Problem is a Java application problems, because here is the network communication sector, so only describes the TCP problem, description and explanation about the problem, see a Java plate post: http://bbs.csdn.net/topics/392160230

Problem: the business layer request call slowly, under normal circumstances millisecond can call in a few minutes if a problem occurs may be finish

The following message is problematic, 10. XX. XX. 85 represents the service side, 10. XX. XX. 55 on behalf of the client, is a fixed IP, one is a floating IP, is actually the same machine

Aimed at the newspaper article has the following phenomenon or conclusion:
1. The service side sliding window size size of 16389 (environment using lo loop, MTU 16436164 36=16389 + 7 + 40)
2. The client to the server issued a length of 16396 packets, but because of the other party of the sliding window restrictions be broken up into 16389 and 7 two packets, but from the time, the each interval between two 16396 message around 200 ms ,


As is normal during fetching message, can be seen from the newspaper article the following phenomena:
1. The service side sliding window size of 16396, is the problem occurring times article not disassemble same size (16389 + 7)
2. The length is larger than 10000 message in the message is very little, in the problem of packet, so many 16389 message
Article 3. The message does not appear in two packet interval of 200 ms


Consultation:
1. The above analysis of 200 ms is the cause of request latency, analysis is correct
2. If correct, what factors may cause 200 ms delay

CodePudding user response:

Question 1: can the sender and the receiver in the program code output before and after the receiving of time, both subtraction look to whether move closer to 200 milliseconds.
Question 2: can you send a packet length 10000, look to whether time 100 milliseconds apart, thus determine whether delay and packet length.