Torrus: The Data Series Processing Framework
In short, Torrus is an alternative software platform to
MRTG,
Cricket and
Cacti. In most cases it brings more
flexibility and performance. We are able to poll more than 50,000 SNMP
OIDs every 5 minutes from a single moderately equipped server,
with simultaneous threshold monitoring of half of them.
Torrus is designed to be the universal data series processing
framework. Its scalable hierarchical design, application-independent core,
and highly customizable architecture make Torrus an attractive choice
both for small installations and for big enterprise or carrier networks.
Although most of our users deploy Torrus for SNMP monitoring, it might be
useful for data series of any nature. Tobi Oetiker's RRDtool is used for
data storage.
See the Torrus demo
server, provided by
Bluewin AG.
See also
Torrus Wiki
for useful information and tips.
Torrus functional overview slideshow gives some conceptional
and quick explanations (Open Office
presentation, PDF)
Torrus presentation at SwiNOG meeting tells where we are
today (Open Office
presentation, PDF)
Currently Torrus consists of the following parts:
-
Configuration compiler and validator.
It processes the XML configuration files and saves the configuration
into a database.
-
View renderer and the web interface.
They generate HTML and the graphical representation of the datasources
and provide user authentication and authorization.
All generated output is controlled by the configuration parameters
and templates.
The users can easily create their own presentation of data series.
-
SNMP Collector.
Thanks to David M. Town's
Net::SNMP Perl module, Torrus has got an extremely efficient SNMP
data poller, with ability to poll dozens of thousands of SNMP variables
from a single server. Modular collector core architecture allows
further extension with new collector and storage types.
Any datasource can have its own polling schedule.
-
SNMP Device Discovery Tool.
Devdiscover is a new, modular, flexible, and expandable tool for
automatic generation of Torrus configuration files. New device types
and MIBs are easily added as independent Perl modules. Thanks to our
contributors, the number of supported devices and vendors is constantly
growing.
-
Billing reports generator.
Torrus can be configured to store the network traffic usage in an SQL
database (without RRDTools' aggregation and interpolation), and produce
monthly reports on traffic volume and 95th Percentile bandwidth usage.
-
Threshold monitor.
All data, regardless of their type and nature,
can be monitored according to the user-defined rules. The rules can also
include the datasource-specific parameters, e.g. boundary values etc.
The following types of action may be combined for event reporting:
1) Displaying the list of alarmed datasources; 2) E-mail notification;
3) SNMP trap; 4) Any custom executable. Thresholds are specified by RPN
expressions.
See the Project
summary and Download Area at SourceForge.Net.
Current development
snapshots are available for testing.
Documentation
Features of Torrus
-
Data-oriented design. Most of the existing frontends are SNMP-centric.
In Torrus, the SNMP collector is not the central part,
but rather one of possible data sources. It is easy to integrate
some existing RRD files that are updated by some external agents.
-
Multiple independent datasource trees in one instance
of Torrus.
-
User authentication and authorization. Users belong to groups,
and each group has privileges to browse a certain subset of trees.
-
Flexible hierarchical configuration in XML format. Multuiple XML
files are processed additively.
-
Externally generated RRD files can be automatically grouped,
based on the filename regular expressions.
-
Flexible representation of the data. One datasource may be displayed in
multiple subtrees of the hierarchy. It allows you to group them at your
convenience.
-
Aberrant behaviour detection based on Holt-Winters prediction
is fully supported (requires rrdtool version 1.1.x). See the
references below for more information.
-
Support for international character sets (Latin1 and Unicode).
-
First open-source complete
implementation
of
Cisco QoS (CISCO-CLASS-BASED-QOS-MIB) discovery and monitoring.
Up-to-date list of ongoing tasks is published in
TODO document.
Product Support
Torrus is free software. It is distributed under the
GNU General
Public License.
Please send all your requests, comments, and bug reports to the
torrus-users mailing list. Only list subscribers are allowed to
post the messages there. All new release announcements are sent to this list.
The development issues of Torrus are discussed in
torrus-devel mailing list.
Copies of all CVS commits are posted to
torrus-commit mailing list.
Archives of the old mailing lists are available for
rrfw-users and
rrfw-devel .
Commercial support for Torrus is provided by:
K-Open GmbH
Switzerland
Tel. +41 79 407 0224
Commercial support includes:
- Software installation and customizing
- Development of custom user data import utilities
- Development of new Torrus features
- Software maintenance contracts
For more details about commercial support, please contact us
by email.
SNMP MIBs
By courtesy of Tobi Oetiker, we are using the subtree of
RRDtool Enterprise ID.
Torrus is now capable of sending the SNMPv2c traps, with the variables
described in the corresponding MIB file.
References
-
Torrus was the new name for RRFW, the Round Robin Database
Framework. The changes are summarized in the
transition roadmap.
Old RRFW documentation is available at
RRFW home page.
-
Development of RRFW releases 0.0.1 to 0.0.7, as well as development and testing
of Cisco class-based QoS monitoring was sponsored by
Cablecom GmbH
Internet Core Routing Services Team.
Cablecom is the leading country-wide cable-TV and Internet
provider in Switzerland.
-
This project would not even start without
Tobi Oetiker's RRDTool.
RRD is a very efficient and popular system to store and
display time-series data.
-
Many concepts and ideas were taken from
Cricket.
Cricket is a widely used and very efficient network monitoring system.
However, its limitations were actually the reason to start this project.
-
Jake Brutlag.
Aberrant Behavior Detection in Time Series for Network Service Monitoring.
LISA 2000 paper.
The article gives the basic theory overview of Holt-Winters forecasting
algorithm, and describes its implementation in RRDTool.
-
Jake Brutlag.
Notes on RRDTOOL implementation of Aberrant Behavior Detection.
The article describes the implementation of
aberrant behavior detection for RRDTOOL.
Credits
Companies that use Torrus
Copyright © 2002-2007 Stanislav Sinyagin
<ssinyagin@yahoo.com>