Greg Meyer wrote:

On Friday 27 June 2003 02:26 pm, andre wrote:


On Friday 27 June 2003 01:45, Greg Meyer wrote:


On Thursday 26 June 2003 02:42 pm, Keld J�rn Simonsen wrote:


Anyway I am not interesed in mimicking what MS does, but in something
that is useful and convenient to the average MDK user.


I personally don't find this useful. When I delete a file, I want it to
be deleted. That, in combination with incremental rotating backups I do
every four hours with rsync give good production and online availability
of a file if deleted by accident.


Your stating that it is completely useless but have implemented a hack
which does exactly the same.



No, I think the difference is that I completely control when something gets deleted for good, and within two days, if I don't need the deleted files because of a mistake, they are gone. The system I see you all raving about could potentially leave files that were intended to be deleted hanging around for a very long time without the user being aware of it.

That would all depend on the policy you define for it. As a matter of fact: why not use the disk close to 100% all the time? The usage by the operational system may be _much_ lower, but the spare space can still be used for snapshots, deleted files, etc. etc. You just need a smart filesystem to do it.

I really don't see the benefits. I haven't really seen any compelling arguments for this other than people saying they like the idea. Why?



Hmmm... IMHO: I really love the way Network Appliance has implemented their snapshot and data recovery features in their (filer) products. It's what makes them rock.

Back in 1999 I managed storage for a 700 person company on a Sun E3000. At that time managing it was a total nightmare. Solaris didn't support group quota's (linux did at that time) and keeping projects (groups) from using too much space meant keeping them on a seperate partitions (or disks). With a fully loaded E3000, 3 storage arrays, some disks raided it still all added up to _many_ filesystems. Backups were a nightmare, and recovering files took a lot of time. Oh yeah, we were serving files out to the unix people with automounted NFS and to the windows people with SMB (Totalnet Advanced Server).

I would have _loved_ to loose the E3K and use a netapp filer instead. The netapp WAFL filesystem rules. A few years later the company did retire the E3K and moved to a netapp filer.

I must say, Microsoft is listening to their customers when it comes to this kind of functionality. I hope linux can provide the same kind of functionality one day, in the filesystem and without any sysadmin hacks... Perhaps with reiser v3,4,5 ?

regards,

Stefan

Attachment: smime.p7s
Description: S/MIME Cryptographic Signature

Reply via email to