By Topic

Modeling, Analysis and Simulation of Computers and Telecommunication Systems, 2008. MASCOTS 2008. IEEE International Symposium on

Date 8-10 Sept. 2008

Filter Results

Displaying Results 1 - 25 of 47
  • [Breaker page]

    Page(s): 1
    Save to Project icon | Request Permissions | PDF file iconPDF (246 KB)  
    Freely Available from IEEE
  • [Breaker page]

    Page(s): 1
    Save to Project icon | Request Permissions | PDF file iconPDF (467 KB)  
    Freely Available from IEEE
  • Controlled Epidemic Routing for Multicasting in Delay Tolerant Networks

    Page(s): 1 - 10
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (2322 KB) |  | HTML iconHTML  

    Delay tolerant networks (DTNs) are a class of networks that experience frequent and long-duration partitions due to sparse distribution of nodes. DTN multicasting is a desirable feature for applications where some form of group communication is needed. In this paper, we examine multicasting in DTNs using controlled flooding schemes. Specifically, we analyze basic multicast routing schemes for fundamental performance metrics such as message delivery ratio, message delay, and buffer occupancy. Further, we study the effects of different controlled Epidemic routing schemes using TTL and message expiration times. Our experiments show that our analytical results are accurate and that with careful protocol parameter selection it is possible to achieve high delivery rates for various system scenarios. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Geographic Constraint Mobility Model for Ad Hoc Network

    Page(s): 1 - 10
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (5442 KB) |  | HTML iconHTML  

    In this paper, we propose a mobility model and present its simulation tool to generate realistic mobility traces for mobile ad hoc network. The mobility model is capable of creating realistic node movement pattern in the presence of geographic constraints by exploiting the concepts of anchors. The model dynamically places anchors depending upon the context of the environment through which nodes are guided to move towards the destination, and obstacles of arbitrary shapes with or without doorways and any existing pathways, in full or part of the terrain can be incorporated which makes the simulation environment more realistic. The characteristics of the proposed mobility model tested on a real world university campus map at various movement patterns are presented that illustrate the impact of the mobility model on the performance of a routing protocol and usefulness of the proposed scenario generation tool. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Increased Reliability with SSPiRAL Data Layouts

    Page(s): 1 - 10
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (6372 KB) |  | HTML iconHTML  

    We evaluate the reliability of storage system schemes consisting of an equal numbers of data disks and parity disks where each parity disk contains the exclusive or (XOR) of two or three of the data disks. These schemes are instances of Survivable Storage using Parity in Redundant Array Layouts (SSPiRAL). They have the same storage costs as mirrored organizations and use very simple parity schemes. Through a novel dynamic analysis of the likelihood of data losses, we show that these schemes are one hundred thousand to a million times less likely to lose data than a comparable mirrored organization. We also found that schemes where each parity disk contains the exclusive or of three data disks performed much better than schemes where each parity disk contains the exclusive or of only two data disks. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Bounding the Performance of BCMP Networks with Load-Dependent Stations

    Page(s): 1 - 8
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (641 KB) |  | HTML iconHTML  

    In this paper, we introduce new bounds on the system throughput and response time of closed, single-class BCMP queueing networks with load-dependent stations. Under the assumption that stations relative service rates are non-decreasing functions of their queue lengths, the bounds derive from the monotonicity of system throughput and queue-lengths and exploit the asymptotic equivalence that exists between closed and open single-class BCMP networks when the number of jobs N populating a closed network grows to infinity. The bounds can be applied when N is sufficiently large and the minimum N which allows their use is given. Experimental results present scenarios in which the proposed bounds significantly improve the accuracy of existing techniques and we analytically show that they are always more accurate than the popular balanced job bounds when N is greater than a given threshold. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Comparison of Single-Rate Multicast Congestion Control Protocols vs. ASMP

    Page(s): 1 - 8
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (5026 KB) |  | HTML iconHTML  

    We present in this paper a simulation-based comparison of two single-rate multicast congestion control schemes (TFMCC and PGMCC) against our proposed adaptive smooth multicast protocol (ASMP). ASMP consists of a single-rate multicast congestion control, which takes advantage of RTCP sender (SR) and receiver reports (RR). The innovation in ASMP lays in the "smooth" transmission rate, which is TCP-friendly and prevent oscillations. The smooth behavior is naturally well suited to multimedia applications as high oscillations of the sending rate may create distortions of audio-video (AV) encoders and decoders. Simulation results, which are conducted with the network simulator ns2 software, showed that ASMP can be regarded as a serious competitor of TFMCC and PGMCC. In many cases, ASMP outperforms TFMCC in terms of TCP-friendliness and smooth transmission rates, while PGMCC presents lower scalability than ASMP. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Runtime prediction models for Web-based system resources

    Page(s): 1 - 8
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (1571 KB) |  | HTML iconHTML  

    Several activities of Web-based architectures are managed by algorithms that take runtime decisions on the basis of continuous information about the state of the internal system resources. The problem is that in this extremely dynamic context the observed data points are characterized by high variability, dispersion and noise at different time scales to the extent that existing models cannot guarantee accurate predictions at runtime. In this paper, we evaluate the predictability of the internal resource state and point out the necessity to filter the noise of raw data measures. We then verify that more accurate prediction models are required which take into account the non stationary effects of the data sets, the time series trends and the runtime constraints. To these purposes, we propose a new prediction model, called trend-aware regression. It is specifically designed to deal with on the fly and short-term forecast of time series which originate from filtered data points belonging to internal resources of Web system. The experiment evaluation for different workload scenarios shows that the proposed trend-aware regression model improves the prediction accuracy with respect to popular algorithms based on auto-regressive and linear models, while satisfying the computational constraints of runtime prediction. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Modeling and Analyzing the Effect of Microarchitecture Design Parameters on Microprocessor Soft Error Vulnerability

    Page(s): 1 - 10
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (5585 KB) |  | HTML iconHTML  

    High performance and reliability are essential for microprocessor design. As semiconductor processing technology continues to move toward smaller and denser transistors, lower threshold voltages and tighter noise margins, microprocessors are becoming more susceptible to transient faults (soft errors) that can affect reliability. The increasing chip soft error rates make it is necessary to estimate process transient fault susceptibility at the microarchitecture design stage. Therefore, it becomes important to understand and to evaluate the implications of design choices and optimizations from both performance and reliability perspectives. This paper explores using predictive models to analyze and forecast the effect of various processor microarchitecture design parameters on reliability and their tradeoffs with performance. The most significant factors affecting microarchitecture structures and processor reliability and its runtime variation are obtained. Experimental results show that the proposed modeling techniques can accurately estimate processor reliability, runtime variation, and the performance/reliability tradeoffs in the early stages of microarchitecture design exploration. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Profiling, Prediction, and Capping of Power Consumption in Consolidated Environments

    Page(s): 1 - 10
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (6683 KB) |  | HTML iconHTML  

    Consolidation of workloads has emerged as a key mechanism to dampen the rapidly growing energy expenditure within enterprise-scale data centers. To gainfully utilize consolidation-based techniques, we must be able to characterize the power consumption of groups of co-located applications. Such characterization is crucial for effective prediction and enforcement of appropriate limits on power consumption-power budgets-within the data center. We identify two kinds of power budgets (i) an average budget to capture an upper bound on long-term energy consumption within that level and (ii) a sustained budget to capture any restrictions on sustained draw of current above a certain threshold. Using a simple measurement infrastructure, we derive power profiles-statistical descriptions of the power consumption of applications. Based on insights gained from detailed profiling of several applications both individual and consolidated-we develop models for predicting average and sustained power consumption of consolidated applications. We conduct an experimental evaluation of our techniques on a Xen-based server that consolidates applications drawn from a diverse pool. For a variety of consolidation scenarios, We are able to predict average power consumption within 5% error margin and sustained power within 10% error margin. Our sustained power prediction techniques allow us to predict close yet safe upper bounds on the sustained power consumption of consolidated applications. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • A Performance Model of Multi-Version Concurrency Control

    Page(s): 1 - 10
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (1821 KB) |  | HTML iconHTML  

    In this article we present a performance model for multi-version concurrency control (MVCC). This type of concurrency control is currently very popular among mainstream commercial and open source database systems thanks to its ability to well cope with read intensive workloads, as in the case of transaction profiles proper of Web applications. To build the model we had to tackle the intrinsic higher complexity of MVCC when compared to traditional concurrency control mechanisms (i.e. 2-phase-locking and optimistic ones), such as the joint use of locks and aborts to resolve direct conflicts among write accesses to the same data item, and the management of multiple data versions. We validate our analytical model via an extensive simulation study, considering both uniform and skewed data accesses, as well as differentiated transaction profiles. To the best of our knowledge, the present study provides the first analytical model of MVCC. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Failure Prediction Models for Proactive Fault Tolerance within Storage Systems

