Method, system and storage medium for managing data migration

Abstract

This paper describes profiling cache replacement, which is a technique for managing data migration between main memory and cache memory to improve system performance. Unlike conventional cache replacement techniques, profiling cache replacement maintains counters with a profiler that counts memory requests that access pages maintained in cache memory and pages maintained in main memory. Based on information gathered by the profiler (eg, about memory access requests) the mover moves pages between main memory and cache memory. For example, the mover may swap high requested pages of main memory (eg, most requested pages of main memory) with low requested pages of cache memory (eg, least requested pages of cache memory). The mover may do so, for example, when a counter indicates that the amount of access requests for high-request pages of main memory is greater than for low-request pages of cache memory. In order not to block memory user operations (eg, client applications) the mover performs page swapping in the background. For this reason the Mover is limited to exchanging pages at predetermined time intervals such as every microsecond.

Description

[0001] related application [0002] This application claims priority under 35 U.S.C. 119 to Provisional Application No. 62 / 293,688, filed February 10, 2016, entitled "Profiling Cache Replacement," the disclosure of which is incorporated herein by reference in its entirety. Background technique [0003] In computing, a cache is a block of memory used to temporarily store frequently accessed data and to allow future requests for cached data to be serviced faster than requests for uncached data. If the requested data is contained in the cache (a scenario known as a "cache hit"), then the request can be serviced by simply reading from the cache, which is relatively faster than accessing the data from main memory. Instead, if the requested data is not included in the cache (a scenario known as a "cache miss"), then the data is recomputed, or, in traditional techniques, the cache is populated with data from its original storage location, which Slower than simply reading data from c...

AI Technical Summary

Problems solved by technology

This overfill and flush traffic can lead to a condition known as "thrashing", where the number of cache misses increases dramatically, and the time it takes to perform cache fills and flushes due to misses can outweigh the work on the data The time it took for the set to perform the originally requested computational operation

Therefore, traditional cache replacement algorithms have flaws that hinder computational operations

Images