Cloud providers may operate large-scale data centers in a few locations. We argue that deploying many small-scale data centers at network edge can significantly improve user experience in terms of latency. Small-scale data centers, however, may not be able to provide elastic services. In this paper, we investigate distributed small-scale data centers with load reallocation where jobs that cannot be suitably processed locally will be reallocated to remote data centers. We formulate an optimization problem for load reallocation in distributed data centers, provide performance comparisons among different alternatives and offer insights on handling multiple job types. We develop online optimization algorithms that can be operated in a decentralized and measurement-based fashion to dynamically reallocate load in response to sudden load surges. The experimental results demonstrate that elasticity can be practically provided by small-scale data centers enhanced with effective load reallocation techniques.

Saving power in datacenter networks has become a pressing issue. While in operation, ElasticTree and CARPO can save power consumed by a fat-tree network by using sleep mode where some components such as ports and switches are turned off when traffic demand in the network is relatively moderate. In this paper, we propose a new approach by exploring the design stage of a datacenter network and focus on how to choose the right switch size that can potentially save the most power during the expected operation of the network. We also consider speed scaling where the power of a switch can be varied by adjusting its processing rate according to its traffic demand. We use analysis and simulation to investigate the power-saving performance of different switch sizes, power-saving modes and traffic demand patterns. Our findings with sleep mode reveal that deploying a large number of small switches is more power-efficient than a small number of large switches when the traffic demand is relatively moderate or when servers exchanging traffic are in close proximity. With speed scaling, the reverse is generally true.

The Internet plays an increasing role in content dissemination as user-generated contents have exploded recently. Cache servers have been deployed to bypass bottlenecks in the network so that contents can be delivered to end users more efficiently. With caches becoming more embedded in the networks, emerging threats follow naturally. A cache pollution attack is one of the most serious threats on caching networks including the current Internet and emerging caching networks such as Content Centric Networking (CCN). In this paper, we propose a detection approach against cache pollution attacks using randomness checks of a matrix. We apply an effective filtering approach and a statistical sequential analysis for detecting low-rate attacks. The results of our experiments show that our approach can detect a cache pollution attack with attack rate of only a few percent of the overall rate.

Content Centric Networking (CCN) is a new and promising architecture that can provide better security and much improved content delivery service than the current Internet. Each router in CCN has a content store that caches objects according to a replacement algorithm and a pending interest table that stores pending requests waiting for matching objects. We propose a novel and flexible resource management system where the content store can store objects as well as pending requests. To each pending request, we add control fields that can be used to control caching resources according to a given objective function. We examine several applications that can benefit from using the proposed resource management system. The experimental results quantify key performance improvements.

With the advent of content-centric networking (CCN) where contents can be cached on each CCN router, cache robustness will soon emerge as a serious concern for CCN deployment. Previous studies on cache pollution attacks only focus on a single cache server. The question of how caching will behave over a general caching network such as CCN under cache pollution attacks has never been answered. In this paper, we propose a novel scheme called CacheShield for enhancing cache robustness. CacheShield is simple, easy-to-deploy, and applicable to any popular cache replacement policy. CacheShield can effectively improve cache performance under normal circumstances, and more importantly, shield CCN routers from cache pollution attacks. Extensive simulations including trace-driven simulations demonstrate that CacheShield is effective for both CCN and today's cache servers. We also study the impact of cache pollution attacks on CCN and reveal several new observations on how different attack scenarios can affect cache hit ratios unexpectedly.

With the convergence between phone and data networks in LTE and 4G, cellular signaling traffic is increasingly carried over IP. Control plane functions, once performed by dedicated machinery, are evolving into large-scale network applications with strict requirements on delay, availability, and processing throughput as mandated by 3GPP standards. In this paper we present DMME, a distributed architecture that implements mobility management for next-generation cellular systems. DMME is a scalable and cost-effective drop-in replacement for the LTE mobility management entity (MME). We evaluate the DMME scheme via analysis and simulation under several deployment scenarios, using mobility patterns drawn from synthetic models and from traces collected in a production network. We also report preliminary performance figures by our DMME prototype implementation. Our results confirm that distributed architectures are a viable choice to reliably support high-throughput, latency-sensitive control plane functions.

LTE is currently considered by many wireless operators as their next generation technology. During a transition phase, LTE will be deployed as an overlay on an existing wireless technology, which acts as an underlay. It is expected that initial LTE deployment will be patchy with coverage holes. In this paper, we model and analyze the impact of dual-technology deployment on signaling load at critical network elements. We consider a case, known as without ISR, where a location update is executed when a UE moves between an overlay and an underlay and another case, known as with ISR, where a location update is made only when a new tracking area or routing area is encountered by a UE. Given a fixed total area of coverage holes in an overlay, our result shows that update rate increases as a square root of the number of coverage holes without ISR while it is virtually constant with ISR.

The LTE architecture consists of eNBs that provide wireless connectivity to each UE, S-GWs that anchor the user plane for mobility, MMEs that provide control-plane support to each UE in its domain such as handover, paging, bearer management and tracking area updates and finally, PDNs that bridge the cellular network to the Internet. In this paper, we investigate LTE architectures with distributed MMEs and with a centralized MME. We present an analysis of the signaling load performance at the MME serving multimedia-capable UEs and compare the performance under both architectures. We propose a multicast paging procedure to alleviate the MME signaling load and compare the MME performance with multicast and unicast paging procedures. Our results show significant benefits of multicasting.

LTE is a next-generation wireless network technology being developed by 3GPP. One unique aspect of LTE is that neighboring eNBs can be directly interconnected through X2 interface to facilitate direct exchange of control messages and to forward user data during handover. In this paper, we investigate the X2 bandwidth requirement for both control-plane traffic and user-plane traffic. We first consider the case where individual UEs cross cell boundaries independently. We then extend to the case where groups of UEs may cross cell boundaries simultaneously. In both cases, we analyze the bandwidth needed on the X2 interface to meet a certain performance objective. Our results generally indicate that the X2 bandwidth requirement is very small compared to Sl-U bandwidth. However, the X2 bandwidth requirement may potentially increase significantly when groups of UEs perform handover simultaneously.

The problem of identifying applications online and directly from traffic flows recently has been a subject of great interest. Traditional techniques relying on port numbers or payload signatures are becoming less effective. In this paper, we present an approach to online identification of applications using statistical behavior analysis. We investigate both host- level identification and flow-level identification. For each level, we define the suitable metrics that can be computed fast and effectively exploited by the identification process. We propose to use decision trees to identify applications with low computation complexity, which is required for high-speed online processing. Our experimental results using BitTorrent, HTTP, SMTP and FTP traffic traces demonstrate that our technique can identify these applications with low error rates and short delay.

The session initiation protocol (SIP) is rapidly being adopted as the signaling protocol for establishing, modifying and terminating multimedia sessions. With the increasing use of SIP in large deployments, it is now becoming apparent that the current SIP design does not easily scale up to large network sizes and SIP servers are not well equipped to handle overload conditions. When a SIP server is operating close to or above its capacity limit, message retransmissions by various SIP timers can cause the network to be severely overloaded and result in an extremely low goodput. In this paper, we first provide a detailed analysis of the behavior of SIP servers under overload. We show that SIP servers are often unable to recover from congestion collapse once it has occurred and that overload can spread throughout a network of SIP servers. We then discuss mechanisms and algorithms for controlling overload in these servers. We found that performing overload control locally at a server provides a simple remedy for light cases of overload; however, it is ineffective in handling higher amounts of load. Finally, we investigate distributed overload control mechanisms for SIP and show that they are effective in controlling overload of SIP servers.

The emerging packet network architectures with decoupled service plane and transport plane may play a prominent role in the future. Traditional packet network architectures are primarily concerned with providing quality "pipes" (better QoS, high bandwidth, more efficient and reliable transport, etc.). The emerging architectures shift the focus towards networks that readily enable new advanced services and allow operators to have deep control over their networks. We describe related architectures with decoupled service and transport planes that are currently being addressed by various standards development organizations such as 3GPP, 3GPP2, ETSI TISPAN, ITU-T NGN, Packet Cable and MSF. We explore various issues that need to be considered with these network architectures with a focus on resource management.

Time-domain Wavelength Interleaved Networking (TWIN) is a cost-effective network architecture that can provide fine-grained, flexible connectivity using passive optics in internal nodes. Each wavelength is dedicated to carry traffic to a single node. We introduce TWIN with wavelength reuse (TWIN-WR), an extension of TWIN that allows the number of nodes to exceed the number of wavelengths. We formulate an associated design problem, decompose it into subproblems, and provide algorithms for solving the subproblems. We experimentally demonstrate that our algorithms come close to upper bounds over a wide range of parameters, and show that the resulting designs are robust in the face of traffic variability and uncertainty. Analysis and experiment show that high efficiency can be maintained as long as the number of nodes does not exceed the square of the number of wavelengths, in principle allowing a network with thousands of nodes using only one hundred wavelengths.

We consider meshed optical transport networks having multiple levels of hierarchy whereby cross-connects in different levels perform multiplexing and demultiplexing functions at different granularities. We present a traffic grooming approach that can be implemented in a centralized or distributed fashion, based on the novel design that takes advantage of algorithm efficiency and a simple threshold mechanism to decide when grooming is economical. Our approach allows nodes to perform grooming and degrooming automatically, which is desired in the next-generation optical network where circuits are setup and torn-down dynamically through signaling. We present several experiments using different network topologies. The results indicate that the flow pattern influences the port requirement behavior, and that the flow thickness influences the optimal value of the threshold.

Link bundling is a way to increase routing scalability whenever a pair of label switching routers in MPLS are connected by multiple parallel links. However, link bundling can be inefficient as a label switched path (LSP) has to be associated with a particular link. In this paper, we show that the efficiency of link bundling can be significantly improved if traffic can be effectively distributed across the parallel links. We propose an IP switch architecture that is capable of distributing flows both inside the switch and among the parallel links based on operations that are relatively simple to implement. The switch requires no speedup, guarantees in-sequence packet delivery for a given flow, avoids complex coordination algorithms, and can achieve LSP throughput higher than the line rate. By means of simulation using IP traces, we investigate the performance of the proposed switch, and show that the switch achieves good load-balancing performance. We describe extensions to the basic architecture which allows for very large bundle size, handles incremental upgrade strategy, improves reliability, and accommodates nonIP traffic.

A time-domain wavelength interleaved network (TWIN) (Widjaja, I. et al., IEEE Commun. Mag., vol.41, 2003) is an optical network with an ultrafast tunable laser and a fixed receiver at each node. We consider the problem of scheduling bursts of data in a TWIN. Due to the high data rates employed on the optical links, the burst transmissions typically last for very short times compared with the round trip propagation times between source-destination pairs. A good schedule should ensure that: 1) there are no transmit/receive conflicts; 2) propagation delays are observed; 3) throughput is maximized (schedule length is minimized). We formulate the scheduling problem with periodic demand as a generalization of the well-known crossbar switch scheduling. We prove that even in the presence of propagation delays, there exist a class of computationally viable scheduling algorithms which asymptotically achieve the maximum throughput obtainable without propagation delays. We also show that any schedule can be rearranged to achieve a factor-two approximation of the maximum throughput even without asymptotic limits. However, the delay/throughput performance of these schedules is limited in practice. We consequently propose a scheduling algorithm that exhibits near optimal (on average within ∼7% of optimum) delay/throughput performance in realistic network examples.

Generalized multi-protocol label switching (GMPLS) is currently being specified as the control plane for next-generation optical networks. This paper investigates how GMPLS affects the performance of lambda-router networks. Our performance study reveals the advantage of using a fully tunable laser at each source in the presence of a limited number of wavelength converters. We propose an enhancement to GMPLS which allows a lightpath request to crankback to a lambda router that can use a wavelength converter to complete the lightpath setup successfully. This effectively makes the wavelength converters shared across the network. It is shown that crankback provides an appreciable improvement in performance when fully tunable lasers at the sources are not universally available.

Destination-based forwarding in traditional IP routers has not been able to take full advantage of multiple paths that frequently exist in Internet service provider networks. As a result, the networks may not operate efficiently, especially when the traffic patterns are dynamic. This paper describes a multipath adaptive traffic engineering mechanism, called MATE, which is targeted for switched networks such as multiprotocol label switching (MPLS) networks. The main goal of MATE is to avoid network congestion by adaptively balancing the load among multiple paths based on measurement and analysis of path congestion. MATE adopts a minimalist approach in that intermediate nodes are not required to perform traffic engineering or measurements besides normal packet forwarding. Moreover MATE does not impose any particular scheduling, buffer management, or a priori traffic characterization on the nodes. This paper presents an analytical model, derives a class of MATE algorithms, and proves their convergence. Several practical design techniques to implement MATE are described. Simulation results are provided to illustrate the efficacy of MATE under various network scenarios.

Multi-protocol over ATM (MPOA) is being considered by the industry as an important short-cut technology that provides an efficient transfer of inter-subnet unicast data in a LANE environment. MPOA was initially considered to carry backbone traffic in the enterprise or campus networks. However, congestion in the public Internet provokes many to consider MPOA as a solution for the carrier or service provider networks as well. In this paper, we investigate the scalability issues of MPOA in the wide area network environment. We use a realistic simulation model driven by real Internet traffic to study crucial metrics such as the SVC setup rate, the number of VCs required, and the percentage of packets switched. We find that MPC ingress cache size provides a three-way trade-off among the percentage of switched packets, the VC usage and the SVC setup rate requirement. We also find that the SVC setup rate is linearly dependent on the packet arrival rate.

In a multiprotocol label switching (MPLS) domain, ATM label-switching routers (LSRs) are potentially capable of providing the highest forwarding capacity in the backbone network. Virtual circuit (VC) merging is a mechanism in an ATM-LSR that allows many IP routes to be mapped to the same VC label and provides a scalable mapping method that can support thousands of destinations. VC merging requires reassembly buffers so that cells belonging to different packets intended for the same destination do not interleave with each other. In this paper, the impact of VC merging on the buffering requirement for the reassembly buffers is investigated. We propose a realistic architecture that supports VC merging. We study the performance of this architecture using an analytic approach and using simulation driven by empirical Internet packet-size distribution. At the cell level, our main finding indicates that VC merging incurs a minimal overhead compared to non-VC merging, in terms of additional buffering. Moreover, the overhead decreases as utilization increases or as the traffic becomes more bursty with longer dependence. The finding has important practical consequences since routers and switches are dimensioned for high utilization and stressful traffic conditions. At the packet level, VC merging generally achieves a higher goodput than non-VC merging with EPD for the same buffer size. We also study the delay performance and find that the additional delay due to VC merging is insignificant at high speed.

VC merging allows many routes to be mapped to the same VC label, providing a scalable mapping method that can support tens of thousands of edge routers. VC merging requires reassembly buffers so that cells belonging to different packets intended for the same destination do not interleave with each other. The impact of VC merging on the additional buffer required for the reassembly buffers and other buffers due to the perturbation in the traffic process is investigated. We propose a realistic output-buffered ATM switch architecture that supports VC merging capability. We analyze the performance of the switch using a decomposition approach, and verify the results using simulation. We investigate the impact of VC merging on loss and delay performance for realistic traffic scenarios. The main result indicates that VC merging incurs a minimal overhead compared to non-VC merging in terms of additional buffering. Moreover, the overhead decreases as utilization increases, or as the traffic becomes more bursty. The finding has important implication since practical ATM switches are dimensioned for high utilization and stressful traffic conditions. We also study the delay performance and find that the additional delay due to VC merging is insignificant for most applications.

