Shulou(Shulou.com)06/01 Report--

This article mainly explains "what is the method of RPC congestion control". The content of the explanation is simple and clear, and it is easy to learn and understand. Please follow the editor's train of thought to study and learn "what is the method of RPC congestion control".

Priority rating

We say that congestion occurs in RPC, which is actually the result of the interaction of resource requests, and the root cause of this interaction is that we do not further separate them, but are redundant to deal with them together. The first thing to improve here is to divide the priority relationship, with each priority corresponding to a queue, such as Q0 and Q1 and Q3, and then define a rule that the smaller the number, the higher the priority.

Queue priority determination

After the queue priority is divided, a very important operation is to confirm the priority. Here, of course, we will not manually set the priority of the request. A more general, more intelligent algorithm is to determine the priority according to the frequency of the request. For users, what they need to do is to specify this rule. Specifically, for example, requests from three queues, Q1 queue, the request frequency is between 0 and 10%. Q2 is 1050%, and the remaining 50 or more are Q3 queues. If converted into the actual number, those with more than 50 requests are classified as Q2 priority queues, and those with less than 10 requests are classified as Q0 queues. If we take a look at the congestion phenomenon seen before, most of the requests that cause congestion will be assigned to the same queue, so that the impact on other priority queues can be reduced in subsequent processing.

Secondly, the counting statistics of frequency should also take into account the problem of time limit, because the frequency statistics should be aimed at a certain period of time, for example, the frequency of A period can not be exactly the same as that of B period, and it can not be used all the time. Here HADOOP-9640 mentions an attenuation algorithm in which the counting results in the previous period gradually attenuate in the next round of calculation through the attenuation factor, which is much smoother than complete zero-clearing statistics.

Queue priority weight setting

In fact, after each request is assigned to each priority queue, there are many ways to deal with it. For example, in the simplest way, a certain batch of requests are extracted from each queue by polling. Here, we mention a more flexible approach, in the context of polling, and then set a theoretical guarantee proportion for each queue. For example, there are 3 priority queues, Q0 and Q1, Q2, and configuration weights, such as 9, 4, and 1. In one poll, 9 requests from Q0 queue, 4 requests from Q1, and one request from Q2 are guaranteed to be processed within the next 15 requests.

Thank you for your reading, the above is the content of "what is the method of RPC congestion control". After the study of this article, I believe you have a deeper understanding of what the method of RPC congestion control is, and the specific use needs to be verified in practice. Here is, the editor will push for you more related knowledge points of the article, welcome to follow!

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.