GSOC IRC Meeting Report for 2012-05-17

[Chris Johns]

On 19/05/12 5:51 AM, Sebastian Huber wrote:
> On 18/05/12 20:23, Joel Sherrill wrote:
> [...]
>> * bdbuf - Community and mentors assist with Bdbuf statistics definition
>
> Since bdbuf is simply a cache for blocks of memory the general cache
> statistics are interesting:
>
> http://en.wikipedia.org/wiki/CPU_cache#Cache_miss
>
> They should be per disk (rtems_disk_device).
>
> The cold misses can be determined with a special driver in the benchmarks.
>

I think there is more in bdbuf we should consider monitoring. The buffer 
resizing and how often that happens. Also the swap out threads, the work 
they do, the time a buffer stays in the caches, maybe more.

I am wondering if we should add some hooks to allow monitoring. This way 
users do not have to rebuild the code to add monitoring functionality 
and the code does not need to change to accommodate evolving monitoring 
needs.

>> * bdbuf - Community and mentors assist with Bdbuf benchmark/reference
>> scenarios
>
> We should start simple
> 1. reading of a large file (large compared to the cache size),
> 2. writing of a large file, and
> 3. random file operations (see libtests/flashdisk01).
>

Is this against different systems ?

Chris