softupdates && write cache && ata tags topic
Hello to all, I understand the basic concept of the folowing techs: softupdates, disk write cache and ata tags. My question is: It is safe to use softupdates + write cache + ata tags (IBM disk)? I read someware that it not safe to use softupdate + write cache (without ata tags) and if it is not safe why FreeBSD 5.0 ships with them enabled? Thanks very much, Nuno Teixeira -- /* PGP fingerprint: C6D1 06ED EB54 A99C 6B14 6732 0A5D 810D 727D F6C6 */ To Unsubscribe: send mail to [EMAIL PROTECTED] with "unsubscribe freebsd-current" in the body of the message
Re: softupdates && write cache && ata tags topic
Hi, > Is it safe to use softupdates + write cache + ata tags (IBM disk)? The summary of *my* experience and knowledge is: It is considered *unsave* to use Soft Updates with WriteCache enabled. I consider it unnecessary to use WriteCache if TaggedQueuing is enabled and working. (The performace gain of WriteCache and TaggedQueuing is more or less the same, the combination of both adds less than 10% of performance and you shouldn't use Soft Updates any more) Soft Updates alone add a huge amount of performance. So I recommend to use Soft Updates with Tagged Queuing enabled. -- Ciao/BSD - Matthias Matthias Schuendehuette , Berlin (Germany) Powered by FreeBSD 5.0-CURRENT To Unsubscribe: send mail to [EMAIL PROTECTED] with "unsubscribe freebsd-current" in the body of the message
Re: softupdates && write cache && ata tags topic
Nuno Teixeira wrote: > I understand the basic concept of the folowing techs: softupdates, disk > write cache and ata tags. > > My question is: > > It is safe to use softupdates + write cache + ata tags (IBM disk)? > > I read someware that it not safe to use softupdate + write cache > (without ata tags) and if it is not safe why FreeBSD 5.0 ships with them > enabled? See the discussion of 3 weeks ago with one of the engineers from Maxtor, on this list. The short answer is that the drive does not honor tags on a commit to stable storage, before returning the request as satisfied, and that's because on a write, ATA doesn't support disconnect with tagged commands. As to why write caching is on be default, there's a lot of going back and forth on that, and it's flipped state maybe 4-5 times, so far. The short answer is that ATA drives are pig-slow with write caching disabled, and users find that unacceptable. -- Terry To Unsubscribe: send mail to [EMAIL PROTECTED] with "unsubscribe freebsd-current" in the body of the message
Re: softupdates && write cache && ata tags topic
Hi, > > Write cacheing is automatically enabled if tagged queueing is enabled > and supported by the disk, so I doubt you're seeing any improvement at > all. I must admit: My statements are based on experience with SCSI Tagged Queuing and SCSI Write Cache. I hope I'm correct if I assume that the ATA features have not only the same name but the same functionality. I wasn't aware, that they are coupled by the driver code... Thanks for your commment/correction - Matthias To Unsubscribe: send mail to [EMAIL PROTECTED] with "unsubscribe freebsd-current" in the body of the message
Re: softupdates && write cache && ata tags topic
Matthias Schuendehuette <[EMAIL PROTECTED]> writes:
> I consider it unnecessary to use WriteCache if TaggedQueuing is enabled
> and working.
> (The performace gain of WriteCache and TaggedQueuing is more or less the
> same, the combination of both adds less than 10% of performance and you
> shouldn't use Soft Updates any more)
Write cacheing is automatically enabled if tagged queueing is enabled
and supported by the disk, so I doubt you're seeing any improvement at
all.
/* enable write caching if allowed and not default on device */
if (ata_wc || (ata_tags && ad_tagsupported(adp))) {
if (ata_command(atadev, ATA_C_SETFEATURES,
0, 0, ATA_C_F_ENAB_WCACHE, ATA_WAIT_INTR))
ata_prtdev(atadev, "enabling write cache failed\n");
}
DES
--
Dag-Erling Smorgrav - [EMAIL PROTECTED]
To Unsubscribe: send mail to [EMAIL PROTECTED]
with "unsubscribe freebsd-current" in the body of the message
