The question of whether an ideal network exists with global scalability in its full life cycle has always been a first-principles problem in the research of network systems and architectures. Thus far, it has not been possible to scientifically practice the design criteria of an ideal network in a unimorphic network system, making it difficult to adapt to known services with clear application scenarios while supporting the ever-growing future services with unexpected characteristics. Here, we theoretically prove that no unimorphic network system can simultaneously meet the scalability requirement in a full cycle in three dimensions-the service-level agreement (S), multiplexity (M), and variousness (V)-which we name as the "impossible SMV triangle" dilemma. It is only by transforming the current network development paradigm that the contradiction between global scalability and a unified network infrastructure can be resolved from the perspectives of thinking, methodology, and practice norms. In this paper, we propose a theoretical framework called the polymorphic network environment (PNE), the first principle of which is to separate or decouple application network systems from the infrastructure environment and, under the given resource conditions, use core technologies such as the elementization of network baselines, the dynamic aggregation of resources, and collaborative software and hardware arrangements to generate the capability of the "network of networks." This makes it possible to construct an ideal network system that is designed for change and capable of symbiosis and coexistence with the generative network mor-pha in the spatiotemporal dimensions. An environment test for principle verification shows that the generated representative application network modalities can not only coexist without mutual influence but also independently match well-defined multimedia services or custom services under the constraints of technical and economic indicators.
With accelerated development since the late 1990s, the Internet has achieved unprecedented success in the history of human network communication technology and application, becoming an information communication infrastructure that is as indispensable in people’s daily social activities as water supply, power supply, and gas supply facilities. However, since this communication network came into being, research efforts in network systems have consistently followed a development route similar to "crossing the river by feeling the stones," tirelessly exploring the "one-size-fits-all" ideal network system in terms of engineering and technical level. Frustratingly, thus far, whether in theoretical models or practical schemes, no unimorphic network system has been found that can not only effectively carry known services with clear application scenarios and ensure economic and functional performance but also adapt to future services with their continuously enriched application scenarios throughout its full life cycle. There are issues of theoretical and practical feasibility in adapting extensible and diversified application scenarios to a unimorphic network system, as well as technical and economic issues in matching diversified application scenarios with multiple unimor-phic network systems. Obviously, under the existing network theory framework, there is no solution to the problem of whether to make services adapt to the network or the network match to services. To achieve a breakthrough, firstly, the first-principles problem of the "ideal network" system-that is, the existence problem-should be made clear; secondly, it is necessary to study what first principles of technology can transform the theoretical existence of an ideal network system into engineering feasibility.
By reviewing the development history of network systems, we can trace back to the root causes of the above problems. Wu [1] summarized and divided the previous network technology development paradigm into three stages, in what may have been the first time the theory of paradigm has been applied to network theory. These three stages include: ① the initial stage of network and service integration, in which dedicated network services form, such as telephone networks, telegraph networks, and television networks [2]; ② the second stage of integrated-service digital networks, which mainly targets the transmission of multimedia services, such as voice, data, video, and image transmission [3]; and ③ the present stage of the separation of networks from services, as represented by the transmission control protocol/Internet protocol (TCP/IP) Internet [4], which aims to achieve the end-to-end integrated carrying of multimedia data communication services under the to customer (toC) condition. Based on the technological evolution of these stages, it is not difficult to see that the development of information communication networks has always revolved around the problem of how to support various or diversified service needs with a unified network infrastructure and technical system under given technical and economic conditions.
In today’s digital, intelligent, and networked era, diversified application scenarios and the demands of vertical industries, such as to business (toB), to machine (toM), and to unknown emerging things (toX), have risen strongly. To meet the needs of emerging application scenarios, Internet service providers (ISPs) and the technology industry have ceaselessly attempted to supplement various protocols or pile up various edge techniques, and the Internet Engineering Task Force (IETF) and International Telecommunication Union (ITU) have adopted more than 9000 request-for-comment (RFC) recommendations or related technical standards [5]. However, the consumer Internet, which is characterized by the separation of networks from services, still presents the following challenges:
(1) Services are generated or provided by the network edges. Although network function virtualization (NFV) technology can support the realization of some network functions, it cannot ensure the relevant carrying performance [6,7].
(2) The evolutionary development route of patch stacking has brought system control complexity to the brink of collapse, resulting in ubiquitous generalized functional security threats and breaches.
(3) Pipelined information connection services, which only aim at the end-to-end transmission of information or data, are far from meeting the needs of a series of emerging services in cyberspace, including cloud computing, edge computing, trusted computing, endogenous safety and security, and artificial intelligence.
In view of the genetic defects of the current Internet, a number of new network systems and applications with distinctive features have developed rapidly, such as identity network [8], named data networking (NDN) [9], mobility first (MF) [10], time-sensitive network (TSN) [11], and smart identifier network [12], some of which have been applied in industrial dedicated networks and achieved good results. Nevertheless, limited by the original design intentions of unimorphic network systems and service models, it is difficult for the traditional Internet to support the progress requirements of the era in terms of the inclusiveness and coexistence of various emerging application scenarios and services. Under the constraints of the huge stock assets of existing network infrastructures and terminal equipment, any innovative network system can only be dwarfed at most by the virtualized overlay of the current system; as a result, such systems can neither fully reflect the functional performance advantages of new network systems and application services nor be deployed in a large-scale and timely manner. Hence, the contradiction between the utilization of stock assets and the application of emerging network systems is becoming increasingly incisive. This is particularly the case within the digital economy era, with deep human-machine-things integration, as the social cost required for upgrading the network infrastructure is increasingly heavy, and the priority consideration of economic benefits often instinctively excludes the large-scale deployment of any revolutionary network system.
