Then, some data need to be removed from cache storage based on access frequencies, the time intervals, and the last access time. Common cache replacement algorithms include least recently used algorithm (LRU), the minimum frequency of visits algorithm (least frequently used), and recently used Ganetespib Phase 3 algorithm (most recently used). In response to these limitations of the traditional cache replacement algorithms, there are a lot of improved ones. These replacement algorithms are based on the LFU and LRU algorithms. They add some other criteria such as time intervals and access frequencies. They are the deformation of the traditional replacement algorithms. In addition, the document size is introduced to distinguish the characteristics of the replacement algorithms, such as the SIZE algorithm.

Load balancing strategies of distributed caching system are mainly used to solve some cache problems. For example, unbalanced tasks and network congestion will decrease the process ability. Cache systems are mainly used to dynamically balance the tasks in the different servers, so as to improve system performance and speed, and to provide better access quality. Now there are polling, minimum connection number, and process ability balance strategies [8].3. Two-Level Cache ArchitectureAccording to the analysis of hot words and repeatability of queries, it has been found that previous hot spots will still be repeated and maintained hot in current user queries. Based on this phenomenon, if the hot words are initialized in the caches, the entire cache hit rate can be improved.

This paper designs a two-level cache based on user query logs and integrates it into a distributed caching system.3.1. Cache StructureThe structure of the cache design is as follows. In each cluster, each cache server adopts a two-level cache structure. The caches consisted of a static cache and a dynamic cache. The static and dynamic caches work together to cache data and enhance the processing speed. In entire cache structure, static cache will work at first and then dynamic cache does. When each query arrives, our system will look it up in the static cache. If it hits, the cached data is returned; otherwise, it accesses dynamic cache to see if it hits. According to different situation, cache system uses different processes. The static cache stores the highest frequency queries and retrieves the results.

By analyzing the query logs in the cluster, the hottest queries are extracted, and the corresponding search results are in pairs stored in the static cache. Data in static cache are relatively fixed. The contents will change only when the static cache needs to rebuild and replace some contents in Dacomitinib the cache. The dynamic cache changes according to user queries; it stores high frequency user queries and retrieval results. The store is dynamic.