This is the mail archive of the
libc-alpha@sourceware.org
mailing list for the glibc project.
Where do we store whole-system benchmark data for glibc?
- From: "Carlos O'Donell" <carlos at redhat dot com>
- To: OndÅej BÃlka <neleai at seznam dot cz>
- Cc: libc-alpha at sourceware dot org, Torvald Riegel <triegel at redhat dot com>
- Date: Tue, 09 Apr 2013 16:24:25 -0400
- Subject: Where do we store whole-system benchmark data for glibc?
- References: <20130409192249 dot GA11662 at domone dot kolej dot mff dot cuni dot cz>
On 04/09/2013 03:22 PM, OndÅej BÃlka wrote:
> Hi,
>
> Dryrun is framework to supply real world data for microbenchmarks
> and analyze workloads.
>
> To keep first commit simple I only added code to handle exp.
> Others will follow.
>
> One question is where to save data? For memset I produced 20MB file
> so adding to git is out of question.
> Is possible http/ftp at sourceware where we could upload these files.
>
> Another issue leaking information. So far I for memset/memcpy/strlen
> need only to know positions and I doubt that somebody can reconstruct
> from it password.
>
> Compiling directory with given function creates three files:
>
> record.so: To record function calls write 'LD_PRELOAD=dir/record.so command'
> or add record.so to /ld.so.preload for global recording.
>
> replay: Calls function as recorded.
>
> show: displays recorded data.
I've already replied to this email, earlier and asked you to look
into leveraging systemtap to create a more powerful framework.
I'm responding again to touch on an orthogonal topic, that is
what do we do with whole system benchmark data that may be from 1MB
to 1000MB in size.
We can't store that data on sourceware.org.
We need to come up with somewhere to store the data.
Amazon EC2?
We need a secure way for users to upload the data (along with
whatever metadata they wanted to tag it with).
Comments?
Cheers,
Carlos.