Applying the P2P paradigm to management of large-scale distributed networks using a Model Driven Approach

Conference Paper (PDF Available) · January 2006with 39 Reads
DOI: 10.1109/NOMS.2006.1687605 · Source: DBLP
Conference: Management of Integrated End-to-End Communications and Services, 10th IEEE/IFIP Network Operations and Management Symposium, NOMS 2006, Vancouver, Canada, April 3-7, 2006. Proceedings
Cite this publication
Abstract
This paper details an application of model-driven development undertaken within the Celtic initiative project Madeira. The objective of Madeira is to apply model-driven approaches to investigate large-scale distribution techniques in network management. So far, the project has concentrated on building a prototype system using the peer-to-peer (P2P) paradigm. For this, one of the major challenges was to provide a model supporting P2P characteristics, such as 1) self-organisation, 2) symmetric communication and 3) distributed control and domain specific concepts for distribution and management as can be found in telecommunications. Our modelling approach had to consider the views of the different participants, such as equipment vendor (Ericsson and Siemens), network operator (BT) and service provider (BT, Telefonica). To minimise the complexity of the model, we have focused on fault and configuration management in a dynamically forming network of transient elements. This paper explains the complexity of the networks we consider and of the management tasks we have to cover. This can be briefly characterized by comprising a multitude of different, sometimes proprietary, technologies and diverse business models. We motivated the application of new paradigms, in our case the P2P paradigm, to simplify management for seamless service provision to customers. Based on this, we provided a case study of how we applied a model-driven approach to capture the complexity of the task and the complexity of the management activities, which ultimately led towards the specification of management information and behaviour of network nodes. The methodology we use is presented in the form of a 'vertical slice', where a logical portion of the project, of limited scope and functionality, is brought from the meta-level right through to the development stage, covering all modelling and architectural work as well as the underlying platform aspects
Network Management Based on Distributed Paradigms
Applying the P2P paradigm to management of
large-scale distributed networks using a Model
Driven Approach
Ray Carroll1, Claire Fahy1, Elyes Lehtihet1, Sven van der Meer1, Nektarios Georgalas2, David Cleary3
1Waterford Instit ute of Technol ogy,
Telecommunica tions Software an d System Group,
Cork Road, Waterford, Ireland.
{rcar roll, vdm eer, eleht ihet , cfa hy}@ tssg.o rg
2BT Group
ICT Futures Research Centre,
Orion Building - Groun d Floor pp13
Adastral Park,
Martlesham Heath
nektarios.ge orgalas@b t.com
3Ericsson Ire land,
Athlone, Ireland
david.clear y@ericsson.c om
Abstract
This paper details an application of model-driven development undertaken within the Celtic
initiative project Madeira. The objective of Madeira is to apply model-driven approaches to
investigate large-scale distribution techniques in network management. So far, the project has
concentrated on building a prototype system using the peer-to-peer (P2P) paradigm. For this one
of the major challenges was to provide a model supporting P2P characteristics, such as 1) self-
organisation, 2) symmetric communication and 3) distributed control and domain specific
concepts for distribution and management as can be found in telecommunications. Our
modelling approach had to consider the views of the different participants, such as equipment
vendor (Ericsson and Siemens), network operator (BT) and service provider (BT, Telefonica).
To minimise the complexity of the model, we have focused on fault and configuration
management in a dyna mically forming network of transient elements.
This paper will explain the complexity of the networks we consider and of the management tasks
we have to cover. This can be briefly characterized by comprising a multitude of different,
sometimes proprietary, technologies and diverse business models. We will motivate the
application of new paradigms, in our case the P2P paradigm, to simplify management for
seamless service provision to customers. Based on this, we will provide a case study of how we
applied a model-driven approach to capture the complexity of the task and the complexity of the
management activities, which ultimately led towards the specification of management
information and behaviour of network nodes. The methodology we use is presented in the form
of a ‘vertical slice’, where a logical portion of the project, of limited scope and functionality, is
brought from the meta-level right through to the development stage, covering all modelling and
architectural work as well as the underlying platform aspects.
