A Summary Description of the A2RD Project
Juliao Braga, Joao Nuno Silva, Patricia Takako Endo, Nizam Omar
AA Summary Description of the A2RD Project
Juliao Braga , , Joao Nuno Silva , Patricia Takako Endo , , Nizam Omar IST - INESC ID, University of Lisboa, Portugal Universidade Presbiteriana Mackenzie (UPM) Universidade de Pernambuco (UPE), Brazil Dublin City University (DCU), Ireland {juliao.braga,joao.n.silva}@[email protected],[email protected]
Abstract.
This paper describes the Autonomous Architecture Over Re-stricted Domains project. It begins with the description of the context uponwhich the project is focused, and in the sequence describes the project andimplementation models. It finish by presenting the environment conceptualmodel, showing where stand the components, inputs and facilities required tointeract among the intelligent agents of the various implementations in theirrespective and restricted, routing domains (Autonomous Systems) which to-gether make the Internet work.
1. Introduction
Autonomous System (AS) is the name given to the networks that making up the In-ternet (Hawkinson and Bates 1996). This cluster of ASes interconnected also calledrouting domains or more commonly the Internet, can be represented as in Figure 1.
Figure 1. How ASs build the Internet: is some i , where < i < . . . ≃ ( ) The ASes establish interconnections through a protocol called the
Border GatewayProtocol (BGP) (Rekhter et al. 2006). BGP is a complex protocol that requires alot of knowledge from the administrators of an AS. In addition to the complexity ofBGP, one must add the complexity of Internet governance, which is partially visiblein Figure 2. a r X i v : . [ c s . A I] S e p igure 2. The Internet Infrastructure Ecosystem. On 02/03/2018 there were 59.959ASes present in the Internet Routing Table Sometimes the human being forgets to update information, especially those relatedto routing policy and that reside on important servers such as
Internet RoutingRegistry (IRR). The IRR is a distributed database of route and route-related infor-mation (Braga 2010). This fragile participation of the human being in constructionand maintenance of IRR objects was the motivation for creating a model of agentsthat would replace human interventions on such objects. So, was implemented the Autonomous Architecture Over Restricted Domains (A2RD) into the domain of anAS, applying as use case over the IRR (Braga et al. 2015). A2RD replaces the hu-man with your agents,
Intelligent Elements (IEs), establishing a new IRR model,named innovation IRR (iIRR), shown in Figure 3.
Figure 3. The innovation IRR model established by A2RD
A special A2RD IEs, named specialized IEs, automatically create objects as de-fined by the
Route Policy Specification Language (RPSL) (Alaettinoglu et al. 1999, http://thyme.rand.apnic.net/current/data-summary lunk et al. 2005). Those objects that can not be created automatically will receivesupport from AS administrators through a human-computer cooperation mecha-nism. Nothing is changed in relation to the present and future IRR structure,characterized by the expectations recommended by the stakeholders to the InternetEngineering Task Force (IETF) and Internet Research Task Force (IRTF) dissemi-nated through of yours formal documents (Meyer et al. 1999, Villamizar et al. 1999,Newton 2004, McPherson et al. 2015, Kisteleki and Haberman 2016). Neither doesit affect the security concerns surrounding the IRR and Internet governance(Kuerbis and Mueller 2017). Similarly, tools that use IRR databases can be usedwithout any modification. A very useful, among others, is the IRR Powertools (IRRPT).The purpose of this article is to summarize the A2RD project and it is complemen-tary to Braga et al. 2018a. Divided into four sections, the first is this introduction.The second section is a description of the A2RD abstract model. The third sectionshows the A2RD implementation model and the fourth section ends the article, withthe conceptual model of the A2RD development environment.
2. The A2RD Abstract Model
A2RD is a project that proposed that proposes a model of Intelligent Elements(agents) divided into layers as can be seen in Figure 4.
Figure 4. A2RD layer model for the ASx domain, where x is the AS number.
The model serves the interest of establishing an architecture of intelligent elementson the administrative domain of ASs. It may exist in any of the 2 possible ASs.However, on 25/08/2018 there were only ASs, originating traffic on the In-ternet, according to CIDR-Report . The number of an AS is unique, controlled by https://ietf.org/ https://irtf.org/ https://github.com/6connect/irrpt < http://thyme.rand.apnic.net/current/data-summaryl > he Public Technical Identifiers (PTI) and is named Autonomous System Number (ASN). Thus, the largest possible value of x is , corresponding to AS61612 ,at the date above. A2RD implementations are independent and restricted to anAS, but with a high degree of interoperability and, of course, intensive coopera-tion because AS administrators depend on the behavior of all others. The PTI hasreserved two contiguous ranges of AS numbers for private use: and (Mitchell 2013). Conveniently, these AS numbers canbe used to designate Intelligent Element domains.The first of the four layers hosts the Intelligent Element called the Controller. Itsidentification is unique and fixed: x:0 , that is, the number placed to the rightside of the : symbol, following the ASN hosting the model. Sometimes, to makeclear which IE is being referenced, IE is used before identification, for example, bystating that the IE Controller is IEx:0 . Thus, if
ASn is the host domain of themodel, then the controller element is
IEn:0 . No IE from the lower layers may existwithout the prior consent of the
IE Controller . It has the property of keepinghimself organized (self-organization) and of ensuring the self-organization of any IEfrom the lower layer.The second layer is represented by the so-called
Specialized IEs . These elementsare identified by suffixes that can range from to . The specialized elementssupport the IE Controller in specific activities required for functionalities rangingfrom ensuring the interoperability of the entire system of implemented IEs to spe-cific functionalities such as servers with end-to-end characteristics that stimulatethe understanding between two architectures: the layer model and the topologicalmodel (Saltzer et al. 1984), access features to bank semantic repositories, propri-etary software (similar to Southern SDN APIs), facilities required for lower-tier IEs,and many others. However, support for the IE Controller is the primary objectiveof the
Specialized IEs . This objective is what determines the features of the secondlayer. It is assumed that some
Specialized IEs may be
Autonomic Elements orintelligent elements that execute automatic processes, such as proprietary softwareand procedures associated with legacy systems, among others. A
Specialized IE can be created with functions that only concern the
IE Controller , especially whenit depends on the functionalities of IEs of the third layer.In the third layer lies the largest agglomeration of IEs, which is why it is called the
IE Colonies . Elements of this layer can be autonomous , autonomic or auto-matic , except legacy and are directly responsible for the most important activitiesof the application, including software reuse. They act under the influence of a highdegree of interoperability and cooperation between them and between IEs of otherlayers and other domains / subdomains. They do not directly participate in inter-connections or exchange messages with other IEs outside the domain, but they doso through IEs in the upper tiers. There is intense semantic interoperability activityon the part of these IEs, which have a high capacity for self-learning due to contin-uous interactions with the domain environment, and produce improvement effectson the knowledge of other IEs of the colony itself and the IEs of the layers the IE < https://pti.icann.org/ > Recognized as end-to-end arguments ontroller . In other words, these IEs favor the learning of the entire cluster ofIEs of the layer model, which is being described. The IEs of the colonies receive anidentification with numerical suffixes, ranging from to .The fourth layer is the Auxiliary IEs . This layer exists, in order to allow the trans-fer of computing demands to a new set of IEs (A2RD successiveness). It reproduces,successively, the first, second, third and fourth layers. This new IEs sequence has anadditional suffix :j:0 for a new
IE Controller responsible for the next four layers.In the second, third and fourth new layers, the IEs identifications are postfixed with :j:id where, j is the colony IE number that originated the new fourth layer and id is a number with the above specifications. A typical application for the fourth layerare subdomains, such as home networks ( homenet ).The use case for the A2RD was the addition and update of objects in IRR server. Theapplication was considered useful mainly because the tasks of the AS administratordid not guarantee the accuracy in its completion nor the permanent need to updatethe objects making the IRR an unreliable system from the point of view of itscontents. A2RD solved this problem (Braga et al. 2017).
3. The A2RD Implementation Model
According to Figure 5, the IEs are arranged and distributed between layers, similarto what was said in the previous section and are implemented in the domain of anyASN.
Figure 5. A2RD implementation model, where x is any ASN.
It is also observed, in the same figure, that the IEs functionally important in inter-domain operations reside in the upper layers. For example, a classification of rele-ance is the intensity of aggregation that an IE possesses, in relation to the auto-*(or self-*) properties. If an IE, however, has some self-organizing capability, it mustparticipate directly linked to the
IE Controller . Even if it participates in the
Aux-iliary IEs layer, an
IE Controller can logically construct a new layer architecture.And so on.On the other hand, the representation of the model is logical (abstraction of thephysical implementation). Physically, the locating if an IE in the domain environ-ment is essential. The best alternative is IP addressing, preferably IPv6, for reasonsof availability. The IE Controller must maintain a table associating the referencereference logic with the IP designated by the IE Controller itself, from the premisethat an IPv6 block should be available at the beginning of the implementation. Inthe implementation of prototypes related to the case study, the Python languagewill be used. When needed, features closer to the operating system will be used(”scripts” and other inherent facilities).