There are two fundamental causes of the above problems: First, a unimorphic network system cannot ensure the carrying performance for various and scalable services; and, second, there is a contradiction between the application demand of diversified network systems and the construction of a unified infrastructure. At present, the network infrastructure and application network systems adopt a tight coupling mode, such as the TCP/IP-based Internet, which means that the implementation of any application network system innovation on the already-deployed information and communication infrastructure involves sizable costs because of necessary upgrading and transformation. However, under the constraints of the existing unimorphic network systems, it is impossible for vendors or network operators to adapt to ever-growing future application scenarios and services in a backward-compatible way using the current tinkering mode of technological evolution in full life cycle of the network system. Even if a future network criterion [13] that is oriented to design for changes is adopted, there will only be a one-by-one development pattern in which one network system is replaced by another, which cannot break through the "impossible service-level agreement (S), multi-plexity (M), and variousness (V) triangle" dilemma proposed in this paper. Therefore, we believe that the issue of how to decouple application network systems and services from the infrastructure environment is the first-principles problem that must be addressed in order to resolve the contradiction between diversity and unity. William Ross Ashby, the author of Introduction to Cybernetics, stated, "In nature, only variety can destroy variety," which is also known as the law of requisite variety [14]. Robert May, a theoretical biologist, pointed out, "Highly diversified ecosystems must have reduced complexity, that is, there must be fewer or weaker interactions between species" [15]. We believe that any method of increasing complexity cannot continuously support diversified development needs; that is, the demand for service diversity in the full life cycle can only be matched by a diversity of network systems, rather than by a unimorphic network system. Hence, an ideal network infrastructure should be able to support the creation or deployment of diversified application network modalities and services throughout its full life cycle of the network system, and a modality designer will be able to design a matching network system according to the expected application scenarios and service requirements, thereby largely avoiding the problem that application scenarios and service scalability are constrained by rigid or solidified network systems (i.e., architectures).
In this paper, we attempt to theoretically demonstrate the sufficiency and necessity of a polymorphic network environment (PNE) as the future network system development paradigm, establish a general deconstruction model of the "impossible SMV triangle" to qualitatively analyze the limitations of any unimorphic network system, and reveal the first principle of the "generative network of networks" system by abstracting a mathematical model of services and resources so as to provide theoretical and methodological support for creating a new paradigm of network system development.
2.The theoretical dilemma of the traditional network system development paradigm
Technological development in human society has transitioned from the first industrial revolution, in which the steam engine realized the facilitation of material transportation, to the second industrial revolution, when electricity realized the convenience of energy transmission, and then to the present information technology revolution, represented by the computer, in which the integration of network technology has realized a leap forward in the development of information network services. We believe that a successful technological revolution that can serve humankind for a long period of time must take into account the following three factors: ① economy, as humanity always expects to achieve a higher performance at a lower cost to facilitate the promotion of technology; ② effectiveness, as a technology must be able to effectively ensure the specific pursuit of quality, requirement, or efficiency in human production and life; and ③ inclusiveness, as the core concept of this technology must be capable of being extended to a variety of application fields in order to meet the requirements of different production and life scenarios.
2.1. Factor analysis of the "impossible triangle"
As network systems are one of the core technologies in the digital field, network system developers have essentially been pursuing the maximum benefits of the characteristics of successful technology, whether in evolution or revolution development. Therefore, given the various problems presented by the current network system, we attempt in this paper to extract the three characteristics contributing to an ideal network and establish a general deconstruction model of the "impossible triangle" of network systems, in order to qualitatively analyze the existing theoretical dilemma presented by past and current network system development paradigms.
Generally speaking, a network system provides supporting capacity for service functions and performance through combination of specific software and hardware resources such as transmission, computation, storage, and security. Therefore, different network systems can be abstracted as different ways of mapping physical/logical resources to service functions and performance, as shown in Fig. 1. Here, it should be noted that economy, effectiveness, and inclusiveness are necessary characteristics that determine whether a technology can become a revolutionary technology. An ideal network system with global scalability in its full life cycle requires three factors: ① S, which represents the quality assurance of network service performance; ② M, which reflects the utilization of physical/logical resources under a given network scale; ③ V, which represents the types of network services that can be carried, including clearly defined expected services and extended services with clear application scenarios in the future. These three factors are referred to herein as the "SMV factors."
However, we have found that there can be irreconcilable contradictions among the SMV factors in the full cycle of any unimorphic network system that has been applied practically or commercially in the past (see Section 2.2 for details); that is, any unimorphic network system cannot simultaneously achieve the scalability of S, M, and V in its full cycle (which is also known as the global scalability of a network). To facilitate a qualitative analysis of the relationship among the SMV factors of any unimor-phic network system, based on the assumption that these network systems have global scalability in their full cycles, we have created a general deconstruction model of the "impossible triangle" for unimorphic network systems, referred to herein as the "SMV dilemma," as shown in Fig. 2.
2.2. Deconstruction analysis of the "impossible triangle"
It is pointed out above that, under the premise of global scala-bility, , and are three necessary factors for measuring network capabilities. This section further analyzes the inherent contradictions among these three factors in a unimorphic network system.
Lemma 1. If a unimorphic network system can achieve the optimal capacities of both and at the same time, then cannot be optimal.
Proof. When M reaches optimal capacity, the actual service-carrying capacity of the network environment reaches the maximum based on the given physical resources and, under the premise of scalability, the physical resources reach a saturated utilization state under the requirements of dynamic service scenarios and the existing network system. Assuming that there are services (where is the number of services) in the network at this time, then the saturated resource utilization state means that there is no new service such that can be carried in the network environment at the same time. Moreover, if is optimal, it means that the service levels of are satisfied; that is, the physical resources required by can be fully guaranteed in a specific period and location. In other words, the demands of for physical resources can avoid absolute conflicts in time and space by multiplexing. At this time, it is required that the demands of for physical resources should have a specific distribution interval in time and space (otherwise, resource conflicts will occur). Because of the scalability requirement of , the attribute combinations of are various and random, so there is a natural contradiction between the distribution regularity of over and the scalability requirement of at this time. To sum up, under the condition of satisfying the optimal capacities of both and , it is impossible to guarantee the optimal capacity of at the same time.
Lemma 2. If a unimorphic network system can achieve the optimal capacities of both and at the same time, then cannot be optimal.
Proof. Assume that there are services in the network at this time. If reaches optimal capacity, according to the scalability requirement, the network should adapt to various service attribute combinations at this time; that is, the services can adopt any service attribute combination. Meanwhile, in order to ensure an optimal , under the premise of scalability, all types of service attribute combinations of should be satisfied in terms of the service level agreement; that is, the required physical resources of can be fully guaranteed at a specific time and space location. As the service attributes of can present various combinations, the service demand for physical resources has a combinatorial randomness in its time and space distribution. In this case, it is necessary to ensure that the distribution of physical resources in the time and space dimensions can meet the upper limit of the demand types of any service attribute combinations when is optimal. Assuming that and represent the time and space dimensions, respectively, the upper limit of the resource supply at this time is recorded as . In this case, for a specific service attribute combination of , there must be a set of time slots and space locations such that the usage of resources is less than . At this point, there must be a service , where the demand of for resources in the time and space locations cannot exceed . Therefore, is not the maximum service volume carried by the network resources, so M does not reach optimal capacity.