Network Management Based on Distributed Paradigms
Overview
Madeira Vision
Madeira Architecture
Modelling Approach
Vertical Slice
Overall Modelling Initiative
MDA Concerns
Meta Layers
Manual Mapping
Example
MDA
Conclusions
References and Future Work
This paper’s aim is twofold. Firstly we will describe our experience in adopting a
model driven approach for the system design using manual mappings between
modelling layers. Secondly we will then present our findings from attempting to,
in parallel with the manual approach, adopt an automated model driven approach
through the use of MDA (Model Driven Architecture ) compliant tools.
Additionally we will also describe some of the difficulties experienced in
adopting the model driven approach in a multi-partner European project.
Network Management Based on Distributed Paradigms
Madeira Vision
Telecommunications networks will evolve to contain very large numbers of nodes involving
networks of networ ks (with more operators, equipment vendors and owners) and with larger
technological diversity and heterogeneity of elements. The requirements placed on management
systems will lead to new paradigms being needed in order to deliver more adaptive and simpler
systems, particularly required will be a new relationship between Configuration Management
and Fault Management functions. Key requirements on the management systems are that they
are capable of self-configuration within a network, capable of inter-network collaboration for
management tasks, are much cheaper compared to today (i.e. lower Operating Expenditure
(OPEX)), more adaptive to service requirements, varying network deployment scenarios and
highly usable in all respects.
The objective of the Madeira project [1] is to investigate large-scale distribution techniques in
Network Management and ultimately to provide novel technologies, namely P2P concepts, for a
logically meshed Network Management System (NMS) that facilitates dynamic behaviour of
transient network elements. This will enable self-managed services and network elements of
increased scale, heterogeneity and transience thereby reducing OPEX. In particular, this research
aims to develop a logically meshed distributed non-hierarchical telecommunications
management system that focuses, within a case study, on the inter-related configuration and fault
management (CM/FM) aspects of a network and its element in large-scale dynamic
heterogeneous networks. It will also consider the relationship between a network management
system and an operators operations support system (OSS) in such an interlinked CM/FM
scenario. It will take a scenario based user-centric approach to requirements definition and
emphasise the services that the system elements shall offer from the perspective of the users (i.e.
Operators and Application Developers).
The focus of this project can be divided into 3 core areas. These are: computing and
communications platform; data modelling and architectural requirements. While these three
areas are intrinsically linked this paper focuses specifically on the modelling concerns and on
how the application of the model-driven approach in this context can benefit the development of
such a management system.
Network Management Based on Distributed Paradigms
Madeira Architecture
Platform
Services
AMC
Core Policy
Applications
Services
Network Element
Node A
Node C
Node B
Node n
Based on the approach of applying P2P concepts to the management domain the Madeira
architecture has been designed with a number of key principles in mind. The most important of
these principles is ‘heterogeneity’ or the ability of the Madeira system to be applied to many
management domains and across heterogeneous devices and platforms. The main vehicles for
this generic management is the usage of policies, notifications and applications.
The Madeira architecture is essentially composed of an Adaptive Management Component
(AMC) and a Platform. The AMC is the component that manages a given node and together
many AMCs can orchestrate the overall behaviour of the meshed network. These AMCs have
the ability to exchange and export Network Management information between peer management
applications and are deployed as an overlay network, communicating using the peer-to- peer
paradigm. The AMC itself is composed of a number of sub-elements which facilitate this
management.
The AMC Core is the is the primary component or ‘brain’ of the AMC and, based on
notifications and policies, orchestrates the services and applications to facilitate the required
network management function. Services are components that provide some functionality
required by the AMC Core. Applications provide the actual management functionality specific
to a particular device. Applications are physically divided into parts that run within each AMC
but are logically connected through peer interactions. For example each AMC will run some
Fault Management application which together with FM applications running on other nodes
constitutes the overall FM application of the entire meshed network. Policies provide the generic
management functionality shared by all AMCs within the sa me management domain. The
generic management specified in policies is then mapped to applications. Notifications enable
inter-AMC communication, via the Madeira platform, about events within the mesh network and
also provide a means to a logically distributed applications.
