From bures at d3s.mff.cuni.cz  Tue Jun  3 13:32:02 2014
From: bures at d3s.mff.cuni.cz (Tomas Bures)
Date: Tue, 03 Jun 2014 13:32:02 +0200
Subject: Seminar - tomorrow 9:00
Message-ID: <538DB232.5020108@d3s.mff.cuni.cz>

Dear all,

for the next two months, we host a RELATE student for a secondment. His 
name is Inti Gonzalez. He is focusing on resource reservation and 
quotas. He is going to give a seminar this Wednesday 9:00.

Primarily, the seminar is meant for the component group, however, I 
believe that others may benefit and provide valuable feedback as well - 
especially the systems group.


2014-06-04 09:00
Inti Gonzalez-Herrera (IRISA, France): Resource reservation in pervasive 
Java-based middleware

Resource management is critical for application domains where components 
share their execution environments but belong to different stakeholders, 
such as smart homes or cloud systems. Yet, current middleware and 
application containers often hide system-level details needed for 
dynamic resource management. In particular, they tend to hide resource 
usage and offer automatic management instead (e.g., CPU, memory and 
I/O). In contrast, system-level containers, such as Linux Containers 
(LXC), allow fine-grain resource management. However, they lack 
knowledge about the application’s structure and its requirements in 
order to provide well tuned resource management. In this paper, we 
propose a flexible and efficient approach to resource management that 
takes advantage of the application’s structure and requirements to 
deploy components on system-level containers to manage resources. Our 
approach follows the models at runtime paradigm, which simplifies the 
development of self-adaptive systems and captures structural information 
of the application. In Squirrel, the application’s model is augmented 
with resource management contracts that are used at runtime to drive 
system-level containers and enforce resource management. We validate 
Squirrel’s feasibility and show its overhead regarding communication, 
CPU/memory consumption, and adaptation performance. The results 
demonstrate negligible impact on performance and only slight memory 
overhead when comparing a resource managed application to its original 
version.



Thank you. Best regards,
Tomas

-- 
Tomas Bures, Ph.D.
Associate Professor
Department of Distributed and Dependable Systems
Charles University
Malostranske nam.25
11800 Prague 1, Czech Republic
http://d3s.mff.cuni.cz
Phone: (+420) 2 2191 4236
Fax:   (+420) 2 2191 4323