minimise groupby work

[gordon chung]

looking at code, when we have 8 aggregates, we build an aggregatedseries 8x, we update it 8x and write 8x.

the initial aggregatedseries is the same across all aggreagtes except for aggreagtion_method and the update logic is the same (combine/groupby) except for the actually aggregation (https://github.com/openstack/gnocchi/blob/master/gnocchi/carbonara.py#L432-L439). to minimise the number of groupby/round_timestamp calls, we could in theory share all the initial work.

attached my profile to see if it's worth it. profile is based on https://review.openstack.org/#/c/358314 code

[OpenStack Infra (hudson-openstack)]

Fix proposed to branch: master
Review: https://review.openstack.org/368226

[OpenStack Infra (hudson-openstack)]

Reviewed: https://review.openstack.org/368226
Committed: https://git.openstack.org/cgit/openstack/gnocchi/commit/?id=8abc05b22ad541df3abe1cbe3200de5a5e9f3ce0
Submitter: Jenkins
Branch: master

commit 8abc05b22ad541df3abe1cbe3200de5a5e9f3ce0
Author: gord chung <email address hidden>
Date: Fri Sep 9 20:37:28 2016 +0000

    share groupings across aggregates

    this changes it so we don't compute a completely new series over
    and over for each aggregte when in reality, they are all the same.
    should save on round_timestamp calculations as well.

    Change-Id: I525fe97b2674eaf06c7170dce8d40523f15443da
    Closes-Bug: #1621498
    Closes-Bug: #1621510