netdev - Re: [PATCH net-next] net: introduce napi_schedule

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 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:	Tue, 28 Oct 2014 23:20:27 -0700
From:	Eric Dumazet <eric.dumazet@...il.com>
To:	Alexei Starovoitov <alexei.starovoitov@...il.com>
Cc:	David Miller <davem@...emloft.net>, netdev <netdev@...r.kernel.org>
Subject: Re: [PATCH net-next] net: introduce napi_schedule_irqoff()

On Tue, 2014-10-28 at 22:13 -0700, Alexei Starovoitov wrote:

> tried 50 parallel netperf -t TCP_RR over ixgbe
> and perf top were tcp stack bits, qdisc locks and netperf itself.
> What do you see?

You are kidding right ?

If you save 30 nsec ( 2 * 15 nsec) per transaction, and rtt is about 20
usec, its a 0.15 % gain. Not bad for a trivial patch.

Why are you using 50 parallel netperf, instead of trying a single
netperf, as I mentioned latency impact, not overall throughput ?

Do you believe typical servers in data centers are only sending &
receiving bulk packets, with no interrupt, and one cpu busy polling in
NAPI handler ?

Every atomic op we remove/avoid, every irq masking unmasking we remove,
every cache line miss or extra bus transaction we remove, TLB miss, is
the path for better latency.

You should take a look at recent commits I did, you'll get the general
picture if you missed it.

git log --oneline --author dumazet | head -100

--
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