Date: Fri, 4 Sep 1998 09:59:40 +0200 (CEST) From: Andrzej Bialecki <abial@nask.pl> To: freebsd-small@FreeBSD.ORG Subject: Flash disk test results Message-ID: <Pine.BSF.4.02A.9809040930290.26306-100000@korin.warman.org.pl>
next in thread | raw e-mail | index | archive | help
Hi, A few days ago I received a SanDisk IDE flash disk sent to me by Goran Hasse of Raditex AB (http://www.raditex.se) - thanks, Goran, for your involvement! Below I describe my first impressions and experiences (with some more to follow in next few days). Well, to begin with the equipment characteristics: it's a 40MB silicon disk which is equipped with normal IDE and power supply connectors. I plugged it into the IDE bus, and BIOS recognized it without any problem as yet another IDE disk. Then I booted FreeBSD, and it also saw it as yet another IDE disk. I then created one slice on it, covering the whole disk (in so called "dedicated" mode - there's no sense to waste space for partition table). Then I went on with disklabeling it (with -B to make it bootable), and finally did newfs. All these steps went as usual. Then I installed slightly modified version of PicoBSD (no MFS in kernel - all binaries were taken from flash) on that IDE flash disk. This required modification of startup scripts, in order to mount the flash (/dev/wd0s1a) as /, and read-only. I'll probably modify this setup further to create after startup a small MFS and mount it on /tmp, and to mount flash read-write, but with noatime (to minimize write operations to the flash). All these operations I did on one machine, and then took out the flash disk and fitted it into another, test machine - an oldish 486DX with floppy, no HDD. (BTW. this begs quite another question of convenient bootstrapping the system using only one machine...). Then I rebooted, and withheld my breath.. :-) Well, in short: it was a total success, and it has shown some other advantages I wasn't quite aware of at that moment. First impression is, that it's blazingly fast - the whole booting process, from turning machine on to the login: prompt, takes ca. 30 seconds, and it can be further tuned down. The setup I tried uses less memory, because now it's possible to keep all binaries on the disk, and not on the ramdisk. The difference is quite substantial, something like 1-2MB of RAM :-). It was a funny story, indeed: at first I didn't notice this, but then I suddenly realized that I forgot to put all of those memory chips in the sockets, and I was running with only 4MB of RAM installed! The picobsd flavor I used was incidentally the "net" version, AND I was already running two sessions using full-blown /bin/sh, running inetd _and_ SNMPd in background! Isn't it amazing?! I can't wait to see what can be done with "router" version, which uses _so_ less memory than "net"... So, in the next few days I'm going to perform some other tests to see what are the capabilities and constraints of various setups. But, in short: it looks _very_ promising for embedded applications, where the 386SX/4MB RAM is a standard, and thus far it seems that you can run many things on such low-end equipment. Andrzej Bialecki -------------------- ++-------++ ------------------------------------- <abial@nask.pl> ||PicoBSD|| FreeBSD in your pocket? Go and see: Research & Academic |+-------+| "Small & Embedded FreeBSD" Network in Poland | |TT~~~| | http://www.freebsd.org/~picobsd/ -------------------- ~-+==---+-+ ------------------------------------- To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-small" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?Pine.BSF.4.02A.9809040930290.26306-100000>