    Page(s): 1 - 8
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (2362 KB) |  | HTML iconHTML  

    The increasingly large demand for data storage has spurred on the development of systems that rely on the aggregate performance of multiple hard drives. In many of these applications, reliability and availability are of utmost importance. It is therefore necessary to closely scrutinize a complex storage system's reliability characteristics. In this paper, we use Markov models to rigorously demonstrate the effects that failure prediction has on a system's mean time to data loss (MTTDL) given a parameterized sensitivity. We devise models for a single hard drive, RAID1, and N+1 type RAID systems. We find that the normal SMART failure prediction system has little impact on the MTTDL, but striking results can be seen when the sensitivity of the predictor reaches 0.5 or more. In past research, machine learning techniques have been proposed to improve SMART, showing that sensitivity levels of 0.5 or more are possible by training on past SMART data alone. The results of our stochastic models show that even with such relatively modest predictive power, these failure prediction algorithms can drastically extend the MTTDL of a data storage system. We feel that these results underscore the importance and need for complex prediction systems when calculating impending hard drive failures. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Quantifying Temporal and Spatial Localities in Storage Workloads and Transformations by Data Path Components

    Page(s): 1 - 10
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (8219 KB) |  | HTML iconHTML  

    Temporal and spatial localities are basic concepts in operating systems, and storage systems rely on localities to perform well. Surprisingly, it is difficult to quantify the localities present in workloads and how localities are transformed by storage data path components in metrics that can be compared under diverse settings. In this paper, we introduce stack- and block-affinity metrics to quantify temporal and spatial localities. We demonstrate that our metrics (1) behave well under extreme and normal loads, (2) can be used to validate synthetic loads at each stage of storage optimization, (3) can capture localities in ways that are resilient to generations of hardware, and (4) correlate meaningfully with performance. Our experience also unveiled hidden semantics of localities and identified future research directions. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Network Information Flow in Network of Queues

    Page(s): 1 - 8
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (2460 KB) |  | HTML iconHTML  

    Two classic categories of models exist for computer networks: network information flow and network of queues. The network information flow model appropriately captures the multi-hop flow routing nature in general network topologies, as well as encodable and replicable properties of information flows. However, it assumes nodes in the network to be infinitely powerful and therefore does not accurately model queueing delay and loss at nodes. The network of queues model instead focuses on finite capacitied nodes and studies buffering and loss behaviors from a stochastic perspective. However, existing models on network of queues are mostly based on unrealistically simple topologies, and lacks the multi-hop flow routing dimension. In this work, we seek to combine advantages from both models. We start with the network information flow model and replace each infinitely powerful node with afinitely capacitied queue system instead. We show that the optimal routing problems for unicast, multiple unicasts and multicast can all be formulated as convex optimization problems. As a necessary step in validating the model for multicast routing, we show that network coding does not change the memoryless nature of traffic. We examine the correctness of the models through simulations and show that they behave differently than traditional link-cost based network flow models. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Optimal Layered Multicast with Network Coding: Mathematical Model and Empirical Studies

    Page(s): 1 - 10
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (5059 KB) |  | HTML iconHTML  

    Recent advances in network coding research dramatically changed the underlying structure of optimal multicast routing algorithms and made them efficiently computable. While most such algorithm design assume a single file/layer being multicast, layered coding introduces new challenges into the paradigm due to its cumulative decoding nature. Layered coding is designed to handle heterogeneity in receiver capacities, and a node may decode layer k only if it successfully receives all layers in 1..k. We show that recently proposed optimization models for layered multicast do not correctly address this challenge. We argue that in order to achieve the absolute maximum throughput (or minimum cost), it is necessary to decouple application layer throughput from network layer throughput. In particular, a node should be able to receive a non-consecutive layer or a partial layer even if it cannot decode and utilize it (e.g., for playback in media streaming applications). The rationale is that nodes at critical network locations need to receive data just for helping other peers. We present a mathematical programming model that addresses the above challenges and achieves the absolute optimal performance. Simulation results show considerable throughput gain (cost reduction) compared with previous models, in a broad range of network scenarios. We further generalize our model for studying the optimal progression of layer sizes. We show that such optimization is non-convex, and apply a Simulated Annealing algorithm to solve it, with flexible trade-off between solution quality and running time. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Optimizing Galois Field Arithmetic for Diverse Processor Architectures and Applications

