yahoo-eng-team team mailing list archive

[Kevin Benton <1667203@xxxxxxxxxxxxxxxxxx>]

Marking as Won't Fix for now because 50ms for a single get_port seems
reasonable. Anything that needs bulk port retrievals should be using
get_ports which will still emit the same number of queries regardless of
port count.

** Summary changed:

- neutron performance degradation 
+ neutron get_port performance degradation

** Changed in: neutron
       Status: New => Won't Fix

-- 
You received this bug notification because you are a member of Yahoo!
Engineering Team, which is subscribed to neutron.
https://bugs.launchpad.net/bugs/1667203

Title:
  neutron get_port performance degradation

Status in neutron:
  Won't Fix

Bug description:
  I test the runtime of neutron get_port() in Newton and Juno, the
  Newton need more than twice as much as Juno, Newton need about 55ms,
  and Juno need about 20ms. I log the time before
  https://github.com/openstack/neutron/blob/master/neutron/db/db_base_plugin_v2.py#L1300
  and after this time.

  I guess any database operation need more time in Newton than Juno.
  The sqlalchemy core already submit a patch to optimize the time for the eager joinloader, which is used most by neutron.
  Even with this patch, the time of get_port() is 35ms, also more than Juno.

  I think neutron need consider the performance, not mainly focus on the
  new features. I know Newton add more relationship for port, this is
  one reason that cause the poor runtime, but that maybe not the only
  reason.

  https://bitbucket.org/zzzeek/sqlalchemy/issues/3915/performance-
  degradation-on-version-10xx

  
  The salalchemy core said we can use baked query extension to fix this poor runtime.

To manage notifications about this bug go to:
https://bugs.launchpad.net/neutron/+bug/1667203/+subscriptions