Skip site navigation (1)Skip section navigation (2)
Date:      Sat, 16 May 2009 21:30:29 +0000 (GMT)
From:      Saifi Khan <saifi.khan@twincling.org>
To:        Abhiman Yashpal Karkera <abhiman.yashpal@gmail.com>
Cc:        freebsd-questions@freebsd.org
Subject:   Re: Query regarding "write" system call.
Message-ID:  <Pine.LNX.4.64.0905162127530.5837@localhost>
In-Reply-To: <38534d590905160831i1c812806qb03bbfc26caa2b62@mail.gmail.com>
References:  <38534d590905160831i1c812806qb03bbfc26caa2b62@mail.gmail.com>

next in thread | previous in thread | raw e-mail | index | archive | help
On Sat, 16 May 2009, Abhiman Yashpal Karkera wrote:

> Hi all,
> I am a newbie to Freebsd OS. I had a query regarding performing writes onto
> a disk.
> 
> Generally when we want to write some data we first copy the data from the
> processes user space to the kernel buffer and hand this buffer to the device
> driver who then goes and initiated a write to the h/w.
> 
> Now my question here is that the kernel buffers are very limited in size ,
> what happens if i have a really huge chunk of data to be written, won't the
> above mechanism of copying from userland to kernel buffer be slow ?
> 
> thanks,
> Abhiman

RDMA is one possibility. 
Zero-copy networking is an illustration of RDMA.

You may also want to look up VIA
http://www.intel.com/intelpress/chapter-via.pdf


thanks
Saifi.



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?Pine.LNX.4.64.0905162127530.5837>