09-05-2014, 12:19 PM
Cache Management Using Multi-Level Hints
Cache Management.doc (Size: 88 KB / Downloads: 15)
Introduction: -
Computer system runs multiple applications at a time. These applications need lots of data which should be kept nearer to the processor so that it should be quickly accessed. The cache memory is used for this purpose as it has faster access than other memory type. But as the size of cache is very small, it is difficult to keep all the data required by the application into the cache. So the hierarchical caches are used in which size of the cache increases from upper level to lower level. Now all the required data is available into the cache. The most recently used data is kept in first level and the least recently used data is kept at last level. To identify and manage these data, hints are used. Hints gives the latest history of the block in the cache According to the requirement, these blocks are promoted from lower level to upper level or demoted form upper level to lower level on the basis of hints.
Now a day, many approaches like hints based multi-level cache like LRU-K are used to increase cache performance by promoting or demoting a memory block into the cache depending upon its latest history. As the block which is moved from cache memory to main memory may be required in future it is kept in lower level cache in the cache hierarchy. So the cost of accessing the block from main memory is minimized. The drawback of existing multi-level cache management techniques is that it uses hints of level L1 only. It does not provide sufficient information about the history of the cache block at all previous cache levels. Therefore an incorrect cache block movement may happen. So, the improved algorithm is proposed which keeps track of the hints of the cache block at multiple levels (K-Step). As the hints are of multiple levels, it gives the correct information about history of cache block and to decide whether to promote or demote the block in the cache becomes easy.
Problem definition:-
1. There is a problem in using hints about how to correctly identify hot or cold data, and then quickly promote hot data to the upper level(s) and demote cold data to the lower level(s). Typically these hints provide just a block’s latest hint information, but miss some important hint history, which reflects a block’s movement among various cache levels.
2. Another problem is how to give a unified management on demote and promote hints. Demote and promote hints are managed separately which may bring an incomplete view on a data and an additional management cost.
Objective:-
1. In this project the objective is to solve the problem of detecting hot or cold data which is solved by using multiple step history hint information, i.e. K-step hints, (Hint-K).
2. Another objective is developing a model which compare the activeness of data blocks in any cache level and perform unified management on demote and promote hints.