Lemma 3. If a unimorphic network system can achieve the optimal capacities of both and at the same time, then cannot be optimal.
Proof. When M reaches optimal capacity, the actual service-carrying capacity will reach the maximum and, under the requirement of scalability, the physical resources will reach a saturated utilization state under the requirements of dynamic service scenarios and the existing network system. In this case, service carrying means that the abovementioned services can be transmitted in the network without being discarded. At the same time, the of the network will reach optimal capacity, which means that the network will support different types of service attribute combinations at this time; that is, it must adapt to all service attribute combinations, including the burstiness of service. In this case, because the service burstiness caused by any combination of service attributes is unknown, if there are any two concurrent services and that need to occupy the same physical resources in the same space location and time slot, then one of them will be unable to be effectively carried at the expected location and time slot. Meanwhile, since all the above services are required to be effectively carried when M reaches an optimal state, services and should not be discarded; then, one of these services (e.g., ) must actually be realized at a changed space location and time slot via a path change, caching, or other means. The service-level agreement of the service that cannot be realized at the expected space location and time slot cannot be guaranteed, so cannot be optimal.
To sum up, when the optimal capacities of the two-pole combinations of SM, SV, or MV are achieved, the third-pole optimization cannot be satisfied. Based on the above analysis, we put forward the following Theorem 1.
Theorem 1. The impossible triangle theorem of unimorphic network systems.
Under a realistic physical environment and the premise of satisfying the global scalability of full network cycle, it is impossible for any unimorphic network system to simultaneously achieve optimal S, M, and V capabilities.
3. Breakthrough in the theoretical dilemma of the existing development paradigm
3.1. Breakthrough in the "impossible triangle" dilemma via dimensional upgrading of the solution space
Throughout the history of the Internet, it can be seen that the available practical network systems have always been strongly related to the application characteristics of known or conserved service fields. Moreover, in different scenarios, there are both specific and general requirements for network systems and diversified service demands. It has been proven by the SMV dilemma that, under the premise of global scalability in full network cycle, no unimorphic network system can become an ideal network that is designed for change. Therefore, in order to fundamentally resolve the contradiction between the diversity of services (including service functions, performance/efficiency, and technical economy) and the unity of the network infrastructure, it is first necessary to break through the theoretical dilemma.
The SMV dilemma of unimorphic network systems highlights that unimorphic technology systems can only be statically mapped to the two-pole intersection spaces of SV, SM, and MV under the premise of satisfying global scalability in its full life cycle (theoretically, network resilience and service trust should also be included), and cannot form a complete intersection in the SMV space. All the available network systems to date have been based on well-defined or clearly outlined application fields, service types, carrying modes, technical and economic indicators, and so forth in the design of the related network baselines and the physical layer, logical layer, and interface layer protocols, including the data formats involved in the protocols and the methods of matching between the network protocols and the hardware devices. Scalability considerations have been limited, only being included in determined application fields, and designs beyond preset scenarios or addressing future applications have been a "best effort" affair (at least in previous implementation schemes).
From a macro perspective, despite the existence of the "impossible SMV triangle" dilemma in unimorphic network systems, an ideal network must support or carry diversified application scenarios and services throughout its full cycle. The first principle of an ideal network’s design must be to pursue the smooth updating of global scalability; thus, in specific application fields and service scenarios, complete SMV intersection is not always a sufficient and necessary condition. Therefore, under the condition that scal-ability is restricted to be a local variable or a compromise among the SMV performances and efficiencies, differentiated clustering of service requirements is a common practice in engineering implementation (e.g., asynchronous transfer mode (ATM)), although different clustering combinations may have different degrees of consideration for the three attributes of SMV. Hence, a specific service attribute will assign a specific boundary to any one of the three SMV factors. Within this boundary, the adaption contradiction between unified software/hardware resources and diversified services can be alleviated to some extent.
Based on the above analysis, we infer that adopting different network systems to carry different network service clusters (i.e., meeting the cybernetic law of requisite variety) is the only effective way to break through the SMV dilemma (although this is yet to be proved in theory). However, aside from special or customized network scenarios such as dedicated networks, generally speaking, the types and distribution of services of a network system in its full cycle are always dynamically changing. The service adaptability that can be met in the current stage may not be able to avoid a function/performance mismatch caused by the rigid network system in the next stage. Therefore, when scalability is a global variable in the full cycle of a given network system, the issue of whether there is an ideal network with , and completely intersecting in the SMV space is the first-principles problem that must be solved in the network system design and engineering.
The mathematician Bernhard Riemann introduced the temporal dimension into geometry, thereby laying a mathematical foundation for later great scientific theories such as relativity [16]. Inspired by Riemann, we intend to introduce the temporal dimension on the space dimension of the three factors , and to dynamically realize the SMV instantiation process of network systems throughout their full cycle. A combined morphology of multiple unimorphic network systems is adopted at different times to carry diversified services based on application scenario clustering, so as to break through the SMV dilemma in three-dimensional (3D) space by increasing the number of spatiotemporal dimensions to obtain a solution.
Theoretically, a unimorphic network system designed in a certain operation stage or historical period can only support clearly defined application scenarios according to a compromise scheme among the factors of , and , and cannot effectively carry scalable services that meet the requirements of future applications in the full network cycle. However, if multiple network systems with incomplete SMV intersection can be mapped in the SMV 3D space at the same time, complete SMV intersection can be obtained equivalently in the SMV & time (SMVT) spatiotemporal dimensions. As a result, the ideal network system cannot exist in the SMV 3D space; it is only in the SMVT spatiotemporal dimensions that an ideal network system that meets the scalability requirements of diversified application scenarios and services in the full cycle of the network system can exist. Based on this idea, the criterion for a future network should include an extended definition of the spatiotemporal dimensions; that is, the future network should not only support the coexistence of multiple unimorphic network systems in the spatial dimension but also support the evolution and revolution of any network system with incomplete SMV intersection in the temporal dimension (within the full network cycle).
Theorem 2. On the basis of SMV 3D capacities, complete SMV intersection can be obtained by adding the temporal dimension .
