Networked systems form the backbone of modern society, underpinning critical infrastructures such as electricity, water, transport and commerce, and other essential services (e.g., information, entertainment, and social networks). It is almost inconceivable to contemplate a future without even more dependence on them. Indeed, any unavailability of such critical systems is - even for short periods - a rather bleak prospect. However, due to their increasing size and complexity, they also require some means of autonomic formation and self-organization. This paper identifies the design principles and open research issues in the twin fields of self-organization and resilience for networked systems. In combination, they offer the prospect of combating threats and allowing essential services that run on networked systems to continue operating satisfactorily. This will be achieved, on the one hand, through the (self-)adaptation of networked systems and, on the other hand, through structural and operational resilience techniques to ensure that they can detect, defend against, and ultimately withstand challenges.

There is a need for new architectures and designs of resilient networked systems that are capable of supporting critical services and infrastructures. The arguments have previously been well rehearsed, but much remains to be done, not least to demonstrate the feasibility of building such systems. Key among the remaining challenges is how to specify and realise appropriate components that interact with each other to produce a resulting resilient system. This paper reviews the state of the art, describes recent contributions, and looks ahead to future research and prospects.

We propose a cross-layer routing framework in the SDN (software-defined networking) domain to cope with regionally-correlated challenges. By taking advantage of the failure detection model, GeoDivRP calculates multiple geodiverse paths for resilient network communications. Coupled with the optimization model, it realizes the minimized delay-skew product when decoupling traffic onto multiple paths. We evaluate our framework using MPTCP (Multipath TCP) in the face of regionally-correlated failures and it presents better performance compared to the single path routing. We further demonstrate our web framework to automate the OpenFlow experiment by programmatically importing network topologies and execute challenge emulations using the user-provided challenge regions.

With the increasing frequency of natural disasters and intentional attacks that challenge communication networks, vulnerability to cascading, and regional‐correlated challenges is escalating. Given the high complexity and large traffic load of communication networks, these correlated challenges cause substantial damage to reliable network communication. In this work, we extend the GeoDivRP routing protocol to consider delay‐skew requirement when using multiple geographically diverse paths for telecommunication networks under area‐based challenges. We present a flow‐diverse minimum‐cost routing multicommodity flow problem. Furthermore, we present a nonlinear delay‐skew optimization problem to balance between delay and traffic skew on paths. We investigate the tradeoff between the delay and skew in choosing multiple geodiverse paths. We implement GeoDivRP in ns‐3 to use the optimized paths given by the two optimization solutions and demonstrate their effectiveness compared to open shortest path first Equal‐Cost Multi‐Path routing in terms of overall link utilization. It guarantees the delay‐skew constraint provided by the upper layer while satisfies the traffic demand imposed by multiple routing commodities in the telecommunication networks. © 2015 Wiley Periodicals, Inc. NETWORKS, Vol. 66(4), 335–346 2015

Internet and web services have seen widespread adoption in recent years and are now tightly integrated into society's daily activities. An important emerging part of the Internet is clouds that provide low-cost configurable computing resources, allowing businesses to reduce their hardware, software, and personnel costs. Increasingly, enterprises now use such cloud resources to host web applications. While clouds provide an excellent business model, most existing public and private cloud infrastructures are based on monocultures that allow attackers to focus their efforts on a single hardware/software platform and facilitates the rapid spreading of successful attacks. In this invited paper, we describe a methodology and mech- anisms that make clouds and hosted applications considerably more resilient to attacks and correlated failures by introducing diversity at every level of the cloud: physical interconnect, network components, processor platforms, storage management, virtual machine monitors, operating systems, and application processes. Our goal is to defend against attacks by continuing to operate correctly even when part of the infrastructure fails and to substantially raise the difficulty of executing a successful attack by requiring the attacker to simultaneously target different hardware and software choices. Furthermore, by geographically spreading applications among different datacenters using diverse network connections (in service provider and access medium - wired vs. wireless), the cloud will be resilient against physical infrastructure attacks and large-scale disasters.

End-to-end communication in highly-dynamic airborne networks is challenging due to the presence of highly mobile nodes and the inherent nature of wireless communication channels. Domain-specific protocols are required that can address these challenges and enable reliable transmission of data in this environment. We develop the ANTP (airborne network and transport protocols) suite that operates in this highly-dynamic environment while utilising cross-layer optimisations between the physical, MAC, network, and transport layers. We show how each component in the ANTP suite outperforms the traditional TCP/IP and MANET protocols through simulation using ns-3. Having verified these protocols through simulation and analysis, the next step towards deployment of the ANTP suite is developing a cross-platform implementation of the protocols. Towards this end we present an architecture for the protocol stack to be implemented in the Python programming language.

A thorough understanding of the network behaviour when exposed to challenges is of paramount importance to construct a resilient MANET (mobile ad hoc network). However, modelling mobile and wireless networks as well as challenges against them is non-trivial due to dynamic and intermittent connectivity caused by channel fading and mobility of the nodes. We treat MANETs as time-varying graphs (TVGs) represented as a weighted adjacency matrix, in which the weights denote the link availability. We present how centrality-based attacks could affect network performance for different routing protocols. Furthermore, we model propagation loss models that represent realistic area-based challenges in wireless networks.

Great Plains Environment for Network Innovation (GpENI) is an international testbed for future Internet research. A key component of GpENI is programmable network virtualization (GpENI-VINI). The scope of this paper is to present the framework, implementation and integration experience with network virtualization in GpENI. In particular, this is described through our experience of implementing and integrating the XORP (eXtensible Open Router Platform) routing platform into GpENI-VINI. Preliminary results on measurements and validation are presented.

