1

Topic: The analog critical section is necessary

Kind time of days, All!
There is a problem.
The server, is a lot of kernels, procedure without parameters, fulfills 3 seconds. Returns to all asking (well almost) same.
It cause 10  time at an o'clock.
She reads many pages from a cache (5 at an o'clock). We will consider that hit in a cache of 100 %.
Total, 30 seconds at an o'clock superfluous queue of requests.
To all returns same, but each time works on a data set in which the heap of records in a second flies.
...Here, there was a thought,  this miracle.
That anew enumerated this all not each call, and time in 10 seconds. (For example)
But my attempts to create reference . A label which we read with (rowlock) in transaction - did not lead to success.
Sensation that competing flows in  fly, and read cleaned, but yet the cache not filled anew.
Here.
Question, as though to organize the performance unit, that the first entered into it was fulfilled (With DIRTY READING), and remaining - ?
Here somehow so, thanks in advance.

2

Re: The analog critical section is necessary

...And yes, reading IN transaction should remain dirty for the main requests.
But it would be desirable to surround with its any mechanism a-lja critical section...
That competitors did not enter, and did not begin the , master process yet did not end.

3

Re: The analog critical section is necessary

Makar4ik wrote:

... And yes, reading IN transaction should remain dirty for the main requests.
But it would be desirable to surround with its any mechanism a-lja critical section...
That competitors did not enter, and did not begin the , master process yet did not end.

well, yes, I cannot repeatable read expose for all procedure.
She then dies.
Only dirty reading.

4

Re: The analog critical section is necessary

Question:
Whether the such can help????

begin tran
update table_lock (with updlock) set lock=lock
---And here something it is done
commit

5

Re: The analog critical section is necessary

Yes, thanks while asked a question, seemingly, were invented by the answer.

6

Re: The analog critical section is necessary

Makar4ik wrote:

a question:
Whether the such can help????

begin tran
update table_lock (with updlock) set lock=lock
---And here something it is done
commit
sp_getapplock

Also it is not necessary to invent .

7

Re: The analog critical section is necessary

aleks222 wrote:

it is passed...

sp_getapplock

Also it is not necessary to invent .

...
All right, Alex, the visitor...
Thanks that there...
Itself I will invent a semaphore.

8

Re: The analog critical section is necessary

...More truly, not.
Already prompted the operating decision.
Thanks, question it is closed.

9

Re: The analog critical section is necessary

Makar4ik wrote:

... More truly, not.
Already prompted the operating decision.
Thanks, question it is closed.

Inventors of waterproof gunpowder?
So, would share with the public - though will laugh over what.

10

Re: The analog critical section is necessary

aleks222 wrote:

it is passed...
Inventors of waterproof gunpowder?
So, would share with the public - though will laugh over what.

so, , will laugh over what...

11

Re: The analog critical section is necessary

Makar4ik wrote:

it is passed...
So, , will laugh over what...

And what decision prompted?