From mboxrd@z Thu Jan  1 00:00:00 1970
From: Stephen Hemminger <stephen@networkplumber.org>
Subject: Re: What is the max size of packets rte_eth_tx_burst()
 can send practically/theoritically?
Date: Tue, 28 Feb 2017 22:24:19 -0800
Message-ID: <20170228222419.1ced1cb2@xeon-e3>
References: <CAFj4wcL2PVD3YHArPb7gXmMr_YZoFodaTaEKbU0YtNmk1R3GUw@mail.gmail.com>
Mime-Version: 1.0
Content-Type: text/plain; charset=US-ASCII
Content-Transfer-Encoding: 7bit
Cc: dev@dpdk.org
To: Joo Kim <itsolution@gmail.com>
Return-path: <dev-bounces@dpdk.org>
Received: from mail-pg0-f49.google.com (mail-pg0-f49.google.com [74.125.83.49])
 by dpdk.org (Postfix) with ESMTP id 7639A2A5E
 for <dev@dpdk.org>; Wed,  1 Mar 2017 07:24:27 +0100 (CET)
Received: by mail-pg0-f49.google.com with SMTP id p5so17036863pga.1
 for <dev@dpdk.org>; Tue, 28 Feb 2017 22:24:27 -0800 (PST)
In-Reply-To: <CAFj4wcL2PVD3YHArPb7gXmMr_YZoFodaTaEKbU0YtNmk1R3GUw@mail.gmail.com>
List-Id: DPDK patches and discussions <dev.dpdk.org>
List-Unsubscribe: <http://dpdk.org/ml/options/dev>,
 <mailto:dev-request@dpdk.org?subject=unsubscribe>
List-Archive: <http://dpdk.org/ml/archives/dev/>
List-Post: <mailto:dev@dpdk.org>
List-Help: <mailto:dev-request@dpdk.org?subject=help>
List-Subscribe: <http://dpdk.org/ml/listinfo/dev>,
 <mailto:dev-request@dpdk.org?subject=subscribe>
Errors-To: dev-bounces@dpdk.org
Sender: "dev" <dev-bounces@dpdk.org>

On Tue, 28 Feb 2017 20:28:21 -0800
Joo Kim <itsolution@gmail.com> wrote:

> Hello,
> 
> 
> For example, as I understand,  ovs-dpdk code  uses a buffer of size 32
> when it transmits via rte_eth_tx_burst().
> 
> I think it can transmit more packets in a bust.
> I know there should be a balance between throughput and latency. But, I am
> wondering what max size packets  a dpdk application can give  to
> rte_eth_tx_burst()  practically/theoretically?
> (I see some max macro in ixgbe_rxtx.h)
> 
> Thanks

You could theoretically transmit a burst size up to the configured number of TX descriptors.
The downside is you will add latency and have to handle the TX ring getting full more often.
Bigger burst sizes really don't win that much. 50% of the gain happens by just sending 2 at a time.