Optimized Caching Strategy: A Hybrid of Least Recently Used and Least Frequently Used Methods

The eviction algorithm is what makes it the core, and at the same time, the soul of it is very relevant in a caching system. The proposed method in this paper is adaptive. The cache is also LRU and LFU algorithm-based and this algorithm uses a 'temperature' model and assigns counters to the element in the cache. The counter, in turn, can be seen as a monitor of the elements; it shows their “temperature”. An element with a time-to-live parameter modulated when the counter increases with respect to a threshold will be put into the LFU cache instead of the LRU cache. The time the element is evicted away from the LFU cache is decided by the TTL parameter. The basic idea of this algorithm is that the parts that have been recently and most frequently used will be in the LRU cache and the rest will be in the LFU cache. With counter monitoring of the temperature and access frequency of each item, we adaptively adjust the placement of elements to get the cache hit and eviction rate effectiveness. In this paper, it is shown that the combination of LRU and LFU evictions with the thermal model is a good way to improve cache efficiency in most cases, compared with LRU and LFU used together, but the performance improves through the thermal model consistently.