716323 - File and prioritize Memshrink bugs in Necko

Status: RESOLVED WONTFIX

(Core :: Networking, defect)

Description

[Brian Smith (:briansmith, :bsmith, use NEEDINFO?)]

In bug 576902 comment 9, njn wrote:
> > AFAICT, this bug is over-prioritized for the Memshrink effort. For
> > Firefox, we are almost definitely better off working on other
> > things in Necko/PSM for Memshrink. My patch only saves 18KB of
> > memory per SSL connection.
>
> Which Necko/PSM things are more important?  How many SSL connections
> do we typically have at any one time?

This bug is about providing Nicholas an answer to that question. We (Necko team) should document what things we can do within Necko to reduce memory, and work with the performance team to prioritize them. File bugs, set the priority (P1, P2, etc.) for the Necko team, and estimate a priority for the performance team (status whiteboard: [Memshrink-P1], [Memshrink-P2]). (These priorities might not match.)

I suspect the first order of business is ensuring that we have a good accounting of all the memory we use in Necko.

Comment 1

[Nicholas Nethercote [inactive]]

Thanks for filing this.

> I suspect the first order of business is ensuring that we have a good
> accounting of all the memory we use in Necko.

The main thing not accounted for at the moment is the Necko buffer cache, but that's slated for removal in bug 715770.  I don't recall seeing any other network stuff high in DMD's output, but that's an imperfect data point.

BTW, we don't prioritize MemShrink bugs prior to meetings, to give a chance for all attendees to see the new bugs and CC themselves if they want.  Also, we use a colon before the priority, e.g. "[MemShrink:P1]".  The bugzilla searches on https://wiki.mozilla.org/Performance/MemShrink will miss bugs that are marked e.g. "[MemShrink-P1]".

Comment 2

[Nicholas Nethercote [inactive]]

We decided in the MemShrink meeting that this bug doesn't need the MemShrink tag, but any bugs filed as a result of this can have the MemShrink tag.

Comment 3

[Nicholas Nethercote [inactive]]

Can this bug be closed? It has just two blocking bugs and so doesn't appear to be serving any useful purpose.