[OpenAFS] RE: mkdir() performance on AFS client

2014-03-31 Thread Robert Milkowski
Update - Oracle has implemented a similar optimization as Linux has, so now VFS layer will check if an object already exists and if it does it won't even call a fs-specific callback. AFS and NFS benefit most. This is in Solaris 11 + SRU17, and will be part of 11.2 release as well. > -Origin

Re: [OpenAFS] Re: mkdir() performance on AFS client

2014-01-24 Thread Jeffrey Altman
On 1/24/2014 8:56 AM, mi...@task.gda.pl wrote: >> >> The Windows cache manager even takes things a step further by >> maintaining a negative cache for EACCESS errors on {FID, user}. This >> has avoided hitting the abort threshold limits triggered by Windows >> that >> assumes that if it can list a

RE: [OpenAFS] Re: mkdir() performance on AFS client

2014-01-24 Thread milek
> > The Windows cache manager even takes things a step further by > maintaining a negative cache for EACCESS errors on {FID, user}. This > has avoided hitting the abort threshold limits triggered by Windows > that > assumes that if it can list a directory it must be able to read the > status of a

Re: [OpenAFS] Re: mkdir() performance on AFS client

2014-01-23 Thread Jeffrey Altman
On 1/23/2014 11:53 AM, Andrew Deason wrote: > On Thu, 23 Jan 2014 11:43:50 + > Simon Wilkinson wrote: > >> The real question here is how widely we should be applying the abort >> threshold - should it apply to all aborts sent to a particular client, >> or should we be more selective? There ar

[OpenAFS] Re: mkdir() performance on AFS client

2014-01-23 Thread Andrew Deason
On Thu, 23 Jan 2014 11:43:50 + Simon Wilkinson wrote: > The real question here is how widely we should be applying the abort > threshold - should it apply to all aborts sent to a particular client, > or should we be more selective? There are a lot of competing views > here, as it depends on w