Network Management Based on Distributed Paradigms
Modelling Approach
Vertical Slice
– Cross-section of the modelling aspects of the
project for a specific scenario.
– Useful to evaluate our approach.
Adopt Model Driven Approach
Use MDA to model vertical slice.
– Many concerns among consortium.
There are a number of key decisions that were made in terms of the approach taken to modeling
in this project. The first decision was to adopt a ‘Vertical Slice’ approach. The vertical slice is
essentially a full cross-section of the modeling and architecture aspects of the project for a
specific scenario. The purpose of this is to evaluate our approach, highlighting problems or
issues, which can then be addressed and re-worked. It also serves to establish req uirements
towards the Platform.
Clearly the second decision was to investigate the application of MDA within this vertical slice.
This in itself has lead to many interesting issues arising as a result of a consortium of pan-
European partners with divergent concerns. These concerns and the overall conclusions drawn in
regard to these are outlined later in the presentation.
Network Management Based on Distributed Paradigms
Vertical Slice
The system development lifecycle process is a proble
m
-solving process where requirements
maybe considered problems occurring within a domain, systems that address the requirements
may be considered solutions residing within an environment, and pr oblem solving involves
understanding a problem, solving the problem, and implementing the solution [2].
The diagram above shows how both MDA and the Vertical Slice concepts fit together. The aim of
the vertical slice approach is to develop the platform independent models around a specific
scenario. Instead of trying to tackle all aspects of the problem domain, we instead select a set of
requirements and design our solution based on this chosen subset. This subset then provides the
initial input to our Platform Independent Model (PIM). The chosen scenario was orientated around
a basic configuration and fault management use case where nodes must first configure themselves,
then identify a fault and re-configure based on the identified fault.
Based on the architecture described earlier we saw that applications are considered as separate
logical functions from the underlying AMC, where the AMC provides the supporting
infrastructure for these management applications. As such the vertical slice scenario drives the
specific requirements of both the AMC and the Applications themselves. Also it should be clear
that, as the supporting infrastructure, the AMC is a pre-requisite of the Application and as such the
vertical slice must first address the AMC model. The relationship between the application models
and the AMC model is not a linear one, as to design the AMC we must consider its applications
yet the applications are inherently reliant on the AMC. This is the main aim and strength of the
vertical slice approach as it will rapidly produce a prototype and identify major issues and areas of
concern.
With this PIM we can then generate a Platform Specific Model (PSM) based on mappings to our
chosen platform. To represent the platform specific model, we started by identifying the necessary
components of our system, as defined by the original scenario, and the platform requirements for
implementing a P2P network management system. Once a stable representation of the basic
functionality has been developed we can then build extensions that consider all other aspects of
the entire problem domain. This is an iterative process where every model is documented,
reviewed and updated by the partners according to their area of expertise.
Network Management Based on Distributed Paradigms
Overall Modelling Initiative
The case of a vastly distributed, non hierarchical, logically meshed peer-to-
p
eer management
system for heterogeneous networks generates a set of models that try to capture the complexity
of the P2P Network Management System. The slide above highlights the overall modelling
initiative and the partition of the work between the partners. Every partner, in their area of
competency, developed specific models in coordination with the other partners.
Some important points are:
Mappings from PIM to PIM have been done manually
We are investigating tools and process to automate the mapping from PIM to PSM and PSM to
PIM
Having the underlying infrastructure specification is necessary to specify the mapping.
The Mapping itself is inspired by the best practises in software engineering. In our case it
covers different areas: P2P Networks (Protocols and APIs) and Policy-based management.
Network Management Based on Distributed Paradigms
MDA Concerns
Support in meta model to handle dynamic
behaviour
Complexity of mapping between the various
layers.
The ability to generate software via a tool set
from model.
How much of the mapping has to be hand coded
Impacts on OSS total solution, where some of the
system is not built in a MDA fashion
MDA Concerns: Using MDA to Build a Distributed Management System.
As described previously, the Madeira project is concerned with management of dynamic radio
networks. Our approach is one centred on the adoption of meta-modelling principles. Our
problem offers many difficult challenges and differs from the current adoption of model driven
development. The main issues we face are concerned with the management of a loosely
coupled dynamically forming communication platform [3].
If we analyze the conceptual approach of MDA, which is at the core of meta-model [4]
thinking, it clearly illustrates both the strengths and weaknesses of the MDA approach. From
our experience it allows you to think of the problem you want to solve in a different way. You
can abstract the problem and break it up into smaller parts, which can be solved individually. It
gives you a way to visualise the problems and to solve them without being swa mped in the
immensity of all the issues that need to be solved. This is extremely important and is why all
modelling is useful and used. However there is another saying that illustrates the problem with
MDA just as succinctly, “The devil is in the detail”. This is true of almost all software
systems, however in the field of distributed real-time network management it is even more so.
The consequence of this means that while it is very important that the high level modelling and
design is done correctly it is equally vital that any design/design methodology is capable of
dealing with minute detail.
Two noticeable problems with the MDA theory arise from this assertion. Firstly in order to
deal with the detail the model becomes more and more complex. This may be unavoidable but
it can lead to the loss of the advantage that the model gave you in the first place, i.e. a level of
abstraction. Secondly changes made to the PSM in order to account for the details that are
abstracted out of the PIM are not reflected back into the PIM, which over time makes the PIM
increasingly irrelevant and not better than standard textual descriptions (possibly even worse as
it becomes more expensive in tools/expertise to maintain).
Network Management Based on Distributed Paradigms
Meta-Layers
M3-Meta-meta Model
Model of Object Primitives
M2 – Meta Model
Management and Middleware
Architecture
M1 – PIM (AMC Model)
• Instance
As specified by the MDA we adopted the MOF 4-layered architecture as our overall modelling
framework. At the same time, and in order to address the MDA concerns of the consortium, we added
more detail in the meta-levels. By doing this we aimed to ease the mapping between layers and also to
ensur e that we addressed the problem of capturing beh aviour within the meta-layers.
At the top level is the Madeira meta-meta-model (M3) which was based on the Model of Object
Primitives [5]. The MOP provides a basic set of model concepts on which the other model layers could
be built. Key to MOP is that it incorporates policy and events so behaviour is considered from the
outset, at the highest level of abstraction.
At the M2 equivalent level is the Madeira meta-model. Here the modelling constructs defined in the
meta-meta model begin to become influenced by the problem domain, incorporating concepts that are
useful to describe our particular system. The Madeira meta-model is based on work carried out on the
Middleware and Management Architecture (MAMA) [6]. This premise of this work was to integrate
the ‘best of breed’ approaches to management and middleware in an attempt to define a unified
management and middleware architecture. As such MAMA adopted concepts from prominent
distributed processing and management standards/approaches like the RM-ODP, TINA-C, CORBA and
CIM. While this was adopted as the basis for the Madeira meta-model, it needed to be extended to
incorporate concepts for a peer to peer management paradigm, in line with the concerns expressed
previously. As a result Connections, Transactions and Distributed Application elements were added to
the meta-model to represent P2P network management concepts that were key to the Madeira scenario.
The AMC model utilises the newly defined concepts within the preceding meta-layers to describe a
single Adaptive Management Component. One of the key aspects of this is the representation of policy,
notifications (events) and applications in such a way as to describe the management mechanism of a
Madeira node.
Also at this level there are a number of Application Models. As the AMC is a generic tool for applying
any management function to the network we needed to define specific models of the management
application. These models were based around the vertical slice scenarios, specifically Configuration
and Fault Management.
Network Management Based on Distributed Paradigms
Manual Mappings
Entirely UML based
‘Simple’ UML modelling tools
No standard mappings.
No formal mapping between layers
Simple stereotypes
– No control mechanism for meta-model
conformance
• Hand-coded
As a result of the MDA concerns expressed within the Madeira project two separate approaches
to the modelling were adopted. The approach descried here is best described as a Manual
Mapping Approach. The manual approach describes the method of mapping between the meta-
layers using only a ‘mental-model’ and ‘simple’ modelling tools. This essentially involves
creating each model based on it’s preceding meta-model using simple UML stereotypes. As a
result there is no formal mapping between layers and hence no strict means of ensuring that the
each model conforms to its meta-model. Also with no formal mappings defined then it is very
difficult to maintain consistent models as each developers mapping can vary. Using this
approach the models defined essentially become the primary design specification and are used to
hand-code the Madeira system.
We began our Vertical Slice approach using the basic AMC model as this is a pre-requisite for
all the application models. Each model element in the AMC was stereotyped from meta-model
elements as described in the previous paragraph. An example excerpt is shown on the next slide.
Network Management Based on Distributed Paradigms
Example
Meta-Model
Model (AMC)
In this example we can see that elements from the Meta-model on the left are mapped to the
model on the right using basic UML stereotypes. So for instance a Module in the meta-model
contains one core object and many service objects, where both Service and Core contain
Attributes and Methods. A Service is invoked by a policy Action where the Core parses a Policy.
In the AMC model we can see that an AMC is a Module and as such contains an AMC Core and
AMC Services. AMC Core and AMC Service have attributes and methods.
The model here is too flexible and cannot be automatically be verified and constrained. For
instance the AMC as a Module strictly speaking should have only one Core and any number of
Service however using stereotypes there is no way to verify that this is actually the case. This
approach is error-prone and is only useful as a roadmap for developers to build the software,
often requiring much manual validation and correction. Simply put this approach is not formal
enough.
Network Management Based on Distributed Paradigms
MDA
Formal mappings
validation of conformance is inherent
Inheritance of constraints and behaviour
Standard representation of mapping
More consistent models
Possibility of defining grammars and
generating code and parsers.
The alternative approach is to use an MDA tool to develop the models where each meta-level becomes
an instantiation of the previous meta-model. In this way stereotypes become not only text-based typing
but physical instances of a related meta-element. This allows constraints (and behaviour) expressed in
the meta-model to be inherited by subsequent instantiations of this model. Formal standard mappings
ensure that the models built are consistent across the development lifecycle. This approach can provide
huge advantages in software development as it allows early testing and validation of models before
implementation as well as more consistency and conformance. One of the main benefits of MDA is
obviously the generation of code. As the model is formally represented in a way that can be machine
manipulated it is also possible that language grammars can be easily built from that model and that
resultantly the development of parsers can be simplified.
In this case we started by specifying the MOP model in our chosen MDA tool. This provides the basic
building block upon which all the modelling work would be built. When specifying the Meta-Model the
MOP-based meta-meta model is then used as it’s meta-package. This enables the meta-model to be a
direct instantiation of the MOP meta-model and hence inherit its properties and constraints. Verification
of this model is automatically carried out by the tool as any invalid instantiation of a meta-meta model
element will not be allowed. However while the manual mapping approach allowed us to reach a much
greater level of maturity in our models; we were not able to achieve a similar result with the MDA
approach. This was due to a number of reasons which are detailed in our conclusion.
Network Management Based on Distributed Paradigms
Conclusions
MDA holds a lot of potential
Model instantiation and verification
Constraint inheritance through layers
Grammars and Parsers
MDA tools too immature
– Unstable
Not able to model some simple constructs properly
MDA Standards are still being developed
Domain specific elements in meta-layers
More expressive modelling constructs
Instinctive mappings
MDA tools, like MDA itself, demonstrated extraordinary potential. However at this early stage
in the evolution of both there still remain a number of issues.
So far our experiences of many of the existing MDA tools is that while much progress has been
made they are still very immature in their nature. Also the ability to represent some basic
modelling constructs was lacking in some of these tools which presents a major obstacle to
system developers/modellers.
Also a number of the key MDA standards are still under development. MOF and UML are
standards of the OMG and all the Modelling tools implement these specifications. For model
transformation (validation, checking consistency, etc.) the important OMG standard is QVT
(Query View Transfor mation). The final specification of QVT is going to be finalized, and
publicly available, by July 2006. Similarly for MOF version 2.1. The compliant tools for these
core technologies should be available next year. So the available tools are not mature because of
the non-finalized standards and lack of cases studies.
