[rtg] rtg status?

[Bryan Wann]

On Wed, 7 Jan 2009, Jacob Steinberger wrote:

> In all honesty, if I wanted to compress/average data based on time, I'd 
> use another product. I use RTG for two reasons - MySQL, and raw data. 
> I'd hate to see either go away, in any iteration.

On a side note, I'm interested in what other people are doing for table 
indexing/optimization.  I realize some people may optimize for graphing, 
others may optimize for something else like billing.

I started out with two KEYs, one based on 'id' and another based on 
'dtime'.  Trying to display large number of graphs (i.e. a loaded switch) 
simutaneously started to be unpossible by the time a table had 20-30 
million rows.  The optimizer was only using the 'id' column and then 
performing a filesort on the rest.  (select blah from ifOctets where id=x 
and dtime >y1 and dtime <y2 order by dtime).

I decided I should never have more than one row for a given interface and 
time.  I blew away my two KEYs and replaced them with "UNIQUE KEY 
(id,dtime)".  This got rid of the filesort and made graphs render an order 
of magnitude faster.



Met vriendelijke groet/kind regards,
bryan