Grid Configuration and Application Monitoring in GridGain
Florin Pop, Maria-Alexandra Lovin,
Catalin Negru, Valentin Cristea
University POLITEHNICA of Bucharest, Romania
Faculty of Automatic Control and Computers
Emails: florin.pop@cs.pub.ro, alexandra.lovin@cti.pub.ro,
catalin.negru@cs.pub.ro, valentin.cristea@cs.pub.ro
Nik Bessis, Stelios Sotiriadis
School of Computing & Maths,
University of Derby, Derby, United Kingdom
Bucharest, Romania
Emails: n.bessis@derby.ac.uk, s.sotiriadis@derby.ac.uk
Abstract—Monitoring large scale distributed systems such
as Grids environments represents a means for obtaining a
quantitative and qualitative measurement of performance by
collecting information relevant to environment and applications
running on the Grid. This paper proposes a solution for
Grid monitoring on GridGain middleware platform using
both indirect information gathered inside the platform and
an addition tool based on MonALISA and ApMon. A useful
feature would be to allow the monitoring system to present
the operator with suggestions computed based on the history
of monitored parameters for jobs with a longer execution time
or based on some theoretical, model based assumptions that
relate existing configuration to the values of some performance
parameters.
Keywords-Distributed System, Monitoring, GridGain, Mon-
ALISA, Scheduling Service
I. I NTRODUCTION
Grid monitoring solutions focus on different directions
of monitoring: on configuration monitoring of the entire
structure of the Grid or per Grid node, on monitoring
applications running on the Grid or on a combination of
both configuration and applications performance monitoring.
Although starting from version 3, GridGain platform [1]
becomes oriented more towards cloud development, Grid
middleware remains an important feature of the platform and
thus the need for monitoring in order to evaluate and enhance
the performance of applications running and of resources
usage. The solution for GridGain follows a combined ap-
proach for both configuration and applications monitoring,
by gathering a large number of parameters of both types.
As previously proven on the existing monitoring solutions
for GridGain, information extracted from Grid metrics and
configuration available with GridGain API, GridGain SPIs
and other monitoring tools. Also, the analysis of the existing
Grid monitoring solutions indicates MonaLISA [2] and the
set of available tools as the solution that can be easily
integrated with GridGain.
The solution is designed for applications running on a
Linux/Unix environment and uses the Java implementation
part from MonALISA. Monitoring information is sent to a
central location from where it can be accessed by requesters
and further viewed in a separate module. According to the
taxonomy introduced in [3], the solution belongs to the
category of third level of monitoring tools, as a hierarchy
of publishers system, based on the provision characteristics,
complexity of components defined in Grid Monitoring Ar-
chitecture (GMA).
The remainder of this paper is organized as follows: Sec-
tion 2 presents the main issues regarding Grid monitoring;
Section 3 details the proposed solution at architectural level
and presents the technologies used and the way the different
modules are combined, with a small portion dedicated to
changes at monitoring level starting from version 3 of
GridGain; Section 4 focuses on the data visualization and
management layer of the solution, exploring the possibility
of analyzing the gathering information in order to make
configuration arrangement proposals; Section 5 tries to eval-
uate the solution based on the parameters gathered referring
to characteristics such as number, accuracy and relevance;
Sections 6 and 7 contain a number of possible development
and improvement directions, while Section 8 presents the
conclusions of this paper.
II. GRID MONITORING.STATE OF THE ART AND
SOLUTIONS
A traditional Grid monitoring process consists of four
stages [2] [21]. In the first phase, generated events lead to
the acquiring of measurements specific to the environment
and monitored application. The next phase involves process-
ing acquired data according to different criteria or groups
of events, followed by a distribution of the information
to requesters. Presentation is often organized in a GUI
application by using a stream of events or recorded trace
data stored as archive or database, a process known as
consumption. The data presented involves further processing
to extract conclusions. Grid monitoring systems need to
follow a certain schema for data representation and ongoing
research tries to establish a standard for the schema and
structure of description. Monitoring information or state-
ments are described by a time-stamp, indicating the moment
of time when the information was taken of the measurement
statement was true.
2012 Fourth International Conference on Intelligent Networking and Collaborative Systems
978-0-7695-4808-1/12 $26.00 © 2012 IEEE
DOI 10.1109/iNCoS.2012.82
155