Adding domain specific elements within the meta-layers resulted in a much more expressive and
instinctive meta-language for the development of the Madeira system. This also made the
mapping process between layers much less painstaking as the required mappings became much
more obvious. In particular introducing behaviour primitives at an early stage ensured that at all
stages of development system behaviour was considered.
Network Management Based on Distributed Paradigms
References and Future Work
Evaluate our design
– through implementation and formal evaluation
metrics
Evaluate Vertical Slice approach?
Advance the MDA approach
– MDA adoption limited
– Continue
The next stage of this work is to validate the overall Madeira vision by implementing the
specified system and devising formal metrics on which this system can be evaluated and
compared to traditional distributed management techniques. Another task which we are planning
is to carry out a qualitative review of the vertical slice approach. This review will attempt to
ascertain whether the vertical slice approach provided significant benefits to the development of
the system.
The work we have been able to in terms of MDA has been limited. Our ultimate aim would be to
use MDA to generate code based on the models provided and then to evaluate this code against
the hand-written code. This we feel would provide an extremely useful metric as to the current
state of MDA. Most of the code generation is linked to the sequence diagrams for Fault
Management and Configuration Management. We used stereotypes and constraints but we
mapped to real world objects manually since there is no final implementation of the OMG
standards. We still investigating the potential of many commercial tools (Rational XDE, Borland
Together, SparxSystem) and free plugins for Eclipse (IBM Model Transformation Framework,
IBM Design Pattern Toolkit, MerlinGenerator).
References
[1] The Madeira Project, http://www.celtic-madeira.org/
[2] Sinan Si Alkhir, “Understanding the model driven architecture”, published in “Methods &
Tools” October, 2003, an international software engineering digital newsletter published by
Martinig & Associates. http://home.comcast.net/~salhir/UnderstandingTheMDA.PDF
[3] Martin Zach, Daryl Parker, Liam Fallon, Christian Unfried, Miguel Ponce de Leon, Sven van
der Meer, Nektarios Georgalas, Johan Nielsen. “CELTIC Initiative Project Madeira: A P2P
Approach to Network Management”, Eurescom Summit 2005 Ubiquitous Services and
Applications Exploiting the Potential. 27 - 29 April 2005 Heidelberg, Germany
[4] Model Driven Architecture (MDA), MDA Guide Version 1.0. Copyright © 2003 OMG,
http://www.omg.org/mda/mda_files/MDA_Guide_Version1-0.pdf
[5] Nektarios Georgalas, “The Model of Object Primitives (MOP)”, Succeeding With Object
Databases: A Practical Look At Today's Implementations With Java™ And Xml, October 2000,
464 pages, ISBN 0-471-38384-8
[6] Sven van der Meer, “Middleware and Application Management Architecture”, PhD Thesis,
Berlin, Germany, September 25, 2002. http://edocs.tu-berlin.de/diss/2002/vandermeer_sven.htm
  • Article
    Full-text available
    In this paper, we detail how integrated Traceability Maps for software systems can provide a tool improving change impact analysis and product lifecycle management. This work is based on previous publications on architectural artifacts and Traceability Maps for Contracts. Using the NGOSS Lifecycle, we show a simple language for Implementation View Contracts and use off-the-shelve analysis software to show the effects of a Traceability Map for static analysis. This work then provides the basis for runtime analysis of software systems. The paper concludes with a brief description of our current work on federation aspects, in other words maps of exposed parts of domains of software systems. This work is integrated into joint research on autonomic networks and systems, thus our use cases and examples are focusing on distributed, autonomic network management systems.
