2016-09-08 15:57:08 |
gordon chung |
description |
this is sort of related to bug 1621498.
basically, we are calculating round_timestamp quite a bit and more often then not for the same timestamps. for example, we calculate this for each aggregate+sampling but it should only be different per sampling[1]. similarly, it's the same timestamps for each split per sampling and same groupby times per sampling. this second case can probably be handled by bug 1621498.
[1] https://github.com/openstack/gnocchi/blob/b8a1172c7914d14eaa0979db69057f25fe4e162b/gnocchi/storage/_carbonara.py#L426-L427 |
this is sort of related to bug 1621498.
basically, we are calculating round_timestamp quite a bit and more often then not for the same timestamps. for example, we calculate this for each aggregate+sampling but it should only be different per sampling[1]. similarly, it's the same timestamps for each split per sampling and same groupby times per sampling. this second case can probably be handled by bug 1621498.
i think it'd be good to start caching the results possibly?
[1] https://github.com/openstack/gnocchi/blob/b8a1172c7914d14eaa0979db69057f25fe4e162b/gnocchi/storage/_carbonara.py#L426-L427 |
|