NFS soft hard performance difference

NFS soft hard performance difference

Post by Richard Aulet » Fri, 28 Sep 2001 06:30:55



I am well aware of the difference and implication of choosing
soft versus * an NFS mount. But I was curious in addition
to the myrid reasons not to use soft, might soft NFS mounts also
show a reduction in performance? It would seem to me, that soft
would require more processing overhead, but sufficient to be
noticable?

-Rich Auletta

 
 
 

NFS soft hard performance difference

Post by Barry Margoli » Fri, 28 Sep 2001 06:48:00




>I am well aware of the difference and implication of choosing
>soft versus * an NFS mount. But I was curious in addition
>to the myrid reasons not to use soft, might soft NFS mounts also
>show a reduction in performance? It would seem to me, that soft
>would require more processing overhead, but sufficient to be
>noticable?

Off the top of my head I don't think it should make a noticeable
difference.  Soft mounts eventually stop retransmitting, but by that time I
think the exponential backoff results in the retries being pretty far
apart, so they're not really affecting much.

--

Genuity, Woburn, MA
*** DON'T SEND TECHNICAL QUESTIONS DIRECTLY TO ME, post them to newsgroups.
Please DON'T copy followups to me -- I'll assume it wasn't posted to the group.

 
 
 

NFS soft hard performance difference

Post by Casper H.S. Dik - Network Security Engine » Fri, 28 Sep 2001 17:59:36


[[ PLEASE DON'T SEND ME EMAIL COPIES OF POSTINGS ]]


>I am well aware of the difference and implication of choosing
>soft versus * an NFS mount. But I was curious in addition
>to the myrid reasons not to use soft, might soft NFS mounts also
>show a reduction in performance? It would seem to me, that soft
>would require more processing overhead, but sufficient to be
>noticable?

No, that is not true.  Soft and hard mounts only make a difference at
the retry level; NFS operations will fail for both hard and soft mounts;
hard mounts will just continue to press on where soft mounts give up.

Casper
--
Expressed in this posting are my opinions.  They are in no way related
to opinions held by my employer, Sun Microsystems.
Statements on Sun products included here are not gospel and may
be fiction rather than truth.

 
 
 

NFS soft hard performance difference

Post by Richard Aulet » Sat, 29 Sep 2001 00:21:06


> >I am well aware of the difference and implication of choosing
> >soft versus * an NFS mount. But I was curious in addition
> >to the myrid reasons not to use soft, might soft NFS mounts also
> >show a reduction in performance? It would seem to me, that soft
> >would require more processing overhead, but sufficient to be
> >noticable?

> No, that is not true.  Soft and hard mounts only make a difference at
> the retry level; NFS operations will fail for both hard and soft mounts;
> hard mounts will just continue to press on where soft mounts give up.

> Off the top of my head I don't think it should make a noticeable
> difference.  Soft mounts eventually stop retransmitting, but by that time I
> think the exponential backoff results in the retries being pretty far
>apart, so they're not really affecting much.

Thanks for the responses, I recently discovered at the company I work
for
(whose name will go unmentioned) that almost all the user/design and
executable
NFS mounts are soft. It is amazing what you can find out by unplugging
an ethernet connection and watching a program error out :) I was
looking for more ammunition to get the sys admins to change everything
over, just not my accounts.

I believe the root problem is the Solaris mount_nfs manual page reads
incorrectly.

 hard | soft
           Return an error if the server  does  not
           respond,  or  continue the retry request
           until the server responds.  The  default
           value is hard.

Assuming normal latter/former order hard would correspond to
"Return an error if the server  does  not respond" and
"continue the retry request until the server responds" to soft. I
remember
scrathing my head over this  ... say 1990 ... on a vax? :)

In the sys admin guide there is an excellent discussion of hard and
soft,
and even later on the man page it clearly says not to use soft and why
...
but that little blurb to any newbie reads "wow I need to change to
soft!".

-Rich

 
 
 

NFS soft hard performance difference

Post by Richard L. Hamilt » Sat, 29 Sep 2001 16:57:24




>> >I am well aware of the difference and implication of choosing
>> >soft versus * an NFS mount. But I was curious in addition
>> >to the myrid reasons not to use soft, might soft NFS mounts also
>> >show a reduction in performance? It would seem to me, that soft
>> >would require more processing overhead, but sufficient to be
>> >noticable?


>> No, that is not true.  Soft and hard mounts only make a difference at
>> the retry level; NFS operations will fail for both hard and soft mounts;
>> hard mounts will just continue to press on where soft mounts give up.


>> Off the top of my head I don't think it should make a noticeable
>> difference.  Soft mounts eventually stop retransmitting, but by that time I
>> think the exponential backoff results in the retries being pretty far
>>apart, so they're not really affecting much.

> Thanks for the responses, I recently discovered at the company I work
> for
> (whose name will go unmentioned) that almost all the user/design and
> executable
> NFS mounts are soft. It is amazing what you can find out by unplugging
> an ethernet connection and watching a program error out :) I was
> looking for more ammunition to get the sys admins to change everything
> over, just not my accounts.

> I believe the root problem is the Solaris mount_nfs manual page reads
> incorrectly.

>  hard | soft
>            Return an error if the server  does  not
>            respond,  or  continue the retry request
>            until the server responds.  The  default
>            value is hard.

> Assuming normal latter/former order hard would correspond to
> "Return an error if the server  does  not respond" and
> "continue the retry request until the server responds" to soft. I
> remember
> scrathing my head over this  ... say 1990 ... on a vax? :)

> In the sys admin guide there is an excellent discussion of hard and
> soft,
> and even later on the man page it clearly says not to use soft and why
> ...
> but that little blurb to any newbie reads "wow I need to change to
> soft!".

Typically, you want "hard,intr" in place of soft.  Otherwise, processes
with hung hard accesses will be unkillable.

--

 
 
 

NFS soft hard performance difference

Post by Casper H.S. Dik - Network Security Engine » Sat, 29 Sep 2001 17:41:33


[[ PLEASE DON'T SEND ME EMAIL COPIES OF POSTINGS ]]


Quote:>Typically, you want "hard,intr" in place of soft.  Otherwise, processes
>with hung hard accesses will be unkillable.

That's why "hard,intr" is the default in Solaris.

(Wasn't in SunOS 4, I think)

Casper
--
Expressed in this posting are my opinions.  They are in no way related
to opinions held by my employer, Sun Microsystems.
Statements on Sun products included here are not gospel and may
be fiction rather than truth.

 
 
 

1. What is the difference between soft link and hard link?

Hi..
   I want to know what is the difference between soft link and
hard , and if they will use inode ?
   Where are the documents about this?
   Thanks a lot.

--

Department of Computer & Information Science
National Chiao-Tung University                            
Hsinchu, Taiwan, R.O.C.

2. Acer monitor--what crap

3. NFS: soft vs hard mounts

4. PCMCIA and IRDA support for open server 5

5. Soft mount VS hard mount (NFS)

6. ldconfig - ZMAGIC???? (solved)

7. NFS breaking Soft and Hard links

8. Strange Linksys EtherFast 10/100 PC Card Problem

9. nfs mount: soft or hard?

10. nfs soft/hard confirmation?

11. Poor NFS write performance from Linux clients to NFS Ver 3 servers.

12. Problem: performance NFS V3 vs NFS V2

13. Performance of soft RAID0 on Solaris 10