Memory wear leveling
In some examples, based on the information relating to the wear of physical memories in computer nodes of the distributed system, a system initiates a migration of a memory page from a first physical memory in a first computer node to a second physical memory in a second computer node. As part of the migration, the system updates a mapping between a first address space accessible by programs in the distributed system and a physical address space comprising memory locations in the physical memories.
1. A non-transitory machine-readable storage medium comprising instructions that upon execution cause a distributed system to:
receive information relating to wear of physical memories in computer nodes of the distributed system, wherein each of the computer nodes comprise a corresponding set of hardware resources, including processing resources, remote from hardware resources of other computer nodes;
based on the information relating to the wear, initiate a migration of a memory page from a first physical memory in a first computer node to a second physical memory in a second computer node, wherein the second physical memory is remote from the first physical memory; and
as part of the migration, update a mapping between a first address space accessible by programs in the distributed system and a physical address space comprising memory locations in the physical memories.
2. The non-transitory machine-readable storage medium of claim 1 , wherein the first address space is a virtual address space comprising virtual addresses in a virtual memory.
3. The non-transitory machine-readable storage medium of claim 1 , wherein the first address space is a guest physical memory address space of a guest physical memory accessible by one or more guest operating systems (OSes) in a virtualized environment.
4. The non-transitory machine-readable storage medium of claim 1 , wherein the information relating to the wear comprises information relating to correctable errors in the physical memories.
5. The non-transitory machine-readable storage medium of claim 1 , wherein the information relating to the wear comprises information relating to cumulative access of the physical memories.
6. The non-transitory machine-readable storage medium of claim 1 , wherein the information relating to the wear comprises information relating to intensity of access of the physical memories.
7. The non-transitory machine-readable storage medium of claim 1 , wherein the information relating to the wear is from hardware in the computer nodes.
8. The non-transitory machine-readable storage medium of claim 1 , wherein the information relating to the wear is from a program that samples virtual memory addresses accessed and determines physical memory addresses corresponding to the virtual memory addresses.
9. The non-transitory machine-readable storage medium of claim 1 , wherein the instructions upon execution cause the distributed system to:
migrate a virtual processor from the first computer node to the second computer node in conjunction with the migration of the memory page.
10. The non-transitory machine-readable storage medium of claim 1 , wherein the initiating of the migration of the memory page is by a hyper-kernel in the first computer node.
11. The non-transitory machine-readable storage medium of claim 10 , wherein the mapping is updated by the hyper-kernel.
12. The non-transitory machine-readable storage medium of claim 1 , wherein the initiating of the migration of the memory page is by a hypervisor in the first computer node.
13. The non-transitory machine-readable storage medium of claim 12 , wherein the mapping is updated by the hypervisor.
14. A distributed system comprising:
a plurality of computer nodes, wherein each computer node of the plurality of computer nodes comprises a physical memory and a corresponding set of hardware resources, including processing resources, remote from hardware resources of other computer nodes,
wherein a first computer node of the plurality of computer nodes comprises:
memory wear-leveling instructions executable in the first computer node to:
receive information relating to wear of a first physical memory in the first computer node,
based on the information relating to the wear, initiate a migration of a memory page from the first physical memory in the first computer node to a second physical memory in a second computer node, wherein the second physical memory is remote from the first physical memory, and
as part of the migration, update a mapping between a first address space accessible by a program and a physical address space comprising memory locations in the physical memories.
15. The distributed system of claim 14 , wherein the updating of the mapping changes a mapping between a first virtual address of the memory page from a first physical address of the first physical memory to a second physical address of the second physical memory.
16. The distributed system of claim 14 , wherein the updating of the mapping changes a mapping between a first guest memory address of the memory page from a first physical address of the first physical memory to a second physical address of the second physical memory.
17. The distributed system of claim 14 , wherein the plurality of computer nodes comprise respective hyper-kernels that collectively form a distributed hypervisor, wherein the program is executable in a virtual machine distributed across the plurality of computer nodes.
18. The distributed system of claim 14 , wherein the information relating to the wear comprises information relating to correctable errors in the first physical memory.
19. A method of a distributed system comprising a plurality of computer nodes, comprising:
receiving, by a memory wear-leveling engine in a first computer node of the plurality of computer nodes, information relating to wear of a first physical memory in the first computer node, wherein each of the computer nodes comprise a corresponding set of hardware resources, including processing resources, remote from hardware resources of other computer nodes;
based on the information relating to the wear, initiating a migration of a memory page from the first physical memory in the first computer node to a second physical memory in a second computer node, wherein the second physical memory is remote from the first physical memory; and
as part of the migration, updating a mapping between a first address space accessible by programs in a virtual machine and a physical address space comprising memory locations in the physical memories.
20. The method of claim 19 , further comprising:
migrating a virtual processor from the first computer node to the second computer node in conjunction with the migration of the memory page.