Proof. First of all, under the condition of relaxing the and constraints, the network can achieve the optimal capacity of service quality ; that is, the service-level agreement of actual services in the network can be fully satisfied at any time period . Meanwhile, for the optimal capability of , the network should adapt to combinations of various service attributes, which essentially means that the network can adapt to the combinations of service attributes that appear at any time in the full cycle . However, in each specific time slot , the overall service attribute combination space will collapse into a specific service set ; in this case, will have fixed service attributes, respectively, and their set will also have certain service attributes. In this context, due to the attribute certainty of each service , the physical resources required in the space and time slot are also determined. At this time, the space-time resources that the network environment needs to reserve for the services in space and time are recorded as , where represents the physical node (space). Based on these definitions, an optimal mathematical model can be established, as follows:
where represents the actual service quality obtained by the service , and represents the threshold value of the service quality guarantee for . The solution of the above optimization model corresponds to the minimum physical resource overhead-that is, the maximum physical resource utilization rate (i.e., M) that meets the service attribute combinations(V)of the current actual services and the service level agreement (S) of all services within the current time slot . At this time, the network can achieve the optimal SMV three-pole capacities at any time slot ; that is, the complete intersection of SMV is realized.
Corollary 1. The physical morphology of the complete intersection of SMV in the SMVT spatiotemporal dimensions should be able to dynamically adapt to different network systems.
The physical meaning of the solution in the abovementioned optimization model represents the method of adaptation between the physical resources and the upper-layer services, where the macroscopic clustering of different adaptation methods corresponds to different network systems. With a change in , due to the scalability requirement of optimal capacity, there will be various attribute combinations of in different time slots; correspondingly, the optimal resource adaptation, Eq. (1), will also have different solution distributions that correspond to different methods for adaptation between the physical resources and the upper-layer services-that is, different network systems. Therefore, in order to realize the complete intersection of SMV in the SMVT spatiotemporal dimensions, the network architecture that actually carries services must have the ability of dynamic adaptation in its full cycle.
In fact, different services have different endogenous adaptation requirements for network systems, due to different characteristics. If the unified physical resources are mapped to differentiated performance indicators of the upper-layer services in a static and fixed way, the service performance will inevitably degrade, or the physical resources will be wasted. Therefore, we plan to deconstruct the functions of the existing network system into the network’s "elements," flexibly divide or aggregate them in the resource dimension, and then create diversified "compound" network systems through combinations of elements using metaphorical "chemical bonds means links between network elements." Finally, we aim to achieve the best adaption of services based on an application network modality similar to a compound morphology, so as to maximize the alignment between service needs and network resource efficiency, as shown in Fig. 3.
The core meaning of obtaining a solution through dimensional expansion includes two aspects: the "elementization" (i.e., breaking down into the smallest level, or elements) of network resources and the "dynamicalization" (i.e., flexible and dynamic assignment) of resource aggregation.
The elementalization of network resources involves reabstracting the layer functions of the information communication network and decomposing them into finer-grained or appropriately grained basic network elements: network baselines. Network baselines are the most basic functional units that encapsulate the set of operations on underlying network resources. The elementi-zation of network resources is the basic condition for the realization of diversity; it ensures that-on the basis of unified network physical resources-diversified network systems can be implemented via resource combination and reconstruction. Similar to the case of the unity of the physical world contradicting the diversity of the material world in the natural environment, a relationship of the unity of opposites is achieved through arrangements and combinations of 97 naturally occurring chemical elements and chemical bonds, all organic and inorganic substances, and living and non-living bodies, following the unified law of conservation of matter and the law of evolution or mutation. This shows that the sustainable development of the macroscopic world, mesoscopic world, and microscopic world cannot occur without the characteristics of diversity and variousness-a concept referred to by some as the "first principle of universe’s development." Elementalized network resources enable the building of diversified application network modalities according to the service function/performance requirements based on given software and hardware; they also support the coexistence of various application network modalities within the same technical physical environment, so as to always meet the requirements of an ideal network in the full network cycle.
The dynamicalization of resource aggregation involves dynamically combining these network baselines and scheduling network resources to achieve any available network functions and network systems. A dynamic resource arrangement or loading/unloading capability ensures that diversified application network modalities can actively adapt to changes in the resource requirements of service. Under limited service types, dynamicalization can flexibly assign software and hardware resources to different network modalities with dynamic changes in different service volumes. Under the trend of network evolution or revolution, dynamicaliza-tion can ensure that the unified network environment can flexibly support the generation of application network modalities and recycle/unload in a timely fashion the software and hardware resources of the network modalities that are no longer active.
In this paper, the technical and physical environment that generates or creates diversified/varied application network modalities via the elementization of network resources and the dynamicaliza-tion of resource aggregation is denoted as the PNE. It is not difficult to see that a PNE disregards the message format, routing addressing technology, protocol specification and operation, and maintenance management of any specific application network modality. Its practice specification mainly focuses on the engineering problem of how to generate or create diversified/varied network modalities by using an integrated technical and physical environment-that is, it requires the development of a technology for generating a so-called "network of networks," for which a network infrastructure similar to a cloud platform is used to provide pooled computing, storage, transmission, exchange, routing, access, and other commercially circulatable software, hardware, and connection resources for upper-level application network modalities (e.g., IPv4/IPv6, identification network, NDN, and geospatial subdivision network). Based on this, the PNE can realize the separation of software and hardware infrastructure from specific application network modalities (including services) and can support the creation, deployment, and coexistence of multiple application network modalities within the same infrastructure environment. Therefore, the PNE can also be called the "base of networks" or the "network of networks."
3.2. Systematic morpha of the PNE
In his Introduction to Cybernetics, the famous cybernetic scholar William Ross Ashby pointed out that "only diversity can destroy diversity;" that is, the control system can manage a diversity of control objectives only by increasing its own diversity as the system complexity increases. The contradiction between diversified network needs and a unified infrastructure has resulted in a technical route that blindly enhances the complexity of unimorphic network systems to achieve scalability in the full network cycle. This route has reached the so-called "ceiling effect," making it increasingly difficult to adapt to complex and diversified vertical industry applications. We aim to follow the concept of maintaining the endogenous growth demand of the network environment by injecting diversity, which is done by transforming the network development paradigm from the conventional unimorphic network system into a multidimensional and diversified network environment. Hence, the technical engineering goal of the PNE development paradigm is to build an open, transparent, unified, resilient, intelligent, and safely isolated network infrastructure and a technological and industrial ecological environment for sustainable and healthy development that can support the rapid creation, plug and play, independent operation, and management of various application network modalities and related services. The system paradigm is shown in Fig. 4.
