Date: Sun, 17 Feb 2002 15:19:46 -0800 From: Terry Lambert <tlambert2@mindspring.com> To: Roy Sigurd Karlsbakk <roy@karlsbakk.net> Cc: Dag-Erling Smorgrav <des@ofug.org>, Thomas Hurst <tom.hurst@clara.net>, hiten@uk.FreeBSD.org, hackers@FreeBSD.ORG, freebsd-questions@FreeBSD.ORG Subject: Re: in-kernel HTTP Server for FreeBSD? Message-ID: <3C703A92.2EBD3E67@mindspring.com> References: <Pine.LNX.4.30.0202171935060.6486-100000@mustard.heime.net>
next in thread | previous in thread | raw e-mail | index | archive | help
Roy Sigurd Karlsbakk wrote: > > > > sendfile() isn't zero-copy, it's just two-less-copies. > > > zero-copy means "zero copy-operations within memory" > > > > To an MCSE, maybe. > > strange ... > > It's interesting that Dr. Scient Paal Halvorsen that recently finished his > thesis about the subject 'Memory buffering / caching in multimedia > streaming systems'. This has the following description > > The INSTANCE operating system enhancements show great improvements for > multimedia streaming. In this thesis, we want to look at > buffering/caching/prefetching mechanisms suitable for multimedia streams > combined with the INSTANCE zero-copy data path. > > See http://www.ifi.uio.no/~paalh/index2.html for more info > > hm > > Am I an idiot here or have someone else misunderstood? First, the URL reference you give is bad; it's one of those terrible web pages that has a stupid frame that prevents you getting the real URL. The publications URL is: http://www.ifi.uio.no/~paalh/publications/ Second, there's no publication there with this title; the closest one is where he's the junior name on the paper: "Evaluation of a Zero-Copy Protocol Implementation". Third, this paper is bases on the idea that data is being replicated in main memory in lots of places; that doesn't really happen on FreeBSD, which uses a unified VM and buffer cache. Fourth, I think the reason he made the joke about MSCE's was that there are ways of doing true zero copy, using DMA directly between devices. There are a couple of people that have done this by, for example, rewriting the Tigon II firmware to permit page aligned buffers, and DMA'ing directly between disk and ethernet controllers. Technically, that's not zero copy, either, since there's a copy on the disk, and one in some controller memory. "Zero copy" usually means "zero unnecessary copies"; but what someone thinks of as "necessary" is really based on their bias towards an existing implementation. 8-). -- Terry To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-hackers" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?3C703A92.2EBD3E67>