I have a slow running query that I'm working on optimizing.
In Client Statistics in Management Studio, this server was taking about 8 seconds waiting time on the north and around 1 second for customer processing time.
I have always thought that server was the number of waiting time on responses and customer processing time was generally related to bandwidth or large data size.
I made many changes to the query and now my wait time at server reply is about 250ms, however, the customer processing time has increased, the total execution time of approximately 9 seconds is slightly slower.
As a result, the return can be exactly the same.
Is there any difference between these two digits but what can be the result?
'Wait for time on server response' l is the time between the elite request packet to the client The first response packet left and returned to the server. 'Client processing time' is the time between the first response packet and the final response packet. BTW, I could not find the documentation to return these claims, but I would say, on the basis of my observations, they are valid educated estimates.
If you run a query with a large latency, the server responds to 'This means that the server takes a long time to produce the first line, it is normal on common questions that have the operator Those who require full sub-queries to evaluate before proceeding (general examples are sort operators).
On the other hand, a very 'waiting time' response on the server 'means that the query was able to quickly return the first row. Although for a long time, 'client processing time' does not mean that the client spent a lot of time processing and the server was waiting on the client. This can mean that the rows are returning from the server result and it takes so long to return to the last line.
What you see is the result of the changes in the query plan, which may have been removed from an operator who was blocking the execution (most probably in a way) and the new plan is a different strategy. Which makes the first result faster (probably uses an index that guarantees the requested order so that no one is necessary), but the last time, the last time. / P>
If you are worried about the customer while holding the customer back (this can be on a large result set) then you should check out wait_type (this is also information and useful) ) Sessions executed during checkout If I'm waiting on the client waiting type server is not incorrect, then you can do a total check, then reset it using the query again, see that ASYNC_NETWORK_IO Icsha type How does long. Of course, make sure there is no other activity on the server during the trial.
No comments:
Post a Comment