[rtg] RTG Indexing

[bill fumerola]

On Fri, Jan 09, 2009 at 10:24:34AM -0700, Leech, Jonathan wrote:
> The impact to the database schema and indexing was that it let us
> flatten our database schema into something like:
> 
> id time ifinoctets ifoutoctets ifindiscards ifinerrors ...
> ------------------------------------------------------------
> 
> which results in many fewer database objects, more efficient indexing,
> easier summarization processes (we do hourly and daily), easier database
> maintentance including purging old data, and simplified querying to
> build summary queries across interfaces and routers.

building a generic configuration interface that lets you stuff multiple
values into a row would be a great addition to rtg. wouldn't be that
hard to implement in the lex/yacc grammar.

building a ifXTable/ifTable specific hack would be beneficial for how
most people use RTG, but as it can be used to poll any SNMP object with
an integral or decimal value a generic version would work better.

> Also we've built a driver for JRTG to write data to RRD4J (similar to
> the RRD files in MRTG). These naturally solve the problem of truncating
> and summarizing data. This code will be released to the JRTG project
> soon.

RRD support for RTG would be nice as well. again, we're overlapping w/
the Cacti functionality -- it'd be nice to consolidate.

collapsing the JRTG and RTG2 projects, allowing virtela to maintain the
java port within the RTG2 project, is also something to seriously consider.

that way, overlaps can be consolidated, pollers and graphers could be
versioned and mixing and matching (java poller, php grapher, etc) is
possible.

create a middleware cacher (sits between poller/grapher and database),
add in thrift support (http://incubator.apache.org/thrift/) for the
interprocess communication, all of the sudden we have a fully featured
product.

dare to dream,
-- bill