2017-07-13 17:00:17 |
zhang zhenzhong |
description |
Too many period db actions in large scale clusters increase the load of database,
for example over 1000 nodes, it will be 2 * 1000=2000 db update for compute_node in 60s in
_update_available_resource, but this db update is not necessary if resource usage not changed
for compute_node
delete the first and second _update() in _init_compute_node can reduce two db update for one node in 60s, if resource usage not changed for this compute_node. |
Too many period db actions in large scale clusters increase the load of database, especially un-necessary db update or query.
For example, over 1000 nodes, it will be 2 * 1000=2000 db update for compute_node table in 60s in _update_available_resource, but this two db update is not necessary if resource usage not changed.
Delete the first and second _update() in _init_compute_node can reduce two db update for one node in 60s, if resource usage not changed for this compute_node.
Then the funtion self._resource_change(compute_node) in _update() make sense |
|