Emerging networked systems require domain-specific routing protocols to cope with the challenges faced by the aeronautical environment. We present a geographic routing protocol AeroRP for multihop routing in highly dynamic MANETs. The AeroRP algorithm uses velocity-based heuristics to deliver the packets to destinations in a multi-Mach speed environment. Furthermore, we present the decision metrics used to forward the packets by the various AeroRP operational modes. The analysis of the ns-3 simulations shows AeroRP has several advantages over other MANET routing protocols in terms of PDR, accuracy, delay, and overhead. Moreover, AeroRP offers performance tradeoffs in the form of different AeroRP modes.

With growing demand for high-speed access to mobile handheld devices, there is a significant cost benefit in deploying fixed wireless-mesh networks for backhaul access. However, enabling reliable broadband access over high-frequency radios (such as millimeter-wave networks) posses a fundamental challenge due to weather disruptions in general and rain attenuation in particular. In this paper, we present an analysis of the impact of precipitation on millimeter-wave mesh networks based on radar measurements of real storms in the Midwest US. Furthermore, we compare two novel algorithms that use physical-layer information to optimize routing at the network layer: P-WARP (Predictive Weather-Assisted Routing Protocol) and XL-OSPF (Cross-Layered Open Shortest Path First). Finally, we present simulation studies to compare the performance of the proposed protocols and evaluate the dependability of the end-user service during weather disruptions.

A flash crowd event can be characterised by a dramatic increase in requests for a service over a relatively short period of time. Often, these events lead to a loss of service because of the saturation of the target server and associated network resources. This paper presents a set of mechanisms that can be used to make Web servers and associated resources more resilient to flash crowd events. Specifically, we present a novel admission control mechanism that uses a detection mechanism we developed in earlier work to adjust the admission rate of HTTP requests to a Web server. We demonstrate, via simulations, that the admission control mechanism can be used to protect a Web server from the effects of a flash crowd event, protect the traffic of other services that are hosted on the same network as a targeted Web server, and in combination with a push-back mechanism reduce the effect of flash crowd traffic on an ISP's network that is serving the Web server. The mechanisms presented here are exemplars that fit within a resilience strategy we are developing - D2R2+DR - which is summarised here.

Emerging network technologies offer new design choices for performing traffic engineering and management, while introducing challenges for existing protocol control mechanisms. There is a tension between new hop-by-hop capabilities and existing end-to-end mechanisms. We examine one such capability, the load-reactive link (LRL), which has a capacity that can be automatically varied in response to the offered load. Before such adaptive links can be considered for deployment, it is essential to understand the behavior of widely deployed end-to-end protocols, in particular TCP. We study the behavior of TCP flows over a particular type of LRL that uses a hysteresis control mechanism for capacity allocation that reacts to current traffic loads. Our results indicate that reasonable performance for a single TCP flow can be achieved with careful choice of control parameters. We also show that poor choice of control parameters make it difficult or impossible for TCP to function. Preliminary simulations have been done with multiple TCP flows, but further investigation is needed to understand the behavior of the aggregation of a large number of flows.

A significant driver for the consumer use of high bandwidth in the near future will be interactive video on demand (IVOD). A range of service types can be deployed, based on a differing sophistication, which must be traded against the network costs (bandwidth) and component costs (switch complexity and memory). The potential aggregate bandwidth required is huge (O(1Pb/s)), and thus it is essential to properly engineer the network to reduce the bandwidth required. This paper describes a variety of IVOD scenarios, and introduces a cost function that captures the combined bandwidth and storage requirements of the network. This cost function is used to compare different network engineering alternatives, particularly program caching and stream sharing. The effects of nonlinear pricing and differing weights of bandwidth and storage are also reflected by the cost function. This cost function can be used by network designers to determine optimal topology, sharing, and caching strategies for desired bandwidth versus memory costs in a particular network deployment. In addition, a simulation model is used to evaluate caching of programs or windows within programs. We show that there are some results that are widely applicable. In particular, the level in the network at which caching should take place is at approximately 80% depth in the distribution tree, above the head end switch in the network hierarchy. We also observe that the bandwidth savings in sharing streams (actually buffered windows of program content) is fairly small for user behavior based on Zipfs law. The overall intent of this work is to evaluate the effects of various server, cache, and sharing strategies on the bandwidth and storage requirements of the network and their proper placement within the network.>

The primary goal of the Axon architecture is to support a high-performance data path delivering high network bandwidth directly to applications. The Axon network interface is described from the perspective of its simulation, and in particular and implementation of error and flow control in hardware. Background is provided on the simulation package that has been used to explore these mechanisms, and a brief overview is given of the Axon architecture is implemented by the simulator. The error control mechanism, the hardware design, and the functional and performance simulation results are outlined. The rate control scheme is described, along with its implementation and simulation.>

An introduction to the Axon architecture is presented. Axon is a host communication architecture for the distributed systems which can support interprocess communication with high throughput and low latency across a very high-speed internetwork. Network virtual storage (NVS) in the Axon host communication architecture for distributed applications is described. NVS extends segmented paged virtual storage management and address translation mechanisms to include segments located across an internetwork. This provides the ability to use efficiently the shared-memory paradigm in distributed systems for high-performance applications, such as scientific visualization and imaging.>

The primary goal of the Axon architecture is to support a high-performance data path delivering VHSI (very high speed internetwork) bandwidth directly to applications. The significant features of Axon are: (1) an integrated design of host and network interface architecture, operating systems, and communication protocols; (2) a network virtual storage facility which includes support for virtual shared memory on loosely coupled systems; (3) a high-performance, lightweight object transport facility which can be used by both message-passing and shared-memory mechanisms; and (4) a pipelined network interface which can provide a path directly between the VHSI and host memory.>