Date: Thu, 25 May 2006 14:01:32 -0700 From: Paul Allen <nospam@ugcs.caltech.edu> To: Poul-Henning Kamp <phk@phk.freebsd.dk> Cc: current@freebsd.org, Olivier Gautherot <olivier@gautherot.net> Subject: Re: FreeBSD's embedded agenda Message-ID: <20060525210132.GD28128@groat.ugcs.caltech.edu> In-Reply-To: <4486.1148584202@critter.freebsd.dk> References: <20060525184618.GC28128@groat.ugcs.caltech.edu> <4486.1148584202@critter.freebsd.dk>
next in thread | previous in thread | raw e-mail | index | archive | help
>From Poul-Henning Kamp <phk@phk.freebsd.dk>, Thu, May 25, 2006 at 09:10:02PM +0200: > In message <20060525184618.GC28128@groat.ugcs.caltech.edu>, Paul Allen writes: > > >Erase cycles are often 1ms in duration. > > ... Which is pretty trivial compared to the seek and latency of a > rotating disc. Sure to do 'one of them', but the part of the message you clipped involved the assertion that fragmentation is free. It isn't, you have to do the erase cycle block-by-block, each one sequentially taking 1ms to complete. Sure you can use flash with large block-sizes but then you have to rewrite-out the entire block out after the erase, and writing takes time too--not to mention negatively impacting filesystem stability. And each erase is counts against the cycle-life of the flash. The trade-offs are not trivial and starting with the assertion "fragmentation is free" is begging the question.
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20060525210132.GD28128>