on 2016 Jan 19 10:46 AM
Just the number of physical index and table pages written by actual checkpoint operations?
Is there any I/O optimization; e.g., predictive writes prior to a final checkpoint?
Does DiskWrite include anything else?
Physical temporary file pages?
OUTPUT statements?
UNLOAD statements?
xp_write_file()?
There is only two places in SQLA that the DiskWrite database counter is incremented. They are:
So the counter does not include any writes to any auxiliary files such as unload, xp_write_file, etc (nor output... but this is done by the dbisql[c] client!)
Regarding your question "Is there any I/O optimization?". Yes, prior to a checkpoint a pass over the cache is made to flush dirty pages before the database is locked (from connections) to do the actual checkpoint.
You must be a registered user to add a comment. If you've already registered, sign in. Otherwise, register and sign in.
Sorry, by "logs" I was referring to the transaction log and mirror log, not the console log, etc. I've clarified my answer. LogWrite refers to the number of writes made to the transaction log and is a subset of DiskWrite.
User | Count |
---|---|
67 | |
11 | |
10 | |
10 | |
9 | |
9 | |
6 | |
5 | |
4 | |
4 |
You must be a registered user to add a comment. If you've already registered, sign in. Otherwise, register and sign in.