Re: Concurrent access by writer and readers

• To: netcdf-hdf@xxxxxxxxxxxxxxxx
• Subject: Re: Concurrent access by writer and readers
• From: Russ Rew <russ@xxxxxxxxxxxxxxxx>
• Date: Fri, 23 Apr 2004 12:56:03 -0600

Quincey,

>     This sort of concurrency is not supported by default, but it should be
> possible to achieve it with sufficient tweaking of the caching parameters.
> You can use H5Pset_sieve_buf_size() to turn off raw data caching and you
> can use H5Pset_cache() to turn off metadata caching also.  Obviously,
> performance is not great in these scenarios, but I think it will work.
>     If we want to recover some of the performance given up by these sort of
> tweaks, we could change the internal caches to allow write-through instead
> of write-back caching, which would probably recover a significant chunk of
> the slowdown.

Thanks for the info!  Although we don't test concurrent access in our
normal testing program (nc_test), Steve Emmerson has written a couple
of programs that we can use to test this, which should help us see if
we can get the caching parameters right.  Making it work correctly is
much higher priority than getting good performance at this stage.

--Russ

• References:
  • Re: Concurrent access by writer and readers
    • From: Quincey Koziol