Shulou(Shulou.com)06/03 Report--

This article introduces the knowledge of "what is the importance of consistent hash algorithm". In the operation of actual cases, many people will encounter such a dilemma, so let the editor lead you to learn how to deal with these situations. I hope you can read it carefully and be able to achieve something!

What is the consistent hash algorithm? I believe that many established programmers are not very clear, why? Because it is mainly used for load balancing to solve the problem of relative data consistency. General companies to do a hash, polling, weight are great, with what consistent hash algorithm, but consistent hash algorithm is what, a little pursuit of the product manager still need to know.

There is a hash algorithm in nginx. The use scenario of this hash algorithm is that one of my clients is always connected to a specific back-end server. If it does not change, it will make errors. This can solve some problems in our actual scenarios, such as session problems, redis cache data storage, and so on.

If used for caching scenarios, the normal hash algorithm logic looks like this. If there are 10 cache servers, client 1 data is stored on 1 node (1 modulo divided by 10) client 2 data is stored on 2 nodes. The client 11 is stored on the 1 node, and so on. But suddenly one day, there is a machine aging, your total number of servers has become 9, it is conceivable that if you do the modular operation again, the original data may be gone, maybe you will say, it doesn't matter, there is another copy in the database? Ok after reloading, yes (no knowledge), if for hundreds of millions of traffic scenarios, cache breakdown problems may occur, resulting in a series of crashes.

At this time, it is the turn of the consistent hash algorithm. In fact, the idea and implementation of the consistent hash algorithm is very simple. Whenever you think of the consistent hash algorithm, everyone should think of this ring (I do not know why, whenever I see this ring, I always think of Zhang Heng's seismograph).

The general idea of implementing the algorithm is as follows:

1. Construct a hash ring

2. Map the corresponding node of the server to the hash ring

3. The client requests to find the server node on the hash ring clockwise

That's enough! Such a simple logic, implemented in code, first need to construct a hash ring, the hash ring looks more abstract, how to achieve? Look for the rule and find that there is a KV data pair on the hash ring, key is the hash, and value is the corresponding server, and it is sequential, so you naturally think of SortedMap.

SortedMap sortedMap = new TreeMap (); but how do you implement the hash function? This is more interesting. The process of generating a hash function is not described here. Anyway, you should choose a hash function with a relatively uniform distribution and as large an interval as possible. It is recommended to use ketama or FNV1_32,FNV code to implement it as follows:

Final int p = 16777619; int hash = (int) 2166136261L; for (int I = 0; I

7; hash + = hash > 17; hash + = hash

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.