In contrast to the research coordinates of traditional network development paradigms, the research coordinates of the PNE development paradigm are focused on how to construct a unified network infrastructure environment that is characterized by on-demand sharing, flexible combination, safe isolation, and collaboration with the end users for various application network modalities and bearer services, as far as possible. The PNE should be able to support the rapid deployment or withdrawal and recovery of various application network modalities in a form similar to the plug-in application. This effectively deals with the problem of lacking a realistic test environment for research on emerging network theories, technologies, and application services, thereby enabling network operation enterprises to provide a large-scale and real testbed environment for innovative network technologies without affecting their commercial services. As a result, the threshold for new application network modalities and services to enter the market will be reduced, replacing the old one-by-one updating model with a new model based on the smooth upgrading and updating of networks, and thus creating a so-called "four-in-one" harmonious development pattern that involves the utilization of stock assets, network technology innovation, industrial development, and market application. In this way, a new ecological environment can be built for the integrated incremental development of network modalities, service applications, and network infrastructures.
Currently, the Internet adopts a "patch-stack" type of evolution technology to meet the needs of different bearer services. This method of patch stacking and complexity increase will inevitably lead to protocol bloat, a loss of focus, and uncontrolled complexity, and will even prevent the Internet from evolving to meet the new requirements of network resilience and flexibility. Moreover, such engineering and technical efforts clearly violate the first principle of cybernetics [14] and are doomed to be unsustainable. Considering the "anything over IP" mode, although the multimedia service field seems to have achieved a certain degree of technical success under the toC condition, it is still unable to avoid the shackles of a false proposition at the scientific theory level. The elementization of network resources, software-defined interconnection (SDI), and other similar arrangement technologies enables the separation of specific network systems from the underlying physical support environment, so that a variety of network systems oriented to different application scenarios can be implemented in the same technical and physical environment. This fundamentally resolves the contradiction between the unified infrastructure and scalable service function performance (if the corollary that "structure determines diversity" can be proved theoretically, the PNE belongs to the engineering and technical expression of such "structures"). The elementization of network resources involves re-abstracting the layer functions of the information communication network and decomposing them into finer-grained or appropriately grained basic network elements-that is, network baselines. As the basic unit of network functions, network baselines-like the elements in functional molecular structures-can be combined into more advanced forms of varied network modalities, so that the rigid hierarchical structure of traditional networks can be replaced by elemental and modular structures, thereby achieving the goals of the flexible creation or remodeling of network systems. The PNE separates the network baselines from the application network modalities, extracts the common factors of network systems, introduces functional elements such as computing processing, data storage, and intelligent services to form a network baseline element pool, and uses the available or future diversified application network systems to construct various desired application network modalities through combinations of the scalable network baseline elements and SDI.
The fact that network baseline elements can be dynamically constructed, added, and removed reflects the PNE’s microscopical adaptability. The network baseline elements are not unchangeable. According to the stage of technical development and the understanding of engineering practice, the network baseline elements can be divided and combined with different granularities. In specific engineering implementation, the network baseline elements can be roughly divided into two categories: basic elements and extended elements. Basic elements can be regarded as the software and hardware components that constitute the basic functions of networks; they generally include network baseline functions such as protocol systems (including physical/link/network levels), switching modes (circuit/packet), communication mechanisms (point-to-point, point-to-multipoint, and broadcast), and signaling systems (common channel signaling and channel-associated signaling). Extended elements can be defined by various software; they allow users to perform customization flexibly according to different network usage scenarios, application network modality functions, and performance requirements. The extended elements include not only combinations and variations of basic network elements (e.g., protocol system, switching mode, communication mechanism, and signaling mechanism) but also all aspects of a network, such as the computing/storage/forwarding mechanisms, scheduling strategy, service quality and security guarantee mechanisms, and operation and maintenance management. It is straightforward to infer that basic elements are more universal and standardized, while extended elements are personalized and targeted. Basic elements and extended elements are not strictly divided. With the widespread application of or the formation of consensus on some extended elements, they can be standardized and identified as basic elements by standardization organizations. By referencing the physical world, the elementization of network baselines is an expression of the unity of the network world, while the dynamic aggregation of network baselines and resources is a diversity-realization mode of the network world; full-dimensionally definable network baseline technology then serves as a "chemical bond" and is an technical engineering means for realizing the dynamic aggregations of network baselines and resources.
Full-dimensionality definable network baseline technology refers to an engineering implementation technology based on domain-specific architecture (DSA) and the software-defined hardware (SDH) method. It can define network baselines such as the protocol architecture, switching mode, and control mechanism, and aggregates network resources such as computing, storage, switching, and interconnection. By using full-dimensionality definable network baseline technology, a set of standardized network baseline elements, and software and hardware interfaces, the PNE can realize a software-definable unified network infrastructure for supporting a variety of application network modalities and services with quality commitment in the full network cycle. It can be predicted that, with the large-scale deployment of polymorphic network element devices to facilitate the rapid construction of diversified application network modalities, the implementation modes of the network baseline elements and full-dimensionality definable network baseline technology will be regarded as the standard of the PNE and of polymorphic network element devices and will become the good-practice guideline for generating and creating diversified application network modalities for mass coverage networks (by partition and domain). Therefore, the problems of how to decompose the functions of different technical network systems at the "compound" level into basic network elements at the "element" level and how to design a suitable software and hardware collaborative arrangement mode to carry out higher-level combinations at levels analogically similar to chemical bonds or molecules are the primary engineering and technical problems that must be solved in the PNE.
Full-dimensionality definable network baseline technology applies the idea of software definition to all levels of the network software and hardware resources, fully explores the network baseline elements, constantly enriches the pool of network baseline elements, and provides the developers or operators of application network modalities with SDI and DSA capabilities or standard interfaces. This breaks through the fixated network devices and rigidified hierarchical architecture of the traditional Internet, creating a new era of customizing desired application network modalities and services based on flexible combinations of network elements and network baseline technology.
4. The sufficiency and necessity of a PNE
4.1. Network modeling and mathematical description
The PNE extracts the basic elements of various application network modalities to form an element pool of network baseline techniques and constructs a dynamic and plastic network environment infrastructure based on SDI and DSA, so as to ensure that diversified network modality instances can be run on demand. Let the number of network baseline elements be ; then, the network baseline library can be expressed as . Let represent the network modality and let represent the combination of different baseline elements; then, the specific technical system of the network modality can be expressed as
The future network system will use various resource types in general information technology to provide ubiquitous information services. Without loss of generality, the physical resources involved in the network system are expressed respectively as follows: computing resources , storage resources , and bandwidth resources . Set the corresponding total amounts of each resource as , and .
