From: Dave Taht <dave.taht@gmail.com>
To: Jim Reisert AD1C <jjreisert@alum.mit.edu>
Cc: "cerowrt-devel@lists.bufferbloat.net"
<cerowrt-devel@lists.bufferbloat.net>
Subject: Re: [Cerowrt-devel] cerowrt-3.10.44-6 released
Date: Thu, 26 Jun 2014 12:53:11 -0700 [thread overview]
Message-ID: <CAA93jw4MJxR9htBQJm5KzYgP1jOvFYQRF_zFwxnCLi1wV1XcUQ@mail.gmail.com> (raw)
In-Reply-To: <CAK-n8j757OH3udLOqz-TQs3MHoM8nLj2pJoXgGyiSdVgZ-mHYA@mail.gmail.com>
I have been blowing it up with continuous traffic, saturating the cpu,
talking to multiple devices, using crypto and ethernet, now for
several days.
I am using sfq on the wifi, rather than fq_codel, at least temporarily.
Occasionally, using the rrul test, I can get it to temporarily either
lose a route or deassociate from wifi, or something else like netperf
itself blowing up, but it always comes back within a few seconds. So
far, using the rrul_be test I haven't seen that happen. I do see cpu
absolutely maxed at present with rrul and not rrul_be.
I have given up on tuning the wifi aggregate queue length (formerly
12, now 48) as much as I had before - as the previous setting had
adverse effects on download throughput, so I'll put that fix in the
next release and live with the added latency until we can pull
together the make-wifi-fast project.
I see an improvement on the simultaneous upload/download test from
10mbit up/.2 down to 6up,3down, latency (as measured on the main
client) of about 80ms (which still sucks about 10x worse than what
seems possible)
I ran out of space on my packet captures...
I am going to be adding more clients, switching back to fq_codel, and
adding in ipv6 if this continues to take the load I am throwing at it.
Then adding impairments to degrade the wifi to more marginal levels.
Interestingly I have not got any instruction traps, nor the famed dma
overrun bug/recovery in this release so far... I am encouraged, but
until I setup a mac with bad connectivity and have things run for
days and days I won't sleep well.
And I had done a build from this codebase for the picostation and am
watching that crash under normal load regularly. That box only has
32MB of memory, where the cerowrt box has 128MB memory.
On Thu, Jun 26, 2014 at 12:28 PM, Jim Reisert AD1C
<jjreisert@alum.mit.edu> wrote:
> Been running since Wednesday evening, no issues so far.
>
> --
> Jim Reisert AD1C, <jjreisert@alum.mit.edu>, http://www.ad1c.us
--
Dave Täht
NSFW: https://w2.eff.org/Censorship/Internet_censorship_bills/russell_0296_indecent.article
next prev parent reply other threads:[~2014-06-26 19:53 UTC|newest]
Thread overview: 4+ messages / expand[flat|nested] mbox.gz Atom feed top
2014-06-25 3:37 Dave Taht
2014-06-26 19:28 ` Jim Reisert AD1C
2014-06-26 19:53 ` Dave Taht [this message]
2014-07-06 19:45 ` R.
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
List information: https://lists.bufferbloat.net/postorius/lists/cerowrt-devel.lists.bufferbloat.net/
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=CAA93jw4MJxR9htBQJm5KzYgP1jOvFYQRF_zFwxnCLi1wV1XcUQ@mail.gmail.com \
--to=dave.taht@gmail.com \
--cc=cerowrt-devel@lists.bufferbloat.net \
--cc=jjreisert@alum.mit.edu \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox