Subject: QoSA2009 notification From: QoSA2009 Date: Thu, 19 Mar 2009 21:04:14 +0000 To: Aniruddha Gokhale Dear author, Congratulations - your paper has been accepted for publication and presentation at QoSA 2009. Please take into consideration the attached reviews in the final version of the paper. Final papers should not exceed 16 pages, must be written in English, and prepared according to Springer's LNCS style (guidelines are available at: http://www.springer.de/comp/lncs/authors.html). The paper must be personally presented at the COMPARCH 2009 Conference by one of the authors. A registration for your presenter must be received through the online registration by April 30th. The deadline for camera ready copies is April 11th. We will send you further instructions in the next week regarding final submission. cheers Raffaela and Ian --------------------------------------------- Paper: 31 Title: A Model-transformation Approach to Improving the Quality of Software Architectures for Distributed Real-time and Embedded Systems -------------------- review 1 -------------------- PAPER: 31 TITLE: A Model-transformation Approach to Improving the Quality of Software Architectures for Distributed Real-time and Embedded Systems Summary: This approach introduces a heuristic model-transformation algorithm to optimize the component allocation to middleware containers in distributed, real-time, and embedded (DRE) systems. The algorithm allocates components with similar quality-of-service properties to the same container, which can reduce the end-to-end latency of the resulting system significantly. Former approaches only optimized the component allocation to hardware nodes. The approach has been prototypically implemented for the Lightweight CORBA Component Middleware (LwCCM) in context of the Generic Modeling Environment (GME). The authors demonstrate the applicability of their approach in a case study involving a component-based system from the avionics domain and achieve a 70 percent end-to-end latency reduction. Comments: This is a well motivated and written paper about a topic highly relevant for QoSA. The authors introduce a fully implemented method and present a convincing case study, which clearly points out the possible improvements to the quality of service of a DRE system. The paper appears technically sound and original. I definitely recommend this paper for acceptance. While the paper is well written and easy to understand, some details of the approach remain unclear, as they are not explained in the paper. In line 12/13 and 17/18 of the transformation algorithm similar QoS policies of software components are combined. It is not clear when two QoS policies are similar and how they are combined. The paper would benefit if the author could give an example clarifying this step. It is unclear if the algorithm always terminates, because of the repeated optimizations. Furthermore I question the scalability of the approach. The authors mention DRE systems composed out of hundreds of components, but their case study only includes a system with 4 components. Is the approach still efficient when applied on large-scale systems? Minor remarks: - Fig. 2 contains connections from receptacle to receptacle 'get_data()', what does this mean? Normally you would connect a receptacle to a facet. - the authors use the term QoS policy multiple times before they explain it - the title of the paper is too generic and should be made more concrete -------------------- review 2 -------------------- PAPER: 31 TITLE: A Model-transformation Approach to Improving the Quality of Software Architectures for Distributed Real-time and Embedded Systems The paper presents a heuristic based approach addressing the issue of component allocation to different middleware containers, to reduce the end to end latency. The approach is evaluated in a reusable component assembly available in the avionics domain. Comments: * The paper is, in general, well written, although a bit long. * The reference to Real-Time systems seems to be quite artificial, as the approach does not require such a target platform. * It would be good to clarify the difference between a task and a component, as the real-time domain is addressed. * The use of the term “Quality” is used in a confusing way. The paper starts claiming to address the QoS requirements. Later on, the authors refer to, e.g., QoS configurations, QoS properties, QoS policies, QoS optimizations, Quality of software architectures etc. Moreover, the simulations only show that the end to end latency is reduced. It would be interesting to see if other “quality” factors can be improved as well. * The proposed approach is claimed to reduce the memory footprint of distributed real-time embedded systems. It is not clear to me how this is done (and by how much). Minor comments: * Title: to improving -> to improve (or for improving) * it is stated in the paper that the schedulability analysis determines the priorities- This is not true. Usually, the schedulability analysis decides whether a given priority assignment is feasible. -------------------- review 3 -------------------- PAPER: 31 TITLE: A Model-transformation Approach to Improving the Quality of Software Architectures for Distributed Real-time and Embedded Systems The paper presents a model-transformation-based algorithm to justify deployment and configuration decisions with respect to their implications on the system’s quality. Therefore, the authors firstly discuss impediments to the quality of distributed, real-time, and embedded systems. Building upon this discussion, the authors then present their approach to address impediments to the quality that are stemming from a sub-optimal deployment and configuration. The claimed merits of their approach are validated with a case study that documents significant improvements. The authors conclude with a classification of related work and a summary of the lessons learned during the research that underlies the development of the presented approach. Strengths: The paper builds upon a very clearly defined research methodology which spans a description of shortcomings, the presentation of an approach to overcome these shortcomings, as well as an empirical validation. The presented approach advances the state-of-the-art and seems to be able to improve deployment and configuration decisions for distributed, real-time, and embedded systems. Weaknesses: The authors should better show how their approach is superior to other related approaches (e.g. by empirically comparing related approaches and deepen the discussion of related work). In general, quality attributes of software architectures can be estimated by various prediction methods which at least allow identifying good architectural decisions and separating them from bad ones which deteriorate certain quality attributes. The authors, however, do not discuss these techniques at all. -------------------- review 4 -------------------- PAPER: 31 TITLE: A Model-transformation Approach to Improving the Quality of Software Architectures for Distributed Real-time and Embedded Systems __Summary__ The paper presents a heuristic algorithm to optimizing the doployment of the components of a distributed realtime system in order to minimize end-to-end latency and memory footprint. The algorithm is implemented as an automated model transformation based on graph rewriting. __Comments__ The key contribution of this submission is the heuristic algorithm presented at page nine. - Good structure, sufficient background material to make the paper self-contained - Model transformation is in the background, but could have been of interest to conference audience; interesting details on the implementation are completely omitted - since the algorithm is the core of the contribution, the Authors could have dedicated more space to explaining it more clearly. The description given does not provide sufficient insights on the high-level rationale of the design decisions; for example, it is explained that the algorithm aims at reducing the size of SQ_i, but it does not clarify why this is 'per se' a good thing from a QoS optimization perspective. The approach assumes that co-locating components with _similar_ (whatever that means) QoS policy goes in the direction of reducing latency; in my opinion the generality of this assumption is questionable, and without further arguments it cannot be taken for granted. - the presentation of the algorithm should have been completed by showing, by an example, how the algorithm changes an initial deployment plan and produces an improved configuration. - conclusions drawn after Sec. 4 are not sufficiently justified. Furthermore, the difference in the order of magnitude between inter-container and intra-container communication latency is not given, so the figures that result from the experiment (%improvement) are not easily generalizable. - It is not clear whether memory-footprint reduction was done by the heuristic or as a further step by means of PAM. The results of the experiment do not consider it anyway.