<<< Chronological Index >>>    <<< Thread Index >>>

tracerouteDB update


Hi All,

After Manuel's update of last week (fixing a bug the both of
us had overlooked in the source code), I have done another
extensive comparison of a merge_packets (assign routeid to packet)
run with old ascii and new sql DB.

In the ~7000 examined packets some differences were seen but
these are well understood; verification with the stored records
showed the new queries to be correct in all cases.

I will now move on with a last thorough check of the update process
and a comparison of a merge_packets run for a day where records have
all been create by the new update process. In this case the routeids
themselves cannot be compared (new vectors could get different ids
in the old and new code for processing RVEC files) but hopcounts
should be the same.

If both tests are succesfull, I hope to switch TTM production to the
SQL DB by the end of this week.

However, some other (less urgent) things still remain to be done:

- verbosity of updatedb: very verbose output only with debug flag
- packaging: no dependencies on /home/manuel

- documentation:  waiting for feedback from me
- cgi queries:    waiting for final tests / sw check by me

- performance:    once we've started production runs, we have to check
  if query performance is adequate or should be improved (e.g. by
  building more complex 'cache' that could result in less SQL queries)

  Also, we may have to enhance the cgi query to limit performance decrease
  of ginkgo by (unintentional) abusive queries.

-- Rene





<<< Chronological Index >>>    <<< Thread Index >>>