4. The A2RD Environment Conceptual Model
The Figure 6 shows the environment conceptual model , named
Structure for Knowl-edge Acquisition, Use and Collaboration Inter A2RD Agents (SKAU) in which eachimplementation of A2RD, into an AS, is represented as an agglomeration of IEs ina four layers model (11).
Figure 6. Structure for Knowledge Acquisition and Use (SKAU)
The other components of the SKAU are dynamically constructed from non-structured databases, in this experiment, from the Request for Comments (RFCs)database containing documents authored by network operators, engineers and com-puter scientists, documentary methods, behaviors, research, or innovations appli-cable to the Internet, all of them, working in groups of the IETF and IRTF, andmaintained by RFC-Editor . hese SKAU components can be described as following: • RFCs are captured / updated and stored locally (1); • A set of tools responsible for acting lexically and syntactically on RFCs (2),transforming them into intermediary databases (3); • Other tools (4), like
Semantic Distillation , that act on the intermediarydatabases producing inputs for the construction of
Domain Datasets (6) andso these into
Training Data Sets (8). Also, these tools will support for providepart of the knowledge base (9) (Isotani and Bittencourt 2015); • Learning algorithms (7) that support the construction and use of
TrainingDatasets to renew the knowledge base and meet the demand of agents ofA2RD models in the process of developing and applied intelligent actions. • A database, named
IIBlockchain (10) built by each implemented A2RDmodel and stored together in the Git Hub (so, in cloud), that serves as sup-port for the process of collaboration and effective interaction, inter / intraagents of the models (Braga et al. 2018b). The
IIBlockchain cloud interactswith the learning algorithm and knowledge base allowing agents to exercise offline and online computation .Each AS can implement an A2RD, which is controlled by the IE named IE Con-troller , and receives the identification x:0 , where x is the AS Number (ASN).
5. Thanks
From Juliao Braga: Supported by CAPES – Brazilian Federal Agency for Supportand Evaluation of Graduate Education within the Brazil’s Ministry of Educationand was also supported by national funds through Funda¸c˜ao para a Ciˆencia e aTecnologia (FCT) with reference UID/CEC/50021/2013.
References
Alaettinoglu et al. 1999 Alaettinoglu, C., Villamizar, C., Gerich, E., Kessens, D.,Meyer, D., Bates, T., Karrenberg, D., and Terpstra, M. (June 1999). Routingpolicy specification language (rpsl). Technical report, RFC Editor. RFC2622. < > . (Obsoletes RFC2280) (Updated-ByRFC4012, RFC7909) (Status: PROPOSED STANDARD) (Stream: IETF, Area:ops, WG: rps) (DOI: 10.17487/RFC2622) . Acessado em 03/02/2018.Blunk et al. 2005 Blunk, L., Damas, J., Parent, F., and Robachevsky, A. (March2005). Routing Policy Specification Language next generation (RPSLng). Technicalreport, RFC Editor. < > . (Updates Offline computation is the computation done by the agent before it has to act, and onlinecomputation is the computation done by the agent between observing the environment and actingin the environment (Poole and Mackworth 2010)
FC2725, RFC2622) (Updated-By RFC7909) (Status: PROPOSED STANDARD)(Stream: IETF, WG: NON WORKING GROUP) (DOI: 10.17487/RFC4012) .Acessado em 03/02/2018.Braga 2010 Braga, J. (2010). Pol´ıticas de roteamentos: como resolver aimpossibilidade de implementa¸c˜ao na tecnologia hop-by-hop e o futuro. GTER29. Dispon´ıvel em < ftp://ftp.registro.br/pub/gter/gter29/01-PoliticasRoteamento.pdf > . Acessado em 25/05/2010.Braga et al. 2015 Braga, J., Omar, N., and Granville, L. Z. (2015). Uma propostapara o uso de elementos inteligentes em dom´ınios restritos da infraestrutura dainternet. In Anais CSBC 2015 - WPIETFIRTF, Recife, Pernambuco, Brasil.Braga et al. 2017 Braga, J., Omar, N., and Thome, L. F. (2017). Acquisition anduse of knowledge over a restricted domain by intelligent agents. In Proceedingsof the SouthEast Conference, ACM SE ’17, pages 203–207, New York, NY, USA.ACM.Braga et al. 2018a Braga, J., Silva, J. N., Endo, P. T., and Omar, N.