a1vin-tian commented on pull request #7153:
URL: https://github.com/apache/skywalking/pull/7153#issuecomment-866574790
1. The worst case of this change is that the executed `prepare requests`
concurrently and then call `batchDAO.synchronous.` But it is also faster than
before because prepare requests are executed concurrently.
2. The reason for accumulating a batch of requests is because of
the following reasons:
1. There will be no small fragments requesting to write to
the storage, reducing the pressure of the storage and improving the write
performance
2. In ES, call each `batchDAO.synchronous` takes at least 10
seconds to cause even if one piece of data is submitted, what will wait 10
seconds. Therefore, it is important to increase the number of each request and
reduce the number of requests, Can reduce all time.
For example:
At `batchDAO.synchronous` stage.
`maxSyncoperationNum` = 50000
Suppose there are 210100 requests.
waited:
thread 1: [50000,50000,5050] 3time
thread 2: [50000,50000,5050] 3time
not waited:
thread 1: [1000,50000,50000,4050] 4 times
thread 2: [1000,50000,50000,4050] 4 times
Suppose there are 45000 requests.
waited:
thread 1: [45000] 1 time
thread 2: [45000] 1 time
not wait:
thread 1: [1000,44000] 2 times
thread 2: [1000,44000] 2 times
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
[email protected]
