Re: Potential regression: high server-side memory consumption during import

On Sat, Mar 03, 2018 at 04:27:36PM +0100, Stefan Sperling wrote:
> Right now I have no idea how to make the current code in trunk any
> faster without reverting r1778923.

Thinking about this some more, I think reverting r1778923 is the
only way to make it faster.

The authz cache key is based on the checksum of the contents of the
authz rules file. This means the entire file needs to be read and
checksummed in order to perform a cache lookup.

It looks like this is where the overhead we're seeing comes from
relative to trunk with r1778923 reverted: The pre-r1778923 code
only reads and calculate the checksum of the file once per connection.

The cache provides a small benefit at connection setup becuase the
authz model doesn't need to be recalculated if already cached due to
another connection. But it doesn't really help on a per-request basis
because of the overhead involved in doing a cache lookup is too large.

Which leads me to believe that r1778923 may have been based on wrong
assumptions about performance. The new authz is not fast enough to
significantly reduce per-request overhead.
Received on 2018-03-03 17:02:57 CET