Analysis of the draft IEEE 802.11 wireless local area network (WLAN) standard is needed to characterize the expected performance of the standard's ad hoc and infrastructure networks. The performance of the medium access control (MAC) sublayer, which consists of distributed coordination function (DCF) and point coordination function (PCF), is determined by simulating asynchronous data traffic in a 1 Mbps ad hoc network, and asynchronous data and packetized voice traffic in a 1 Mbps infrastructure network. The simulation models incorporate the effect of burst errors, packet size, RTS threshold and fragmentation threshold on network throughput and delay. The results show that the IEEE 802.11 WLAN can achieve a reasonably high efficiency when the medium is almost error-free, but may degrade appreciably under harsh fading. The results also show that time-sensitive traffic such as packet voice can be supported together with other intensive traffic such as packet data. However, an echo canceller is required for packet voice systems.

Many systems in ATM environments require the queue to be monitored for impending congestion before it happens. We characterize impending congestion by measuring how long the queue length has exceeded a certain threshold. Based on the idea of the profile curve, we introduce a new performance measure called the conditional overflow probability which gives a quantitative measure on the congestion state. We develop a matrix analytic method to compute the conditional overflow probability and the profile curve, and present a corresponding algorithm to estimate the two measures using real traffic data.

In this paper, we investigate the performance of two burst-level admission control protocol for ATM networks that are based on simple random access techniques: fell-and-wait and tell-and-go. We develop an analytical model based on reduced load approximation to obtain throughput, end-to-end transfer delay and blocking probability characteristics far an arbitrary network topology with multirate traffic. We discuss the performance of these two protocols under varying propagation delays and peak bit rates. We discuss how the presence of multirate traffic introduces the issue of fairness into these burst-level admission control protocols, a phenomenon not observed previously.

The paper presents a new cell switching architecture for ATM-based networks. The proposed helical switch is a multistage interconnection network which implements the self-routing technique with efficient buffer sharing. Although the switch may route cells along multiple paths, the connection-oriented mode required by the ATM-based network is supported. Cell sequence integrity is guaranteed by introducing a virtual helix which forces cells routed along different paths to proceed in order and fill the internal buffers uniformly. The performance of the helical switch is investigated under uniform and nonuniform traffic patterns. Unlike single-path multistage networks such as buffered banyan networks which can degrade significantly under nonuniform traffic, the helical switch is shown to be quite robust with respect to nonuniform traffic conditions.>

The authors proposed a method for decoding interpolatively encoded data. This class of coding scheme achieves higher estimation gain and are symmetric with respect to time which makes them a good candidate for storage application. They showed that different trade-off parameters are involved and investigated their relationships. These parameters are estimation gain, delay experienced by the encoder and the decoder and end-to-end signal-to-noise ratio. They also showed the implementation and the effect of incorporating quantizers in the circuits. Specifically, they investigated two extreme open and closed loop architectures and compared their performances. Generalization of the above algorithm to noise feedback coding can be achieved easily.>

A novel cell switching architecture for asynchronous transfer mode (ATM)-based networks is presented. The proposed helical switch is a multistage interconnection network which implements the self-routing technique with efficient buffer sharing. Although the switch may route cells along multiple paths, the connection-oriented mode required by the ATM-based network is supported. Cell sequence integrity is guaranteed by introducing a virtual helix which forces cells routed along different paths to proceed in order and fill the internal buffers uniformly. Under a uniform traffic pattern, it is shown that the minimum achievable throughput is 5/6 per output line.>

A report is presented on a queueing analysis and a simulation study of a switch fabric based on a buffered banyan structure whereby buffers are placed at the output links of each switching element. When buffers are located at the input links, it is well known that maximum throughput is limited to approximately 0.45 under a uniform input traffic pattern. This bottleneck is due to the head of the line (HOL) contention at each switching element and is intrinsic to input queueing. The authors propose a buffered banyan switch built from smaller knockout switches which are output-buffered switches. With small knockout switches as the basic switching elements, the complexity of the overall switch fabric is manageable and no internal clock speedup is required. Furthermore, it is shown that with the proposed output-buffered banyan switch, a maximum throughput of 1 can be achieved.>