Memory management for busy virtual machine guests
Systems and methods for enhancing memory management for virtual machines. An example method may comprise: accessing, by a hypervisor running on a host computer system, a data structure exposed by a virtual machine managed by the hypervisor, wherein the data structure includes an estimated next access time for one or more memory pages of a plurality of memory pages associated with the virtual machine; estimating a read latency time associated with the virtual machine; identifying, using the data structure, a memory page associated with an estimated next access time that satisfies a predefined condition with respect to the read latency time; and swapping out the memory page.
1 . A method, comprising:
accessing, by a processing device executing a hypervisor running on a host computer system, a data structure generated by and exposed by a virtual machine managed by the hypervisor, wherein the data structure includes an estimated next access time for memory pages of a plurality of memory pages associated with the virtual machine, and wherein the estimated next access time for the memory pages of the plurality of memory pages associated with the virtual machine is sorted based on a likelihood of access of the memory pages;
estimating, by the hypervisor, a read latency time associated with the virtual machine, wherein the read latency time includes an expected read time and a virtual machine exit time;
identifying, by the processing device executing the hypervisor and using the data structure, a memory page associated with the estimated next access time that satisfies a predefined condition with respect to the read latency time; and
swapping, by the hypervisor, out the memory page.
2 . The method of claim 1 , further comprising:
making the memory page inaccessible by the virtual machine.
3 . The method of claim 1 , further comprising:
assigning the memory page to a second virtual machine managed by the hypervisor.
4 . The method of claim 1 , wherein estimating the read latency time further comprises analyzing a lapsed time from a last access to the memory page.
5 . The method of claim 1 , wherein estimating the read latency time further comprises analyzing a frequency of access to the memory page.
6 . The method of claim 1 , wherein the predefined condition is satisfied when the estimated next access time is larger than or equal to the read latency time adjusted by a factor value.
7 . The method of claim 6 , wherein the factor value is set to be dependent on a memory stress metric.
8 . The method of claim 1 , further comprising:
identifying the plurality of memory pages each associated with the estimated next access time that satisfies the predefined condition with respect to the read latency time; and
determining one of the plurality of memory pages as the memory page.
9 . The method of claim 8 , further comprising:
comparing the estimated next access time of the plurality of memory pages.
10 . The method of claim 1 , further comprising:
accessing, by the hypervisor, a second data structure in a guest memory space of a second virtual machine managed by the hypervisor; and
identifying the memory page using the second data structure.
11 . A system comprising:
a memory; and
a processing device operatively coupled to the memory to:
access, by a hypervisor running on a host computer system, a data structure generated by and exposed by a virtual machine managed by the hypervisor, wherein the data structure includes an estimated next access time for memory pages of a plurality of memory pages associated with the virtual machine, and wherein the estimated next access time for the memory pages of the plurality of memory pages associated with the virtual machine is sorted based on a likelihood of access of the memory pages;
estimate, by the hypervisor, a read latency time associated with the virtual machine, wherein the read latency time includes an expected read time and a virtual machine exit time;
identify, by the hypervisor and using the data structure, a memory page associated with the estimated next access time that satisfies a predefined condition with respect to the read latency time; and
swap, by the hypervisor, out the memory page.
12 . The system of claim 11 , wherein the processing device is further to:
make the memory page inaccessible by the virtual machine.
13 . The system of claim 11 , wherein the processing device is further to:
assign the memory page to a second virtual machine managed by the hypervisor.
14 . The system of claim 11 , wherein the predefined condition is satisfied when the estimated next access time is larger than or equal to the read latency time adjusted by a factor value.
15 . The system of claim 14 , wherein the factor value is set to be dependent on a memory stress metric.
16 . The system of claim 11 , wherein to estimate the read latency time associated with the virtual machine, the processing device is to analyze a lapsed time from a last access to the memory page.
17 . The system of claim 11 , wherein to estimate the read latency time associated with the virtual machine, the processing device is to analyze a frequency of access to the memory page.
18 . A non-transitory machine-readable storage medium storing instructions which, when executed by a processing device, cause the processing device to:
accessing, by the processing device executing a hypervisor running on a host computer system, a data structure generated by and exposed by a virtual machine managed by the hypervisor, wherein the data structure includes an estimated next access time for memory pages of a plurality of memory pages associated with the virtual machine, and wherein the estimated next access time for the memory pages of the plurality of memory pages associated with the virtual machine is sorted based on a likelihood of access of the memory pages;
estimating, by the hypervisor, a read latency time associated with the virtual machine, wherein the read latency time includes an expected read time and a virtual machine exit time;
identifying, by the processing device executing the hypervisor and using the data structure, a memory page associated with the estimated next access time that satisfies a predefined condition with respect to the read latency time; and
swap, by the hypervisor, out the memory page.
19 . The non-transitory machine-readable storage medium of claim 18 , wherein the instructions, when executed by the processing device, cause the processing device further to:
make the memory page inaccessible by the virtual machine.
20 . The non-transitory machine-readable storage medium of claim 18 , wherein the instructions, when executed by the processing device, cause the processing device further to:
assign the memory page to a second virtual machine managed by the hypervisor.