chukwa-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Eric Yang <>
Subject Re: priorities for the next release
Date Mon, 15 Jun 2009 17:35:46 GMT
Hi all,

Byts could mean bytes per second, bytes.  Incorrect assumptions could lead
to really wrong results.  First, let's set the ground rule that all metrics
output should standardize base unit of bytes, and milliseconds for
timestamp, and bytes per second for rate.  Metrics/column header should
append unit to the end.  For example:


By using the first principle rule, we could scrap the headers that don't
apply to the general rule, and replace their header with the common ones.

Once the data has been cleansed, the list of common units become readily
available for implementation.  It would be the time to write the automated
unit converter.  

Once automated conversion is done, database should be converted to use the
new format with a new set of dictionaries.  It's fair size of code, and I
think it would help if more developers participate to iron out each part.


On 6/14/09 9:23 PM, "Jiaqi Tan" <> wrote:

> Hi Ari,
> I think I have a parallel development roadmap for the next month
> (which is all the time I have left at Yahoo! unfortunately) for
> rolling out the visualization and diagnosis stuff, so my hands are
> pretty much tied up with that for now while I'm on payroll at Yahoo!.
> Mac? Eric? Any thoughts on this?
> Jiaqi
> On Sun, Jun 14, 2009 at 9:19 PM, Ariel Rabkin<> wrote:
>> Hi all.
>> It seems like 0.2 is going smoothly, and I think we ought to talk
>> about next steps at our 2pm phone call.
>> I'm currently working on CHUKWA-185, tailing a whole directory.
>> After that, my next priority is CHUKWA-25, adding cluster name to
>> archive schema.
>> What's everyone else working on?
>> Is anyone working on CHUKWA-260, the unit conversions?  I would vote
>> to make it a priority.  Jiaqi, you expressed enthusiasm for it before.
>> Do you have time to get to it soon?
>> We don't yet have an open issue for the timezone/time conversion
>> problems, but I think those should also be a priority. It would remove
>> a whole bunch of quirks from hicc.
>> --Ari
>> --
>> Ari Rabkin
>> UC Berkeley Computer Science Department

View raw message