From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from mail-yh0-f50.google.com (mail-yh0-f50.google.com [209.85.213.50]) by dpdk.org (Postfix) with ESMTP id A7FCB9AD4 for ; Thu, 26 Mar 2015 00:47:37 +0100 (CET) Received: by yhjf44 with SMTP id f44so19366810yhj.3 for ; Wed, 25 Mar 2015 16:47:37 -0700 (PDT) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=gmail.com; s=20120113; h=mime-version:in-reply-to:references:date:message-id:subject:from:to :cc:content-type; bh=cjqp2/zvmohy5QuaI1YLy13fq4xbUagfZGE5QJvUR70=; b=JlqIwnVe/LkGBJ0ZDTyZt6DyfXQrFB9WU9C3BwXZbcrHBsmtDgrkoxlWtcyPte/kYc EoZ05muOlvJ05eLt61FCnszYlB5QoaCEkj0B+pC4DO5Yhh2Z9eMYfRTLgWDH3qUWQzq4 9+7L3ocpLAa/bRBfq4pamv8FGECxx/QUl+WvS99kftSSHNxGTDQMGJt1vXB0vcbGv2EQ TOLY1WY9EqPgFCeTv5D0enns0Z9jEMfH9/HdTJ3sy1e45liTAC7gi1/TOVKqvQosCrQp lE5nKfKmd8sXZDrJixRi2unFnRQwmKBbHeTTNfBQi7BVGhXN8yZCB4PWvcdbYHSSJsA5 T5/w== MIME-Version: 1.0 X-Received: by 10.236.7.170 with SMTP id 30mr12909750yhp.46.1427327257119; Wed, 25 Mar 2015 16:47:37 -0700 (PDT) Received: by 10.170.174.3 with HTTP; Wed, 25 Mar 2015 16:47:37 -0700 (PDT) In-Reply-To: <20150325173248.GA2824@bricha3-MOBL3> References: <20150325150156.GA5620@bricha3-MOBL3> <20150325173248.GA2824@bricha3-MOBL3> Date: Thu, 26 Mar 2015 07:47:37 +0800 Message-ID: From: Cheng Kevin To: Bruce Richardson Content-Type: text/plain; charset=UTF-8 X-Content-Filtered-By: Mailman/MimeDel 2.1.15 Cc: dev@dpdk.org Subject: Re: [dpdk-dev] DPDK testpmd, Virtual Disk IO limitation X-BeenThere: dev@dpdk.org X-Mailman-Version: 2.1.15 Precedence: list List-Id: patches and discussions about DPDK List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Wed, 25 Mar 2015 23:47:37 -0000 Hi Mr.Bruce Richardson I expect the testpmd is able to store the payload of the coming packets and that's all. However, as you can see, my program hits the bottlenecks. The performance is unable to reach 800mbps anymore. My end-goal is simple, finding a way to store the payload without decreasing the speed of packet forwarding. Thanks Kevin 2015-03-26 1:32 GMT+08:00 Bruce Richardson : > On Thu, Mar 26, 2015 at 12:20:42AM +0800, Cheng Kevin wrote: > > Mr. Bruce Richardson > > > > Yes, you are right. This really bother me. > > > > Is there any way to get rid of system call? Maybe some DPDK threading > API? > > Maybe i should use a extra nic card for posting the data out through > > internet, instead of writing on the disk - ex. fwrite. > > > > Or you have some better advises?? > > > > Thanks > > Kevin > > > > Hi Kevin, > > what is your end-goal that you are trying to get to? > > /Bruce > > > > > 2015-03-25 23:01 GMT+08:00 Bruce Richardson >: > > > > > On Wed, Mar 25, 2015 at 10:06:48PM +0800, Cheng Kevin wrote: > > > > Hi all, > > > > > > > > I am a beginner of DPDK. Recently, i am interest in DPDK vHost > app - > > > > testpmd. > > > > > > > > And i have been tracing on testpmd.c and iofwd.c for a while. > > > > > > > > Also add some code inside iofwd.c for storing the payload of > packets. > > > > > > > > Everything goes fine, and the performance is great as expected. > > > > > > > > But when i use fwrite to store the payload into a file, > > > > > > > > the performance decrease from 800mbps to 3mbps (input stream is 1 > > > Gbps). > > > > > > > > Is is caused by the limitation of Virtual Disk IO? How can i > solve it? > > > > > > > > I have tried to search the answer, some people say "pthread" might > > > solve > > > > the problem. > > > > > > > > Can someone give me some hint, i really appreciate for your help. > > > > > > > > > > > > Best Regard, > > > > > > > > Kevin Cheng > > > > > > Two general issues you will hit writing to disk: > > > 1) IO, including disk IO, is slow > > > 2) System calls are slow. > > > > > > You are probably hitting both bottlenecks. > > > > > > /Bruce > > > >