    Page(s): 1 - 10
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (2573 KB) |  | HTML iconHTML  

    Galois field implementations are central to the design of many reliable and secure systems, with many systems implementing them in software. The two most common Galois field operations are addition and multiplication; typically, multiplication is far more expensive than addition. In software, multiplication is generally done with a look-up to a pre-computed table, limiting the size of the field and resulting in uneven performance across architectures and applications. In this paper, we first anaylze existing table-based implementation and optimization techniques for multiplication in fields of the form GF(21). Next, we propose the use of techniques in composite fields: extensions of GF(21) in which multiplications are performed in GF(21) and efficiently combined. The composite field technique trades computation for storage space, which prevents eviction of look-up tables from the CPU cache and allows for arbitrarily large fields. Most Galois field optimizations are specific to a particular implementation; our technique is general and may be applied in any scenario requiring Galois fields. A detailed performance study across five architectures shows that the relative performance of each approach varies with architecture, and that CPU, memory limitations and fields size must be considered when selecting an appropriate Galois field implementation. We also find that the use of our composite field implementation is often faster and less memory intensive than traditional algorithms for GF(21). View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • TCP over WiMAX: A Measurement Study

    Page(s): 1 - 10
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (6591 KB) |  | HTML iconHTML  

    We present active measurement results from a commercial IEEE 802.16/WiMAX-based network, with primary focus on TCP performance. We compare four TCP variants, namely New Reno, Cubic, Vegas and Veno, using throughput, round-trip time (RTT), and retransmission rate metrics. While all TCP variants achieve similar throughput, they do so in different ways, with different impacts on the network performance. We identify adverse effects of TCP window auto-tuning in this environment and demonstrate that on the downlink, congestion losses dominate wireless transmission errors. We reveal several issues for this WiMAX-based network, including limited bandwidth for TCP, high RTT and jitter, and unfairness during bidirectional transfers. Such a network environment may be challenging for many wireless Internet applications, such as remote login, VoIP, and video streaming. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Taxis: Scalable Strong Anonymous Communication

    Page(s): 1 - 10
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (979 KB) |  | HTML iconHTML  

    Anonymity protocols are a privacy-enhancing technology for Internet-based communication. Two important characteristics are the strength of anonymity provided, and the overhead required for anonymous communication. In this paper, we focus on the latter characteristic, and develop simple performance models for two anonymous communication protocols. Practical Buses and Taxis. We show that the message latency of Practical Buses scales quadratically with the number of participants, while that of Taxis scales linearly with the number of participants. Both models are validated with experimental measurements from prototype implementations. We show that Taxis provides more scalable anonymous communication, without compromising the strength of anonymity provided. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • An Analysis of Hard Drive Energy Consumption

    Page(s): 1 - 10
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (5875 KB) |  | HTML iconHTML  

    The increasing storage capacity and necessary redundancy of data centers and other large-scale IT facilities has drawn attention to the issue of reducing the power consumption of hard drives. This work comprehensively investigates the power consumption of hard drives to determine typical runtime power profiles. We have instrumented at a fine-grained level and present our findings which show that (i) the energy consumed by the electronics of a drive is just as important as the mechanical energy consumption; (ii) the energy required to access data is affected by physical location on a drive; and (iii) the size of data transfers has measurable effect on power consumption. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • A Markov Chain Model to Account for Multi-Rate Transmission and Node Cooperative Behavior in IEEE 802.11 Data Link Protocol

    Page(s): 1 - 10
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (5460 KB) |  | HTML iconHTML  

    This paper presents an analytical model which extends Markov Chain models previously proposed for IEEE 802.11 data link protocol. Unlike already published models, the proposed Markov Chain accounts jointly for the following three factors. First, in addition to collision, it accounts for frames that may be lost due to a low signal-to-noise ratio value at the link receiver (frame corruption). Second, it accounts for the various link transmission rate options which affect both the frame transmission time and probability of frame corruption. Third, it accounts for co-operative protocol variants of the IEEE 802.11 protocol. According to these protocols, a third node (relay) - besides the sender and receiver - may help improve the link capacity. In such a case, the active participation of the relay node must be properly captured by the analytical model. All the above factors are combined into one single Markov Chain model by leveraging a multi-dimensional approach. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Operational Analysis of Parallel Servers

