloud including a plurality of disk-read-only-memories (disk-ROMs) coupled to an interconnect network suitable for memory read operations and block read/write operations, each of the disk-ROMS including a memory array that is read/write block accessible to access blocks of consecutive memory locations and random read memory accessible to access random memory locations;
a plurality of processor complexes coupled to the shared memory cloud, the plurality of processor complexes to read blocks of data from and write blocks of data into consecutive memory locations of one or more of the plurality of disk-ROMS to provide disk-like access to the shared memory cloud, wherein each processor complex maps the addresses of one or more of the plurality of disk-read-only-memories into processor address spaces to provide memory read access thereto, the plurality of processor complexes to read random locations of data from random memory locations of one or more of the plurality of disk-ROMS to provide main-memory-like access to the shared memory cloud; and
wherein a total memory space of the shared memory cloud provided by the plurality of disk-ROMs is partitioned into a plurality of memory space partitions, and the plurality of memory space partitions are respectively mapped for usage by the plurality of processor complexes.
29. The network computing system of claim 28 , further comprising:
one or more packet routers coupled to the plurality of disk-read-only-memories (disk-ROMs) and the plurality of processor complexes, each of the one or more packet routers to route read and write packets towards one of the disk-ROMs or processor complexes in response to a routing table and each destination address of the read and write packets.
30. The network computing system of claim 29 , wherein
one or more of the plurality of disk-read-only-memories (disk-ROMs) in the shared memory cloud include a packet router coupled to other disk-read-only-memories (disk-ROMs) and the plurality of processor complexes, the packet router to route read and write packets towards one of the disk-ROMs or processor complexes in response to a routing table and each destination address of the read and write packets.
31. The network computing system of claim 29 , further comprising:
at least one power controller coupled to the plurality of processor complexes; and
wherein the at least one power controller to keep the plurality of disk-ROMS powered on while it powers off one or more inactive processor complexes.
32. The network computing system of claim 31 , wherein
upon request, the at least one power controller to power back on at least one of the one or more inactive processor complexes to execute a process including one or more instructions.
33. The network computing system of claim 28 , wherein
software drivers in each processor complex initiate the block read/write accesses with the plurality of disk-ROM devices and cooperate with each other to coordinate accesses into each disk-ROM device to avoid disk-ROM resource contentions.
34. The network computing system of claim 28 , wherein
software drivers in each processor complex map the addresses of the one or more of the plurality of disk-read-only-memories into processor address spaces to provide memory read access thereto.