Distributed cache system Hotspot key solution

1. Read access is high (million level and above), read access is much greater than write access

In this case, one of the typical business scenarios is the storage configuration information, the general data volume of the configuration information is small, the update frequency is low, but the read traffic is high. There are two ways of optimizing this scenario:

A. Store multiple backups, random reads, and decentralized read access pressures on the same key.

B. Use the local cache. The Local cache basically blocks most of the access, and the amount of read traffic that actually falls on the cache system is minimal. But there is a problem with the local cache: synchronization of data. The solution can periodically update the local cache (updated every n seconds).

2. High write access (million level and above), write access is much greater than read access
One of the typical business scenarios for this scenario is counting statistics, such as PV statistics on a page. This scenario is optimized by splitting a single key statistic into multiple key statistics, assuming that key is a, then a can be split into a1,a2 ... A100,INCR randomly selects a key, reads the total number of all keys, and then the combined statistics. Of course, this way to enlarge the read operation, but based on the reading itself is not high, such an increase in the cluster does not have a large impact.

Distributed cache system Hotspot key solution