netdev - Re: Qdisc: Measuring Head-of-Line blocking with netperf-wrapper

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Date:	Mon, 15 Sep 2014 12:12:24 -0700
From:	Rick Jones <rick.jones2@...com>
To:	Dave Taht <dave.taht@...il.com>,
	Eric Dumazet <eric.dumazet@...il.com>
CC:	Tom Herbert <therbert@...gle.com>,
	Jesper Dangaard Brouer <brouer@...hat.com>,
	"netdev@...r.kernel.org" <netdev@...r.kernel.org>,
	Stephen Hemminger <stephen@...workplumber.org>,
	David Miller <davem@...emloft.net>,
	Hannes Frederic Sowa <hannes@...essinduktion.org>,
	Daniel Borkmann <dborkman@...hat.com>,
	Florian Westphal <fw@...len.de>,
	Toke Høiland-Jørgensen <toke@...e.dk>,
	Tim Shepard <shep@...m.mit.edu>,
	Avery Pennarun <apenwarr@...gle.com>
Subject: Re: Qdisc: Measuring Head-of-Line blocking with netperf-wrapper

On 09/15/2014 11:55 AM, Dave Taht wrote:
> Problems with TCP_RR and UDP_RR are that they are subject to a RTO on
> packet loss.

One possible way to work/kludge around that for TCP_RR would be to have 
a burst-mode test with say five or so transactions in flight at one 
time.  That then should allow for fast retrans on the occasional 
single-packet loss and it will be just an RTT rather than an RTO.  There 
would probably be some interactions with cwnd and bundling though upon 
that retransmission, even with -D set.

> It would be nice if netperf could sprout a "send a user specified
> packet size on this isochronous interval (with a resolution below a
> ms) over X transport" test. Or there was something simpler than owamp
> but as secure.

If you are willing to let netperf spin (ie burn local CPU on the system 
on which netperf is running) over the interval you can configure 
--enable-intervals --enable-spin  and then it will spin rather than rely 
on the resolution of setitimer().  Then the command line would be 
something along the lines of:

netperf -H <destination> -b 1 -w <waittime> -- -m <size>

and it will then make one send of <size> every <waittime>.

happy benchmarking,

rick
--
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html