Managing memory usage in server systems
View Patent ↗A memory manager reduces the impact of memory clean-up operations on server performance. The memory manager monitors local memory usage and other resource usage by the server, such as processor, storage, and network usage. When moderately high memory usage is detected, the memory manager performs memory clean-up operations during relatively low processor, storage, and network usage to reduce the impact of the clean-up operations on server performance. When excessively high memory usage is detected, the memory manager more aggressively cleans-up local memory independently of processor, storage, and network usage.
1. A method for a server system to manage memory in a server, the method comprising the steps of:
monitoring, by the server system, resource usage of a processor and a storage device;
monitoring, by the server system, memory usage for a local memory by the server;
initializing a first clean-up operation that rate limits synchronizing cached pages in the local memory with pages in the storage device to increase the amount of available space in the local memory based on the resource usage and a first level of the memory usage; and
initializing a second clean-up operation that does not rate limit synchronizing the cached pages in the local memory with the pages in the storage device, and independently of the resource usage based on a second level of the memory usage higher than the first level of the memory usage.
2. The method of claim 1 , further comprising determining the first level of the memory usage based on a percentage of the local memory storing the cached pages.
3. The method of claim 1 , further comprising determining the resource usage based on:
a metric of usage of the processor; and
a measure of storage device input/output (I/O) operations and network I/O operations by the processor.
4. The method of claim 3 , further comprising determining the resource usage based on a number of the cached pages that need to be synchronized with the pages in the storage device.
5. The method of claim 1 , further comprising determining the second level of the memory usage based on:
a scan frequency for the cached pages;
a page steal frequency for converting the cached pages into available space in the local memory; or
a measure of swapping of data between the local memory and the storage device.
6. The method of claim 1 , further comprising discontinuing the monitoring of at least some of the resource usage during the second level of the memory usage.
7. The method of claim 1 , further comprising using a module within a kernel space or user space of the server to monitor the resource usage, monitor the memory usage, and initiate the synchronization of the cached pages.
8. The method of claim 1 , further comprising:
estimating an amount of time required to synchronize the cached pages in the local memory with the pages in the storage device; and
synchronizing the cached pages in the local memory with the pages in the storage device based on the estimated amount of time.
9. The method of claim 1 , further comprising determining the second level of the memory usage based on a virtual memory usage and/or a virtual memory efficiency of the storage device as local memory.
10. The method of claim 1 including:
detecting when the memory usage is at the first level; and
while the memory usage is at the first level, limiting memory clean-up operations to reduce an impact of the clean-up operations on server performance.
11. The method of claim 1 including:
detecting when the memory usage is at the second level; and
while the memory usage is at the second level, initializing the second clean-up operation without regard to an impact of the clean-up operation on resource usage and server performance.
12. A server system, comprising:
a server including a processor and local memory;
a storage device; and
one or more stored sequences of instructions, which, when executed by the processor, cause the processor to carry out the steps of:
monitoring local memory usage, processor usage, and storage device usage by the server;
identifying a first memory pressure state for the local memory based on a first level of the local memory usage;
conducting a first clean-up operation for the local memory during the first memory pressure state based on the processor usage and the storage device usage;
identifying a second memory pressure state for the local memory based on a second level of the local memory usage higher than the first level of the local memory usage;
conducting a second clean-up operation during the second memory pressure state independently of the processor usage and the storage device usage;
rate limiting the first clean-up operation during a specified time period; and
not rate limiting the second clean-up operation.
13. The server system of claim 12 , wherein the instructions further cause the processor to carry out the steps of:
monitoring network usage by the server; and
conducting the first clean-up operation based on the network usage; and
conducting the second clean-up operation independently of the network usage.
14. The server system of claim 12 , wherein the instructions further cause the processor to carry out the steps of identifying the first memory pressure state based on an amount of the local memory used for storing the cached pages.
15. The server system of claim 14 , wherein the instructions further cause the processor to carry out the steps of identifying the second memory pressure state based on:
a frequency of page scans in the local memory;
a frequency of converting cached pages in the local memory into available memory space; and/or
an amount of data swapping between the local memory and the storage device.
16. The server system of claim 12 , wherein the instructions further cause the processor to carry out the steps of:
identifying a percentage of dirty pages in the local memory with modified data; and
deferring the first clean-up operation until the percentage of dirty pages drops to a predetermined level.
17. The server system of claim 12 , wherein the instructions further cause the processor to carry out the steps of conducting the first clean-up operation and the second clean-up operation by syncing cached pages in the local memory with associated pages in the storage device and invalidating the cached pages in the local memory.
18. A computer program stored on a non-transitory computer-readable medium for a server system, the computer program comprising a set of instructions operable to:
monitor resource usage by a server;
monitor memory usage of a local memory by the server;
conduct a first clean-up operation for the local memory based on a first level of the memory usage;
initiate the first clean-up operation based on the resource usage;
conduct a second clean-up operation for the local memory based on a second level of the memory usage larger than the first level of the memory usage;
initiate the second clean-up operation independently of the resource usage;
rate limit the first clean-up operation during a specified time period; and
not rate limit the second clean-up operation.
19. The computer program of claim 18 , including instructions operable to identify the first level of the memory usage based on an amount of the local memory used as page cache memory.
20. The computer program of claim 19 , including instructions operable to identify the second level of the memory usage based on scanning and swapping operations for the local memory.
21. The computer program of claim 18 , including instructions operable to determine the resource usage based on usage of a processor in the server, an amount of storage operations by the server with disk memory, and an amount of packet operations by the server over a network.