In this section, we abbreviated Ser as . Set the types of services as and the information volume of each service as . Assume that there are a total of types of network modalities in the network environment, and set the physical resources consumed per bit of service carried by the network modality as , and , respectively. As a network modality can be deemed to be a mapping of the physical resources to the service metrics, the physical resources consumed by the service are determined by a combination of its baseline elements and can be recorded as , and , respectively. Therefore, we get
4.2. Reasoning based on the solution space relationship
Lemma 4. Under the stable state of the PNE, for any service type , there are no two network modalities and such that .
Proof. For any service , if there are two network modalities and such that
then, when (1) indicates that there will be a complete equivalence relationship between and . Considering the modality development and maintenance costs, only one of the modalities and will actually exist;
(2) indicates that the modalities and do not have a completely equivalent relationship, as the resource utilization rate of modality is better than that of modality , modality will be eliminated, and there will only be modality in the network environment.
We use the variable to represent the relationship between modality and service type , where indicates using to bear service type . Set the priority weight of the service type as indicates that the service type can be effectively carried by modality under the current resource constraints (i.e., its service indicators are met). According to the definition above, when a single modality is used to bear all services of the network, the optimal network service quality model that satisfies the physical resource constraints can be represented as follows:
When multiple network modalities are used to bear all network services , the optimal network service quality model that satisfies physical resource constraints can be represented as follows:
It can be seen from Eqs. (4) and (7) that, under the same physical resource constraints, the solution space of a unimorphic network is:
The solution space for using modalities to bear all services is:
Obviously, the solution space of a unimorphic network is a subset of the solution space of a polymorphic network. Furthermore, the above optimization equations can be divided into the following three scenarios:
(1) When both a unimorphic network and a polymorphic network can meet the quality of service (QoS) requirements of all services, , the resource consumption of the polymorphic network will be less than or equal to that of the unimorphic network. The reasoning is as follows: When the polymorphic network selects the same modality as the unimorphic network (set this modality as ) to bear all network service, the resource consumptions of the two are the same. On this basis, when , there can be and , and the PNE will select the modality to bear the service and the modality to bear other services. In this case, the resource consumption of the polymorphic network will be less than that of the unimorphic network.
(2) When network resources are limited, as the service volume grows, the unimorphic network will reach the resource boundary for no less than one type of resource, and the polymorphic network will potentially improve the overall QoS. Assume that the unimor-phic network reaches the resource constraint for the resource ; if the service volume of the services that consume resource increases, then the unimorphic network will be unable to meet the QoS requirements for all services-that is, there is at least one that enables in the objective function of the unimorphic network. Different network modalities have different resource consumptions for the same service. Take the example of the resource consumptions of the IP modality and the content delivery network (CDN) modality for a file transmission service. As adopts a distributed cache strategy to reduce repeated transmission, we have and . If the single modality is adopted to bear all network service, then the consumption of the transmission resource will reach the critical value. In this case, the PNE may adopt to bear part of the file transmission service so that the consumption of the store resource can replace that of the transmission resource . In this case, the overall QoS of the polymorphic network will be higher than that of the unimor-phic network; that is, .
To sum up, we can draw the following conclusions.
·In the service scenario constrained by limited physical resources, if the unimorphic network can effectively provide services (i.e., there is a feasible solution ), then there must be a feasible solution for the polymorphic network; moreover, the resource utilization of the polymorphic network is greater than or equal to that of the unimorphic network. Therefore, the polymorphic network is the sufficient condition to effectively provide services in the scenario of sufficient resources.
·In the scenario of limited physical resources, for a specific service scenario, if there theoretically is a feasible solution, then the polymorphic network can inevitably obtain an effective solution by adjusting the number and types of modalities, while the number of service scenarios that the unimorphic network can bear will be greatly reduced. That is, the polymorphic network will provide better QoS than the unimorphic network in the scenario of limited resources, which is the necessary condition to effectively bear all service schemes.
5. Experiments
The PNE may accept various new technical physical implementations in different technical stages of its full cycle. Based on the current network ecology, engineering technology, application needs, and other factors, this section explores the engineering practice of the PNE.
5.1. Experimental environment
Based on the research and development (R&D) achievements of China’s 13th Five-Year National Key R&D Program "The core technology and principle platform of the polymorphic network," a PNE consisting of one core domain and six access domains was built, as shown in Fig. 5. The core domain consists of four type-I polymorphic network elements and inter-domain controllers, with a link bandwidth of 100 gigabits per second (Gbps) between the network elements. The access domain consists of type-II polymorphic network elements, intra-domain controllers, various fixed/mobile terminals, and so forth. More specifically, the type-I and type-II elements are two different types of routing devices that are designed to support the polymorphic network. The link bandwidth with the core domain is , and the link bandwidth between wired nodes in the domain is 1000 or 100 megabits per second (Mbps). When congestion does not occur, the forwarding delay of a single network element is less than , and the packet loss rate is less than . The maximum switching capacity of the type-I polymorphic network elements is 12 terabits per second (Tbps), and the maximum switching capacity of the type-II polymorphic network elements is 6 Tbps. The two types of network elements were both implemented based on field-definable network baseline technology integrating storage, computation, and forwarding, which can generate various network modalities by definition, support the coexistence and independent operation of various network modalities, and ensure their respective preset performance through elastic isolation among bandwidth resources. The controllers can not only compute and issue the forwarding table of each network modality according to the network topology and routing protocol but also compile the target network code designed with the P4+ programming language, so as to realize the dynamic loading/unloading of the corresponding network modality on the network element devices and the version update of the full network cycle.
We divide the experiment into three parts to progressively verify the feasibility of the PNE and the advancement of its theory: ① verifying the transparency of the PNE when carrying multiple network modalities and ensuring the functional integrity of each network modality; ② measuring the adaptability of the various network modalities in the PNE to diversified services; and ③ testing the isolation among network modalities-that is, in case of resource conflicts, the preset bandwidth/performance of each network modality should be guaranteed.
5.2. Test analysis
5.2.1. Coexistence of diversified network modalities
We aim to verify that the PNE can support the generation, co-network bearing, and operation of various application network systems. As shown in Fig. 5, six well-defined and clearly outlined application network modalities, including IPv4, IPv6, GeoNetwork-ing [17], MobilityFirst, PowerLink [18], and NDN were deployed in the network elements of the PNE, and the corresponding application services, industrial devices (special terminals), and testers were deployed in the access network. Under the condition that the network was not congested, we simulated the applicable scenarios for each network modality and evaluated the basic service capabilities of each network modality. Here, we only discuss the applicable scenarios and test results for non-IP modalities:
(1) GeoNetworking: The GeoNetworking architecture was formulated by the European Telecommunications Standards Institute (ETSI). Its core idea is based on the node’s geographical location information (not the IP address) to route and support the point-to-point, point-to-multipoint, point-to-any point in the area, point-to-area, and other transmission modes using geolocation addressing. It is suitable for communication service scenarios such as the Internet of Vehicles, intelligent transportation, and environmental monitoring. The PNE supports the addressing and routing of messages based on geographical information and can accurately push messages and obtain information within a specified geographical area. Compared with the IP modality, in a situation of message broadcasting within a fixed geographical area, such as Internet of Vehicles communication, the GeoNetworking modality has better efficiency.
