Hi,
When collecting snapshot, there is seemingly discrepency between query
running time and some metrics collected.
For example, on a DPF with 8 nodes, I have 8863201 seconds for
"Total buffer pool read time (milliseconds)" from running "db2 get
snapshot for all on bcudb global"
Then without the "global" option, on each node, there is around
1110000 -- 1120000 milisec for the same metric, e.g. "Total buffer
pool read time (milliseconds) = 1119828"
However, the query itself only runs less than 200 seconds, which is
only 1/5 of the time for "total buffer pool read time" on each node.
How could this be? The snapshot before running the query shows close
to 0 seconds, e.g. "Total buffer pool read time (milliseconds) =
203", so there should be no error.
I guess they multiplied this time with number of prefetchers?
Huaxin
When collecting snapshot, there is seemingly discrepency between query
running time and some metrics collected.
For example, on a DPF with 8 nodes, I have 8863201 seconds for
"Total buffer pool read time (milliseconds)" from running "db2 get
snapshot for all on bcudb global"
Then without the "global" option, on each node, there is around
1110000 -- 1120000 milisec for the same metric, e.g. "Total buffer
pool read time (milliseconds) = 1119828"
However, the query itself only runs less than 200 seconds, which is
only 1/5 of the time for "total buffer pool read time" on each node.
How could this be? The snapshot before running the query shows close
to 0 seconds, e.g. "Total buffer pool read time (milliseconds) =
203", so there should be no error.
I guess they multiplied this time with number of prefetchers?
Huaxin
Comment