Why a file system required in an embedded Device
in [Embedded]
|
Prev: USB Enumeration problem
Next: malloc/free algorithms
From: Vladimir Vassilevsky on 24 Feb 2010 10:59 Jalon wrote: > Hi, > > I have this question in mind, Why do we need a file system as part of any > RTOS when it is intended for a target embedded device(Low memory and less > storage of data). > I can imagine a system having lots of data managed in terms of files in > need of File system but probably not many embedded devices store that > amount of data. > > Is there a particular need/scenario which makes file system required in an > embedded system. Nothing is requred, except for CPU and power supply. Everything else is at your discretion. File system is just a matter of convenience if you need to organize your data. VLV
From: Mark Borgerson on 24 Feb 2010 12:00 In article <88t9o5dj4ooduevlinf9sh10qaajfdpcms(a)4ax.com>, keinanen(a)sci.fi says... > On Wed, 24 Feb 2010 00:12:25 -0600, "Jalon" > <jpeeyush(a)n_o_s_p_a_m.gmail.com> wrote: > > >Hi, > > > >I have this question in mind, Why do we need a file system as part of any > >RTOS when it is intended for a target embedded device(Low memory and less > >storage of data). > >I can imagine a system having lots of data managed in terms of files in > >need of File system but probably not many embedded devices store that > >amount of data. > > > >Is there a particular need/scenario which makes file system required in an > >embedded system. > > Most real time programmers would like to stay away from any kinds of > mechanical disk drives if possible and hence there is not much of a > need for a file system. > > In many RT systems, disk/flash file systems might be usable in the > startup phase to load parameters etc, before the actual RT work > starts. > > The null task is also a usable place in doing non-deterministic > disk-I/O. > > In actual RT processing, you simply can not use blocking I/O requests > such as typical read/write, but use requests like "read sector 6 track > 56", then perform dozens of other operations and at some later time > after hundred of other operations have been performed, handle the > callback or interrupt service routine that actually returned the > requested data. > > While a file system might be usable for loading startup parameters > from a disk/flash and hence require a file system, those systems are a > nuisance for actual RT operations. > I've built dozens of data loggers that collect oceanographic data at 120Hz and write it to a hard disk drive. These are are real-time, multi-processor systems with one processor dedicated to handling the storage chores and the other handling the 120Hz data collection. Coping with the non-deterministic file system does require lots of buffer space and lots of testing. The real tough part is keeping the average power consumption under 15mA at 7.2V. Hey---if it was easy anyone could do it and I'd be out of a job! Mark Borgerson
From: hamilton on 24 Feb 2010 16:33 Mark Borgerson wrote: > I've built dozens of data loggers that collect oceanographic data at Do you have a web site ?
From: Mark Borgerson on 24 Feb 2010 17:52
In article <hm45up$oaa$1(a)news.eternal-september.org>, hamilton(a)nothere.com says... > Mark Borgerson wrote: > > > I've built dozens of data loggers that collect oceanographic data at > > Do you have a web site ? > www.oes.to The web site show mainly interface boards for Persistor CF-2 loggers that are offered for general sale. They are used by oceanographers and others for data collection. The custom loggers that use hard disk or SDHC cards for storage were developed for a single research group and haven't been advertised or sold to others. The data the loggers collect is discussed in this paper: http://mixing.coas.oregonstate.edu/papers/mixing_measurements.pdf Mark Borgerson |