(2018a). Theoretical Foundations of the A2RD Project: Part I . Available at:https://assert.pub/papers/1808.08794, DOI: 10.13140/RG.2.2.22156.97923.Braga et al. 2018b Braga, J., Silva, J. N., Endo, P. T., Ribas, J., and Omar, N.(2018b). Blockchain to Improve Security, Knowledge and Collaboration Inter-AgentCommunication over Restrict Domains of the Internet Infrastructure. In Proceedingof CSBC 2018 - V Workshop pre IETF, page 13, Natal, RN Brazil. Available at:https://arxiv.org/abs/1805.05250.Hawkinson and Bates 1996 Hawkinson, J. and Bates, T. (March 1996).Report on MD5 Performance . Technical report, RFC Editor. RFC1930. < https://tools.ietf.org/rfc/rfc1930.txt > . (Updated-By RFC6996, RFC7300) (AlsoBCP0006) (Status: BEST CURRENT PRACTICE) (Stream: IETF, Area: rtg,WG: idr). Acessado em 06/09/2014.Isotani and Bittencourt 2015 Isotani, S. and Bittencourt, I. I. (2015). Dadosabertos conectados. Novatec Editora, S˜ao Paulo, SP, Brasil.Kisteleki and Haberman 2016 Kisteleki, R. and Haberman, B. (June 2016).Securing Routing Policy Specification Language (RPSL) Objects with ResourcePublic Key Infrastructure (RPKI) Signatures. Technical report, RFC Editor.RFC7909. < > . (Updates RFC2622,RFC4012) (Status: PROPOSED STANDARD) (Stream: IETF, Area: rtg, WG:sidr) (DOI: 10.17487/RFC7909). Acessado em 29/07/2017.Kuerbis and Mueller 2017 Kuerbis, B. and Mueller, M. (2017). Internet routingregistries, data governance, and security. Journal of Cyber Policy, 2(1):64–81.cPherson et al. 2015 McPherson, D., Amante, S., Osterweil, E., Blunk,L., and Mitchell, D. (December 2015). Considerations for Internet RoutingRegistries (IRRs) and Routing Policy Configuration . Technical report, RFCEditor. RFC7682. < > . (TXT = 47996)(Status: INFORMATIONAL) (Stream: IETF, Area: ops, WG: grow) (DOI:10.17487/RFC7682). Acessado em 29/07/2017.Meyer et al. 1999 Meyer, D., Schmitz, J., Orange, C., Prior, M., and Alaettinoglu,C. (August 1999). Using RPSL in Practice. Technical report, RFC Editor.RFC2650. < https://tools.ietf.org/rfc/rfc2650.txt > . (Status: INFORMATIONAL)(Stream: IETF, Area: ops, WG: rps) (DOI: 10.17487/RFC2650). Acessado em29/07/2017.Mitchell 2013 Mitchell, J. (July 2013). Autonomous System (AS) Reservation forPrivate Use. Technical report, RFC Editor. RFC6996. < https://tools.ietf.org/rfc/rfc6996.txt > . (Updates RFC1930) (Also BCP0006) (Status: BEST CURRENTPRACTICE) (Stream: IETF, Area: rtg, WG: idr). Acessado em 03/03/2015.Newton 2004 Newton, A. (February 2004). Cross Registry Internet ServiceProtocol (CRISP) Requirements. Technical report, RFC Editor. < > . (Status: INFORMATIONAL) (Stream:IETF, Area: app, WG: crisp) (DOI: 10.17487/RFC3707). Acessado em 03/02/2018.Poole and Mackworth 2010 Poole, D. L. and Mackworth, A. K. (2010). ArtificialIntelligence: foundations of computational agents. Cambridge University Press.Rekhter et al. 2006 Rekhter, Y., Li, T., and Hares, S. (January 2006). ABorder Gateway Protocol 4 (BGP-4). Technical report, RFC Editor. RFC4271. < > . (Obsoletes RFC1771) (Updated-ByRFC6286, RFC6608, RFC6793) (Status: DRAFT STANDARD) (Stream: IETF,Area: rtg, WG: idr) . Acessado em 07/09/2014.Saltzer et al. 1984 Saltzer, J. H., Reed, D. P., and Clark, D. D. (1984). End-to-endarguments in system design. ACM Transactions on Computer Systems (TOCS),2(4):277–288.Villamizar et al. 1999 Villamizar, C., Alaettinoglu, C., Meyer, D., and Murphy, S.(December 1999). Routing Policy System Security. Technical report, RFC Editor.RFC2725. < >>