(2) NDN: The NDN architecture is content-centric, and its data-transmission mode disregards the storage location of content data; instead, it names all content data and carries out information retrieval through name matching, so as to establish a distributed communication network. It is suitable for service scenarios such as content distribution and hot content pushing. The PNE supports the addressing and routing of messages based on content identifiers, network retrieval, message delivery, network file transmission, and other functions of content files, and supports pushing content files to users’ edge networks through customized messages according to users’ interest requests. Compared with the IP modality, the NDN modality can reduce the users’ request delay.
(3) MobilityFirst: The mobility-centric MobilityFirst network architecture assigns a fixed and globally unique identifier (GUID) to each device (and contents) in the network. These devices are mapped to the corresponding network addresses (NA) according to their locations in the network, thereby realizing the separation of device identifiers from NA and the location determination of these devices during their movement. The PNE supports the addressing and routing of messages based on identities, provides dynamic access for mobile terminals, and supports message delivery and file transfer resuming under the dynamic access of mobile terminals. Compared with the IP modality, this modality can maintain the connection state of end-to-end data transmission before and after the mobile terminal switches to the network, achieving a zero-packet loss rate.
(4) PowerLink: The PowerLink architecture performs the addressing and forwarding of messages based on industrial control device identifiers (IDs) and adopts technologies such as isochronous synchronization and polling sequence to solve the real-time transmission problem of industrial control data. It is suitable for applications such as servo motor control in a factory and remote control outside of a factory. The PNE supports direct message addressing and forwarding based on industrial device IDs. Compared with the IP modality, the PowerLink modality has lower delay and deterministic jitter, thereby satisfying the communication requirements of industrial control devices for high real-time and precise control.
The above experiments show that the PNE can not only support the generation, co-network bearing, and operation of various application network systems (i.e., network modalities) but also guarantee the basic service performance of each network modality, thereby allowing the network services to choose different network modalities to adapt to their communication modes and service quality requirements.
5.2.2. Classified carrying of diversified services
In this section, we aim to verify the law of requisite variety of the PNE-namely, that diversified network services need diversified network modalities-so that the PNE can provide better overall QoS with the same physical resources. In this experiment, we choose two types of network services-video call (peer-to-peer) and video conference (peer-to-peers)-as examples and adopt IP and NDN modalities suitable for content transmission to respectively carry these two types of services. For the two types of services, the clients were deployed on the host of each access domains of the PNE, as shown in Fig. 5, while the server side of the video conference was fixedly deployed on server . We divide the experiment into three types of comparative tests: ① An IP network carries the two types of services simultaneously; ② an NDN carries the two types of services simultaneously; and ③ a PNE network platform is used for classified carrying, an IP network modality is adopted to carry the video-calling service, and an NDN modality is adopted to carry the video-conferencing service. In these three types of tests, the request traffic injected by the clients is the same, and the request traffic of the clients of the video conference is set to be 1.2 times the bandwidth of the link . Therefore, when the IP network carries the video-conference service, link congestion often occurrs on the link due to the point-to-multipoint communication mode of this service.
As can be seen from Fig. 6, there is a low service request delay when the unimorphic IP network carries the video call service; however, when it carries the video-conference service, there is a sharp rise in service requests due to the traffic congestion of the link . When the unimorphic NDN carries the video conference service, as the traffic congestion of the critical path is avoided because of the content addressing and in-network caching, there is a lower service request delay than in the case of the video-conference service carried by the IP network. However, compared with the video-call service carried by the IP network, the service request delay of the video-call service carried by the NDN is higher, because the video call cannot take advantage of the cache of the NDN. Moreover, the routing switching process of the NDN is complicated and time-consuming. These findings also demonstrate from the side that it is difficult for a unimorphic network system to achieve the best performance in all services. Compared with the unimorphic IP network or NDN carrying, the classified carrying of the PNE makes the service request delay of the two types of services optimal, because the PNE can flexibly generate diversified network modalities to carry diversified network services in a unified network basic environment according to the law of requisite variety, which provides the best match between user service, network modalities, and physical resources, thereby providing a higher service quality.
5.2.3. Isolation among network modalities
In the simultaneous operation of multiple network modalities, there may be competitions for resources such as bandwidth, queues, cross switches, and so forth. Therefore, isolation among the network modalities is an important evaluation indicator of the PNE, which relates to the issue of whether each network modality operating independently can obtain the agreed-upon service level. When a network modality is deployed in the PNE, it is allocated with a certain amount of bandwidth resources; if its actual service traffic can always be guaranteed when it does not exceed the pre-allocated bandwidth and is not affected by other modalities, then this network modality will obtain an isolation effect similar to the exclusive use of network resources. In the experiment, the tester is connected to the ports on , and and sends the mixed analog traffic of the above six network modalities from and to . The traffic injected by different network modalities is adjusted to cause the bandwidth conflict on link . The tester conducts a statistical analysis of the delay and packet loss rate of the IPv4 modality messages from the port of . Here, the injected traffic of the IPv4 modality is marked as "IPv4 Traffic," and its reserved bandwidth is set as 4 Gbps, while the sum of the injected traffic of other network modalities is marked as "Other Traffic," which shares a physical link bandwidth of 10 Gbps on link with the IPv4 traffic.
The packet loss rate and average delay of the IPv4 reflects whether it is affected by interference from other network modalities. The experimental results are shown in Fig. 7. It can be seen that:
(1) When the IPv4 traffic is less than 4 Gbps, its packet loss rate tends to be zero, and the average delay is nearly , even if the injected traffic of the other network modalities exceeds 6 Gbps. This indicates that the PNE can ensure isolation between network modalities, allowing them to obtain an operation effect equivalent to the exclusive use of the reserved network resources.
