Index | Thread | Search

From:
Job Snijders <job@openbsd.org>
Subject:
Re: Failing to cache time values should not be fatal early.
To:
Bob Beck <beck@obtuse.com>
Cc:
Theo Buehler <tb@openbsd.org>, jsing@openbsd.org, beck@openbsd.org, tech@openbsd.org
Date:
Mon, 8 Apr 2024 12:45:18 +0000

Download raw body.

Thread
On Sat, Apr 06, 2024 at 12:59:22PM -0600, Bob Beck wrote:
> Now that we deal with time conversion internally we no longer
> have the performance issue, so remove this cache as it adds
> complexity and confuses people where failure modes get added
> at the wrong level.

before applying patch:
Apr 08 11:18:43 Processing time 458 seconds (381 seconds user, 94 seconds system)
Apr 08 11:26:35 Processing time 451 seconds (380 seconds user, 86 seconds system)
Apr 08 11:34:44 Processing time 460 seconds (381 seconds user, 92 seconds system)
Apr 08 11:42:39 Processing time 455 seconds (380 seconds user, 89 seconds system)
Apr 08 11:50:39 Processing time 455 seconds (380 seconds user, 90 seconds system)

after applying patch:
Apr 08 12:10:08 Processing time 484 seconds (401 seconds user, 102 seconds system)
Apr 08 12:18:44 Processing time 460 seconds (384 seconds user, 92 seconds system)
Apr 08 12:26:49 Processing time 465 seconds (389 seconds user, 89 seconds system)
Apr 08 12:34:46 Processing time 461 seconds (389 seconds user, 89 seconds system)
Apr 08 12:42:41 Processing time 456 seconds (384 seconds user, 88 seconds system)

I don't see a noticable slowdown in default rpki-client on a modest amd64 machine.

Seems OK to me

Kind regards,

Job