Write-optimized persistent memory heap management method

Abstract

The invention discloses a write-optimized persistent memory management method. The method comprises the steps of S1, performing root information registration in a persistent memory; s2, constructing adata structure of a heap management memory, and storing the data structure in a DRAM; and S3, in a system fault recovery stage, recovering residual free space in the persistent memory space based onthe registered root information, and reconstructing a data structure of the heap management memory stored in the DRAM to achieve heap management of the persistent memory. According to the method, allmetadata structures allocated to the management memory can be stored in the DRAM, and extra write operation on the persistent memory cannot be caused by allocation or release operation of any persistent memory space; in addition, in the fault recovery process, the reference count of the object can be accurately restored through the number of times of searching the same object while the allocated space is searched, so that the reference count of the object can be stored in the DRAM without worrying about the reference count error or loss caused by the fault.

Description

technical field [0001] The invention belongs to the technical field of persistent heap management, and in particular relates to a write-optimized persistent memory heap management method. Background technique [0002] With the vigorous development of Internet applications, massive data storage and access have brought great challenges to cloud computing and big data processing platforms. In current computer systems, DRAM is usually used as memory, and solid-state hard drives or mechanical hard drives are used as external storage. , the capacity of the external memory is large, but the access delay is high, and it is not only addressable by byte, but can only be accessed in units of "blocks", so the data cannot be directly processed in the external memory, and the data can only be processed after being read into the memory . The memory access speed is fast, but the capacity is small, and data will be lost when the power is turned off, so the computer system often needs to exc...

AI Technical Summary

Problems solved by technology

The lack of fault recovery will lead to a series of serious problems such as memory leaks, dangling pointers, and repeated releases

The second challenge is the need to minimize the extra write operations that heap management brings to persistent memory

The third challenge is that it is difficult to achieve automatic garbage collection based on reference counting while minimizing write operations to persistent memory.

Another representative of this type of solution, NV-Heaps, in addition to the basic persistent heap management functions, also provides support for garbage collection based on reference counting, which saves reference counts in persistent memory. The reference count is updated synchronously at all times, and the transactional nature of the reference count update operation is guaranteed by logging. The ensuing problem is that in order to maintain the correctness of the reference count, the persistent memory has to be frequently written.

[0007] The representative of the second type of scheme is Makalu. Although it effectively reduces the writing of allocation and release operations to persistent memory, it still cannot completely avoid the modification of key metadata when allocation and release operations involve modification of key metadata. Writes to persistent memory

And the non-intrusive garbage collection algorithm used by Makalu has the problem of inaccurate collection, which may cause memory leaks, so it cannot be used to accurately rebuild reference counts

[0008] As mentioned above, the existing persistent heap management technical solutions mainly have three problems: first, the allocation and release of persistent memory will bring a large number of write operations to persistent memory

The second is that the fault recovery based on the non-intrusive garbage collection algorithm cannot accurately identify the allocated space, which may cause memory leaks

The third is that the overhead of implementing reference counting is high, and it is necessary to save and track the modification of reference counting in persistent memory, resulting in frequent writes to persistent memory

Therefore, there is currently no heap management solution for persistent memory that can guarantee the transactional nature of allocation and release operations and provide reference-based Automatic garbage collection support for counting

Images