[94555] in North American Network Operators' Group
RE: [cacti-announce] Cacti 0.8.6j Released (fwd)
daemon@ATHENA.MIT.EDU (michael.dillon@bt.com)
Thu Jan 25 08:32:21 2007
Date: Thu, 25 Jan 2007 13:29:42 -0000
From: <michael.dillon@bt.com>
To: <nanog@merit.edu>
Errors-To: owner-nanog@merit.edu
> > But to start with, just solving the data storage problem is a good=20
> > place to start.=20
> How about something like:
> http://www.hdfgroup.org/whatishdf5.html
That certainly has a lot of support in the scientific community
in similar applications such as astronomy and high-energy
physics. However, if people don't understand the quirks of
dealing with large amounts of time-series data they will
probably look at HDF and consider it to be a poor out-of-date
cousin of the relational database. In fact, it is simply a
different approach to data storage which solves different
problems than relational databases do. A good network management
application would still use relational databases for some data
and only use HDF for the big time-series (SNMP poll data).
But HDF is still a general purpose format and it might be
possible to do better with a specialized time-series database
that has concepts such as deltas, missing values, or even
wavelets.
--Michael Dillon