  • Article
    Full-text available
    In this paper, we detail the use of the Design by Contract (DbC) paradigm, applied to the Technology Neutral Architecture (TNA). Furthermore, we demonstrate how specifications for contracts can be used to manage an Autonomic System, in combination with Policies and Processes. We introduce Traceability Maps as a tool enabling manageability of software artefacts of Autonomic Systems. The work presented is mostly based on four years work within the TM Forum's TNA team. The goal of this paper is to provide the reader with extended guidelines and architectural patterns that will help to develop manageable autonomic software and communication systems. Since we tackle the problem from a technologically neutral angle, this paper will not prescribe a single new technology, but rather provide a means that allows for federating different technologies, each of which offers particular advantages at business and system levels. In particular, it enables business concepts and principles to drive system design and architectures. This may be further implemented using currently available distributed systems information technologies.
  • Data
    Full-text available
    To devise a large-scale network management system, we should consider following issues: heterogeneity of network devices, interoperation between multiple network domains, scal- ability and availability of the management system, and the cost of developing and deploying the system. Traditional network management software does not provide a cost-effective solution for the above issues because they are usually designed for a specific network domain composed of a limited set of network devices, and they have a complex multi-tiered architecture that usually requires a separate failover scheme and development strategy for each tier at the additional costs of development, deployment, and operation. In this paper, we propose a novel network management architecture, CORD, based on consistent hashing. CORD achieves scalable load balancing and sub-50ms failover performance against a node failure with a one-tired architecture. Using XML, CORD provides highly-customizable southbound management interface which integrates network devices from multiple vendors with ease. With the experiments, we have shown that the standard deviation of load balancing on CORD nodes is maintained around 5% and the failover time is below 50ms on average. In addition, we have applied CORD to the optical network management system development to reduce CAPEX and OPEX.
  • Chapter
    Full-text available
    Current network management systems have been impeded by a scarcity of open standards for interoperable management solutions. Information models have made progress in promoting interoperability of traditional “centralised” networks but have still to significantly address the proliferation of next generation networks such as autonomic networks. Such networks impose challenges which include distributed self-control and self-management. The goal of the Madeira project was to utilise novel technologies and methodologies, based on an underlying P2P paradigm, for a logically meshed, distributed Network Management System (NMS) that facilitates dynamic behaviour of transient network elements. In this paper, we describe a solution for a meta-model that attempts to capture the key concepts behind the task of network management of a mesh network. A case-study focusing on the fault management of such a network will be presented with the purpose of verifying the applicability of such a meta-model.
  • Conference Paper
    Full-text available
    To devise a large-scale network management system, we should consider following issues: heterogeneity of network devices, interoperation between multiple network domains, scal- ability and availability of the management system, and the cost of developing and deploying the system. Traditional network management software does not provide a cost-effective solution for the above issues because they are usually designed for a specific network domain composed of a limited set of network devices, and they have a complex multi-tiered architecture that usually requires a separate failover scheme and development strategy for each tier at the additional costs of development, deployment, and operation. In this paper, we propose a novel network management architecture, CORD, based on consistent hashing. CORD achieves scalable load balancing and sub-50ms failover performance against a node failure with a one-tired architecture. Using XML, CORD provides highly-customizable southbound management interface which integrates network devices from multiple vendors with ease. With the experiments, we have shown that the standard deviation of load balancing on CORD nodes is maintained around 5% and the failover time is below 50ms on average. In addition, we have applied CORD to the optical network management system development to reduce CAPEX and OPEX.
  • Article
    This case-study describes how to introduce a pragmatic Model Driven Development (MDD) approach in a large, world-wide financial services software company with a conservative culture and high time-to-market constraints. By the application of agile principles a pragmatic simplified approach was produced by using only a single modeling level and using XML as modeling language. Although simplified, this approach could still significantly improve the application architecture of financial message structure definitions by reducing duplication and automating development tasks.
  • Chapter
    Full-text available
    The 2006 MACE workshop [1, 28] has presented the drivers and challenges of Autonomic Networking [2] and fostered an understanding of emerging principles for this new type of networks [3]. In this paper, we present concepts and principles that define a technological neutral, architectural perspective of Autonomic Networks. The work presented in largely based on work within the Architecture team of the TeleManagement Forum (Technological Neutral Architecture, [4]) and joined research work of industrial and academic research teams (for example Ericsson [5], HP & QinetiQ [6] and Motorola [7]). The goal of this paper is to provide the reader with manageability guidelines and architectural patterns leading to the development of manageable autonomic software and communication systems. We present a component-based, distributed system architecture and an associated set of critical system services that Autonomic Networks require. Since we tackle the problem from a technologically neutral angle, this paper will not prescribe a single new technology, but rather provide a means that allows for federating different technologies, each of which offers particular advantages at business and system levels. In particular, it enables business concepts and principles to drive system design and architectures. This may be further implemented using currently available distributed systems information technologies.
