Memcached: High-performance distributed memory cache server

Memcached: high-performance distributed memory cache server

features :

The U protocol is simple :

n text-line-based protocols

U - libevent -based event handling :

N Library , can realize the increase of the number of connections , O (1) performance

U Internal Memory storage mode

n data stored in memory , restart data disappears , when data reaches a certain value , based on LRU (last recently used) algorithm to delete unused caches

U Memcached distributed non-communication

n the server side does not have distributed functionality , and the implementation is distributed depending on client

N

use of Memcached :

U Save the method :

N Add: save only if the storage space does not exist the same data

N Replace: save only if the storage space has the same data

N Set: save whenever

method of getting the u

N Get: take a data

N get_multi: fetching multiple data

U Delete Data

n Delete

U increment one and minus one operation

N incr

N ADD

Slab Allocation mechanism : allocating , managing memory Mechanisms

divide the memory into blocks of a certain length , block the same size , and Select the appropriate blocks based on the size of the data.

resolves a memory fragmentation problem , but creates a waste of memory ( such as 100bytes of data to be stored in a block of )

tuning : adjusting the block size of a group based on data size

data deletion aspects of Memcached

U Memcached does not really delete data , does not monitor expired data , but checks for expiration at get , This technique is called lazy ( lazy )

U LRU overwrites the least recently used data

Memcached 's Distributed

You are fully implemented by the client : based on the key , the client's algorithm to obtain which memcached server is stored , Fetch according to the same algorithm .

The function Library of the U-client cache::memcached realizes the distributed function

N dispersion based on the remainder : that is, according to the remainder of the number of servers scattered , to obtain the hash value of the key to take the remainder

N Cons : Add or remove Servers , the cost of cache reassembly is large

U consistent Hashing dispersion algorithm

N Find the hash value of each server and configure it to

Memcached: High-performance distributed memory cache server