Resiliency and performance for cluster memory

1. A system, comprising:

a computing device comprising a processor and a memory; and

machine-readable instructions stored in the memory that, when executed by the processor, cause the computing device to at least:

identify a free chunk of byte-addressable memory specified by an allocation structure maintained by a cluster memory host of a plurality of memory hosts;

acquire a chunk of byte-addressable memory from the cluster memory host by performing an atomic compare-and-swap operation on the allocation structure to claim the free chunk of byte-addressable memory;

identify an active set of allocated memory pages and an inactive set of allocated memory pages for a process executing on the computing device;

store the active set of allocated memory pages for the process in the memory of the computing device; and

store the inactive set of allocated memory pages for the process in the chunk of byte-addressable memory of the cluster memory host.

2. The system of claim 1 , wherein the machine-readable instructions, when executed by the processor, further cause the computing device to at least:

detect a change in the active set of a page of allocated memory pages used by the process;

in response to the page moving from the inactive set to the active set of allocated memory pages, remap the page from the chunk of byte-addressable memory of the cluster memory host to the memory of the computing device; and

in response to the page moving from the active set to the inactive set of allocated memory pages, remap the page from the memory of the computing device to the chunk of byte-addressable memory of the cluster memory host.

3. The system of claim 2 , wherein the machine-readable instructions that detect a change in the active set of allocated memory pages used by the process further cause the computing device to at least:

compare the active set of pages identified in a previous time period with the active set of pages identified in a current time period.

4. The system of claim 2 , wherein the machine-readable instructions that detect the change in the active set of allocated memory pages used by the process further cause the computing device to at least:

clear an access bit in a page table at a beginning of a time period, the access bit being associated with the page; and

determine whether the access bit remains unset at an end of the time period.

5. The system of claim 2 , wherein the machine-readable instructions that detect the change in the active set of allocated memory pages used by the process further cause the computing device to at least:

interrupt the process during a time period;

record page numbers of memory locations being accessed when the interrupt occurred; and

identify the active set to include the page numbers recorded as being accessed at an end of the time period.

6. The system of claim 1 , wherein the machine-readable instructions that cause the computing device to store the inactive set of allocated memory pages for the process in the chunk of byte-addressable memory of the cluster memory host further comprise encoding the inactive set of allocated memory pages using erasure encoding.

7. The system of claim 1 , wherein the machine-readable instructions that cause the computing device to acquire the chunk of byte-addressable memory from the cluster memory host further cause the computing device to:

send a request for the chunk of byte-addressable memory to a management host; and

receive a response from the management host, the response identifying the cluster memory host and the chunk of byte-addressable memory on the cluster memory host.

8. A method implemented by a computer, comprising:

identifying a free chunk of byte-addressable memory specified by an allocation structure maintained by a cluster memory host of a plurality of memory hosts;

acquiring a chunk of byte-addressable memory from the cluster memory host by performing an atomic compare-and-swap operation on the allocation structure to claim the free chunk of byte-addressable memory;

identifying an active set of allocated memory pages and an inactive set of allocated memory pages for a process executing on the computer;

storing the active set of allocated memory pages for the process in a memory of the computer; and

storing the inactive set of allocated memory pages for the process in the chunk of byte-addressable memory of the cluster memory host.

9. The method implemented by the computer of claim 8 , further comprising:

detecting a change in the active set of a page of allocated memory pages used by the process;

in response to the page moving from the inactive set to the active set of allocated memory pages, remap the page from the chunk of byte-addressable memory of the cluster memory host to the memory of the computer; and

in response to the page moving from the active set to the inactive set of allocated memory pages, remap the page from memory of the computer to the chunk of byte-addressable memory of the cluster memory host.

10. The method implemented by the computer of claim 9 , further comprising:

comparing the active set of pages identified in a previous time period with the active set of pages identified in a current time period.

11. The method implemented by the computer of claim 9 , wherein detecting the change in the active set of allocated memory pages used by the process further comprises:

clearing an access bit in a page table at a beginning of a time period, the access bit being associated with the page; and

determining that the access bit remains unset at an end of the time period.

12. The method implemented by the computer of claim 9 , further comprising:

interrupting the process during a time period;

recording page numbers of memory locations being accessed when the interrupt occurred; and

identifying the active set to include the page numbers recorded as being accessed at an end of the time period.

13. The method implemented by the computer of claim 8 , wherein storing the inactive set of allocated memory pages for the process in the chunk of byte-addressable memory of the cluster memory host further comprises encoding the inactive set of allocated memory pages using erasure encoding.

14. The method implemented by the computer of claim 8 , wherein acquiring the chunk of byte-addressable memory from the cluster memory host further comprises:

sending a request for the chunk of byte-addressable memory to a management host; and

receiving a response from the management host, the response identifying the cluster memory host and the chunk of byte-addressable memory on the cluster memory host.

15. A non-transitory, computer-readable medium, comprising machine-readable instructions that, when executed by a processor of a computing device, cause the computing device to at least:

identify a free chunk of byte-addressable memory specified by an allocation structure maintained by a cluster memory host of a plurality of memory hosts;

acquire a chunk of byte-addressable memory from the cluster memory host by performing an atomic compare-and-swap operation on the allocation structure to claim the free chunk of byte-addressable memory;

identify an active set of allocated memory pages and an inactive set of allocated memory pages for a process executing on the computing device;

store the active set of allocated memory pages for the process in a memory of the computing device; and

store the inactive set of allocated memory pages for the process in the chunk of byte-addressable memory of the cluster memory host.

16. The non-transitory, computer-readable medium of claim 15 , wherein the machine-readable instructions, when executed by the processor, further cause the computing device to at least:

detect a change in the active set of a page of allocated memory pages used by the process;

in response to the page moving from the inactive set to the active set of allocated memory pages, remap the page from the chunk of byte-addressable memory of the cluster memory host to the memory of the computing device; and

in response to the page moving from the active set to the inactive set of allocated memory pages, remap the page from memory of the computing device to the chunk of byte-addressable memory of the cluster memory host.

17. The non-transitory, computer-readable medium of claim 15 , wherein the machine-readable instructions, when executed by the processor, further cause the computing device to at least:

compare the active set of pages identified in a previous time period with the active set of pages identified in a current time period.

18. The non-transitory, computer-readable medium of claim 15 , wherein the machine-readable instructions that cause the computing device to detect a change in the active set of allocated memory pages used by the process further cause the computing device to

interrupt the process during a time period;

record page numbers of memory locations being accessed when the interrupt occurred; and

identify the active set to include the page numbers recorded as being accessed at an end of the time period.

19. The non-transitory, computer-readable medium of claim 15 , wherein the machine-readable instructions that cause the computing device to store the inactive set of allocated memory pages for the process in the chunk of byte-addressable memory of the cluster memory host further comprise encoding the inactive set of allocated memory pages using erasure encoding.

20. The non-transitory, computer-readable medium of claim 15 , wherein the machine-readable instructions that cause the computing device to acquire the chunk of byte-addressable memory from the cluster memory host further cause the computing device to:

send a request for the chunk of byte-addressable memory to a management host; and

receive a response from the management host, the response identifying the cluster memory host and the chunk of byte-addressable memory on the cluster memory host.