(2) When the IPv4 traffic is greater than 4 Gbps, if the total traffic of the PNE does not exceed 10 Gbps, then its packet loss rate also tends to be zero, and the average delay is nearly . This indicates that the PNE is not limited to the setting of reserved bandwidth. If there is surplus physical bandwidth, it can be provided to a certain network modality to realize elastic sharing of the resource under the condition of a unified infrastructure.
(3) When the traffic of the IPv4 is greater than 4 Gbps and the total traffic in the PNE is greater than , the packet loss rate of the IPv4 is significantly larger, and the average delay is close to . It can be seen that the PNE possesses the dual attributes of modality isolation and resource sharing; that is, the PNE can not only create an application network system that operates independently but also provide resources beyond the upper limit of the physical dedicated network.
6. Summary and prospects
In this paper, we reviewed the current network development paradigm, revealed for the first time that there is an SMV dilemma between the diversity of network services and application scenarios and unimorphic rigid-network systems under the premise of satisfying global scalability in the full network cycle, and then established an universal deconstruction model of the "impossible SMV triangle" to answer the question: Why is it impossible for available practical network systems to break through the SMV dilemma? We also pointed out that any engineering route that does not meet the law of requisite variety of cybernetics cannot achieve the goal of an ideal network. Enlightened by the unity of the physical world and the diversity of the material world, we functionally deconstructed the available unimorphic network system functions and software and hardware resources-such as computing, storage, and interconnection-into elements, and then established a more advanced functional module by using the "chemical bond" effect of SDI. Finally, we were able to generate diversified application network modalities based on the DSA method, which can not only carry well-defined expected services with clear application scenarios but also support extended services with clear application scenarios in the future. In particular, we propose for the first time a polymorphic network development paradigm based on the spatial and temporal dimensions of SMVT, which decouples the application network system from the supporting infrastructure environment. This new paradigm breaks through the current SMV dilemma of an Internet development route based on the separation of services and network, and has the capability to support the coexistence, evolution, and revolution of multiple application network systems and services within a unified network infrastructure environment.
Furthermore, we expounded the core idea and theoretical framework of PNE, established a mathematical model of PNE, and demonstrated the necessity of developing PNE. Finally, we presented a PNE environment test for first-principles verification and conducted a detailed analysis of the experimental results. The preliminary test showed that the prototype network environment can support the simultaneous deployment of more than six application network modalities and well-defined services with clear application scenarios. The theoretical expectation is highly consistent with the experimental results, which proves that the PNE has the technical economy required by engineering practice. In the future, we will theoretically prove the correctness of the corollary that "structure determines diversity;" define the boundary conditions of the related engineering techniques; study key techniques such as the elemental extraction method of various application network systems, the flexible scheduling of software and hardware resources, and trusted services and security guarantees; and develop an element library for generative application network modality and toolkits or toolchains for the compilation environment, network deployment, performance evaluation, operation and maintenance support, and so forth.
Acknowledgments
This work was supported by the National Key Research and Development Program of China (2022YFB2901403) and the Songshan Laboratory Project (221100210900-02).
Compliance with ethics guidelines
Jiangxing Wu, Junfei Li, Penghao Sun, Yuxiang Hu, and Ziyong Li declare that they have no conflict of interest or financial conflicts to disclose.
J. Wu. On the revolution of the information network development paradigm. Sci China Inf Sci, 65 (11) (2022), Article 213301.
[2]
DaiM. Research on networking technology of digital terrestrial television single frequency network. In: Proceedingsof 2020 International Wireless Communications and Mobile Computing IWCMC; Jun15-192020; Limassol, Cyprus; 2020.
[3]
A.S.Acampora, M.Naghshineh. An architecture and methodology for mobile-executed handoff in cellular ATM networks. IEEE J Sel Areas Comm, 12 (8) (1994), pp. 1365-1375.
[4]
H.Yuksel, Ö.Altunay. Host-to-host TCP/IP connection over serial ports using visible light communication. Phys Commun, 43 (2020), Article 101222.
[5]
J. Wu. Thoughts on the development of novel network technology. Sci China Inf Sci, 61 (10) (2018), Article 101301.
[6]
NaikP, ShawDK, VutukuruM. NFVPerf: online performance monitoring and bottleneck detection for NFV. In:Proceedings of 2016 IEEE Conference on Network Function Virtualization and Software Defined Networks (NFV-SDN). 2016 Nov 7- 10; Palo Alto, CA, USA; 2016.
[7]
X.Fei, F.Liu, Q.Zhang, H.Jin, H.Hu. Paving the way for NFV acceleration: a taxonomy, survey and future directions. ACM Comput Surv, 53 (4) (2020), p. 73.
[8]
H.Zhang, W.Quan, H.Chao, C.Qiao. Smart identifier network: a collaborative architecture for the future Internet. IEEE Netw, 30 (3) (2016), pp. 46-51.
[9]
L.Zhang, A.Afanasyev, J.Burke, V.Jacobson, K.Claffy, P.Crowley, et al. Named data networking. Comput Commun Rev, 44 (3) (2014), pp. 66-73.
[10]
D.Raychaudhuri, K.Nagaraja, A.Venkataramani. MobilityFirst: a robust and trustworthy mobility-centric architecture for the future internet. Mob Comput Commun Rev, 16 (3) (2012), pp. 2-13.
[11]
J.Farkas, L.L.Bello, C. Gunther. Time-sensitive networking standards. IEEE Commun Stand Mag, 2 (2) (2018), pp. 20-21.
[12]
H.Zhang, W.Su, W.Quan. Smart collaborative identifier network: a promising design for future Internet. Springer, Berlin (2016).
[13]
LiR, MakhijaniK, DongL. New IP: a data packet framework to evolve the Internet. In:Proceedings of IEEE 21st International Conference on High Performance Switching and Routing (HPSR). 2020 May 11- 14; Newark, NJ, USA; 2020.
[14]
W.R.Ashby. An introduction to cybernetics. Chapman & Hall, London (1956).
[15]
R.M.May. Will a large complex system be stability>. Nature, 238 (5364) (1972), pp. 413-414.
[16]
P.Petersen. Riemannian geometry. Springer, New York City (2006).
[17]
F.Cadger, K.Curran, J.Santos, S.Moffett. A survey of geographical routing in wireless ad-hoc networks. IEEE Commun Surv Tutor, 15 (2) (2013), pp. 621-653.
[18]
L.Seno, S.Vitturi, C.Zunino. Analysis of Ethernet PowerLink wireless extensions based on the IEEE 802.11 WLAN. IEEE Trans Industr Inform, 5 (2) (2009), pp. 86-98.
PDF
(1708KB)
