[Samba] Oplocks...again...

[John H Terpstra]

On Thu, 19 Dec 2002, William Jojo wrote:

>
>
> I would like to know if like the [homes] section mapped to a single user
> at login time (H: drive) would have any of the problems discussed in
> previous emails.
>
> Specifically the oplock break issue. Since one and only one user can have
> this space mapped in our  implementation (because of credential conflicts
> that windows does not allow), how could a user have a detrimental
> experience with oplocks?

It is the MS Windows client that decides whether or not to use file
caching based on the server's ability to handle it. Only Samba allows this
to be controlled on a per share basis.

The performance difference is most noticible where the same file is
repetitively accessed by the same application, but can be very significant
also for MS Excel and MS Word files because of the way that they implement
file access.

>
> Since only that user can open the files whether .xls, .mdb or .doc (don't
> mean to pick on M$-Office) there should be no issue, right?

In single user access the issues are likely to be minor and I would enable
oplocks. If the same user logs in multiple times and opens the same file
on each machine then there is still a risk.

>
> I ask primarily because I've had oplocks off for some time now, but am
> always wondering about performance increases - I'll take them where I can
> get them.

Try it! Let us know what differences you notice.

> My other question is regarding a read-only share. We serve *lots* of
> software from what we call our M: drive. How will oplocks and/or level2
> oplocks help there? I'm not sure I see how they will. I'm at the mercy of
> the randomness of served software, right?

If hte file has been cached in workstation memory it will open much faster
on second and subsequent access.

- John T.

> Any info on this will help.
>
> Thanks Samba Team - You've done a great job so far - it is greatly
> appreciated!
>
>
> Bill
>
>

-- 
John H Terpstra
Email: jht at samba.org