  • Chapter
    Recently, Peer-to-Peer (P2P) architecture is being used to explore better the computing power and bandwidth of networks than a client/server architecture. In order to support the creation of P2P applications, some frameworks were proposed such as JXTA. However, large systems using P2P architecture are complex to be developed, maintained and evolved. Model Driven Architecture (MDA) can support the management of the complexity in the software development process through transformations of Platform Independent Models (PIM) into Platform Specific Models (PSM). In this paper, we apply an MDA approach to allow the development of applications based on JXTA. The JXTA implementation in Java is used to demonstrate our approach. We propose a model transformation definition from an UML model to a Java+JXTA model. In order to validate our approach, we present two case studies.
  • Conference Paper
    Full-text available
    The Madeira project addresses a novel approach for the management of network elements of increasing number, heterogeneity and transience. As next-generation networks exhibit major challenges for today's centralized network management systems, we investigate the feasibility of a peer-to-peer (P2P) approach, facilitating self-management and dynamic behavior of elements within networks. In this short paper we give an overview of the system architecture developed in Madeira and describe the key concepts, like Madeira platform services, adaptive management components and policies that provide the base for building distributed network management applications
  • Conference Paper
    The upcoming peer-to-peer (P2P) and other decentralized, co-operative storage mechanisms allow for decentralized storage of data as well as decentralized search, depending on the specific system. In this work, we assume these kinds of systems to store management information into it. We discuss some of the opportunities, requirements, and challenges when storing network management information decentralized. Additionally, we discuss some issue, when we assume the management plane itself being decentralized, adding the aspect of distributed processing. Distribution of storage and processing resources make the use of such systems very helpful, but still require some changes from today's known systems, which are typically targeted to store content for end-users, for example in file-sharing or content delivery networks (CDN).
  • Understanding the model driven architecture an international software engineering digital newsletter published by Martinig & Associates
    • Sinan Si
    • Alkhir
    Sinan Si Alkhir, " Understanding the model driven architecture ", published in " Methods & Tools " October, 2003, an international software engineering digital newsletter published by Martinig & Associates. http://home.comcast.net/~salhir/UnderstandingTheMDA.PDF
  • The Model of Object Primitives (MOP), Succeeding With Object Databases: A Practical Look At Today's Implementations With Javax And Xml
    • Nektarios Georgalas
  • Understanding the model driven architecture
    • Sinan Si Alkhir
  • Article
    Full-text available
    Die Dissertation zum Thema: Middleware and Application Management Architecture beschreibt einen neuen Zugang für ein integriertes Managementsystem für verteilte Netzwerke, Dienst und Anwendungen. Das Hauptziel dieses Ansatzes ist ein Softwaresystem mit Diensten, welche Komponierbarkeit, Skalierbarkeit, Zuverlässigkeit und Robustheit genau so wie autonome Selbstanpassung ermöglichen. Es konzentriert sich auf Middleware für Management und Kontrolle sowie auf die Nutzung für verteilte Komponenten.
  • Article
    Full-text available
    The vision of the Celtic-Initiative project Madeira is to provide novel technologies for a logically meshed Network Management System that facilitates self-management and dynamic behaviour of nodes within networks. These approaches should enable adaptable services and the management of network elements of increasing number, heterogeneity and transience, thereby reducing OPEX. In this paper, we set the scope for investigations within the project and give an outline of our approach. We present a scenario that challenges today's state of the art in Network Management and upon which we are building our case study for a detailed investigation of feasibility. Finally, we describe a preliminary conceptual system architecture and application data model, and give an insight into the expected final project results.