A cache isn't really a memory, per say, but rather a temporary storage area where a collection of data duplicated from another area on the computer and is used to expedite access times in cases where the actual information would be inefficient to retrieve. It's simply a faster method for a CPU to locate and use information stored in the main memory that is used frequently. This can be extremely useful in having an application run without having to wait each time it is refreshed or revisited in quick succession.



The CPU is not the only thing on your computer that utilizes the convenience of a cache. Web browsers and web servers also find this aspect of storage useful.

Information held within the cache is said to receive hits when a cache client-which is simply the device or web browser using the cache-tries to search a cache for a tagged piece of datum and finds the necessary tag for what it wants. Instead of continuing to the main memory, the client uses the datum from the cache instead without having to retrieve and compute information found in the main memory. Since the data inside the cache is an exact copy of the original data held in the main memory, the data retrieved from the cache responds exactly as the original would.

By the token of definition above, a cache miss is when the cache client cannot find the desired tag for datum that it is trying to locate. In this occurrence, the client would proceed to the main memory to retrieve and compute the necessary datum required by the commands entered.

In order for datum to be in the cache, it is also in the back store. Sometimes, the cache and back store are written at the same time, but not always. It depends upon the commands for the back store and the write commands for the function of the cache. A write-through is a synchronized operation in which both places are written at once. A write-back occurs when the datum is written to the back store after it is written to the cache; and a no-write allocation occurs when only the processor will be accessing the cache, creating a situation in which no rewriting of the back store or the cache is necessary.

There are several types of caches that a client will use while a computer is in operation. Disk caches and BIND caches are two of the more commonly known in the world of caches.

Caches, it is important to note, are not buffers. The functions of both of these computer operations and storage systems are often combined, but the intent for each is different. A buffer is a temporary memory format and is used when the CPU cannot access addressable information directly for a number of reasons. The use of a buffer for these situations increase transfer speed.

Transfer speed is also improved through the use of a cache, but the speed increase comes from the probability that the same datum will be read multiple times from the cache. This is due to the intent of the cache, which is to reduce access to the underlying forms of slower storage such as the main memory. Also important to note is that the cache is usually invisible to neighboring layers since it is often designed as an abstraction layer with that intent in mind.