Automation of decision making for monitoring systems

namibiancurrishInternet και Εφαρμογές Web

12 Νοε 2013 (πριν από 4 χρόνια και 1 μέρα)

72 εμφανίσεις


Włodzimierz Funika,
Filip Szura

Automation of decision making for
monitoring systems

Motivation


Main issue


Automation of
N
etwork monitoring


Rules


Understandable


Easy to create


External Engine (Jboss Drools)


Usage of external monitoring systems


Extension of existing solutions (Zabbix)


Actions described as:


Java code


Unix action scripts



System architecture

System divided

into

three
components:


Server


Decision component


Local
-
Monitor


Data collector


GUI


Presentation

Network 1, 2


monitored
networks

RMI (Remote Method Invocation)




standard communication between system components




System
architecture


Saude
-
Net GUI


Configuration provided during
work by the user


Web page generation,
communication over RMI


Saude
-
Net Server


Configuration stored in XML


Rules stored in DRL file


Rules engine


Saude
-
Net Local
-
Monitor


Dependent on external
monitoring system


XML configuration


Connection over RMI


No actions provided for user
during work


System details

Saude
-
Net(
S
ystem for
au
tomation of
de
cision making for
monitoring systems)


Designed for small and medium size networks


Multiple actions for resources under monitoring


Fully configurable


Rule and actions management is easy for network
administrator


System choose the best action when event occurred in the
monitored network


Can manage multiple networks





Data From
Monitoring
Tools

Demo

List of
available

actions

Trigger the best
action

GUI
component

System details


Preference

tuning value (PTV)





Preference

value (PV)



Ns
-

number of services which are currently

modified

D
-

number of instances of any action

Iw
-

sum of instances of a speci
f
ed action as the best (+1) and other

(
-
1)


PTV
=
Ns
D
x
const
x
Iw
D
PV = oldPV + PTV

System features


Classic monitoring systems


Network events are reported to
netw
o
rk

administrator


Our solution


System automatically react on reported failures


Reaction based on actual network snap
-
shot and knowledge


In effect, system is able to perform any action when
network event occur
r
ed


Action described as
Uni
x

action script


Short Java script located in rule

Implementation

Server
component


Calculating preferences for actions


Invoke actions


Stores actual view of network and actual
knowledge base


Validation of rules

Local
-
Monitor component


Acquir
ing

data from external monitoring
systems (Zabbix)


Send
ing

data to Saude
-
Net server

GUI
component


Create rules


Simple validation


Data representation


Form generation

Presentation component


Extension of Local
-
Monitor


Simple simulator of network

Implementation


Used technologies:


Java


available for most operating systems, the same data representation in different
systems


JSF


less code than in JSP


Jboss Drools


easy to use, available for java


XML


required by Spring


Spring Framework


allows for Dependency Injection, provides a connection to
database


Component architecture


System divided into separate components


Web
-
based GUI


System management by web pages



Conclusions


Achievements


Automatic actions on network failures


Actions described by rules


Easy configuration of rules and actions


Distributed architecture


Limitations


All components must be implemented in Java


Configuration may not be changed during system work


Rules are created only by administrator, our system is not allowed to create
its own rules.


Future work


System should learn automatically from actions taken by the administrator


Extensions of Saude
-
Net system security


Redundancy of Saude
-
Net server component

More details


Poster number: 6


Thank you for your attention
!