    Page(s): 1 - 10
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (1925 KB) |  | HTML iconHTML  

    Multicore processors promise continued hardware performance improvements even as single-core performance flattens out. However they also enable increasingly complex application software that threatens to obfuscate application-level performance. This paper applies operational analysis to the problem of understanding and predicting application-level performance in parallel servers. We present operational laws that offer both insight and actionable information based on lightweight passive external observations of black-box applications. One law accurately infers queuing delays; others predict the performance implications of expanding or reducing capacity. The former enables improved monitoring and system management; the latter enable capacity planning and dynamic resource provisioning to incorporate application-level performance in a principled way. Our laws rest upon a handful of weak assumptions that are easy to test and widely satisfied in practice. We show that the laws are broadly applicable across many practical CPU scheduling policies. Experimental results on a multicore network server in an enterprise data center demonstrate the usefulness of our laws. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Tackling the Memory Balancing Problem for Large-Scale Network Simulation

    Page(s): 1 - 10
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (868 KB) |  | HTML iconHTML  

    A key obstacle to large-scale network simulation over PC clusters is the memory balancing problem where a memory-overloaded machine can slow down an entire simulation due to disk I/O overhead. Memory balancing is complicated by (i) the dfficulty of estimating the peak memory consumption of a group of nodes during network partitioning-a consequence of per-node peak memory not being synchronized-and (ii) trade-off with CPU balancing whose cost metric depends on total-as opposed to maximum-number of messages processed over time. We investigate memory balancing for large-scale network simulation which admits solutions for memory estimation and balancing not availed to small-scale or discrete-event simulation in general. First, we advance a measurement methodology for accurate and efficient memory estimation, and we establish a trade-off between memory and CPU balancing under maximum and total cost metrics. Second, we show that joint memory-CPU balancing can overcome the performance trade-off-in general not feasible due to constraint conflicts-which stems from network simulation having a tendency to induce correlation between maximum and total cost metrics. Performance evaluation is carnied out using benchmark applications with varying traffic characteristics-BGP routing, worm propagation under local and global scanning, and distributed client/server system-on a testbed of 32 Intel times86 machines running a measurement-enhanced DaSSF. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • On Impact of Dynamic Virtual Machine Reallocation on Data Center Efficiency

    Page(s): 1 - 8
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (1790 KB) |  | HTML iconHTML  

    Modern OS virtualization technology allows for "live migration" of virtual servers between physical hosts after the initial consolidation thus providing new avenues for performance optimization. This paper focuses on quantifying by how much data center efficiency can be improved using dynamic (i.e., in response to demand changes) real-location of virtual machines between physical servers. An analytical performance model of dynamic virtual machine reallocation is presented. It allows to estimate the reduction (as compared to the static consolidation) in the number of physical servers required to host the workload (under the same requirements of the maximum number of capacity overloads). Model is validated using simulations and several insights based on the analysis are provided. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Modelling and Validation of Response Times in Zoned RAID

    Page(s): 1 - 10
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (1776 KB) |  | HTML iconHTML  

    We present and validate an enhanced analytical queueing network model of zoned RAID. The model focuses on RAID levels 01 and 5, and yields the distribution of I/O request response time. Whereas our previous work could only support arrival streams of I/O requests of the same type, the model presented here supports heterogeneous streams with a mixture of read and write requests. This improved realism is made possible through multiclass extentions to our existing model. When combined with priority queueing, this development also enables more accurate modelling of the way subtasks of RAID 5 write requests are scheduled. In all cases we derive analytical results for calculating not only the mean but also higher moments and the full distribution of I/O request response time. We validate our model against measurements from a real RAID system. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Characterization of Large-Scale SMTP Traffic: the Coexistence of the Poisson Process and Self-Similarity

    Page(s): 1 - 10
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (5232 KB) |  | HTML iconHTML  

    Network traffic such as Ethernet, Internet, World Wide Web, and TCP/UDP protocols has been extensively studied, with efforts focusing on the Poisson process and self-similarity. However, although SMTP (Simple Mail Transfer Protocol) occupies a significant portion of Internet traffic, it has attracted little attention. This paper shows that large-scale SMTP traffic possesses both the characteristics of the Poisson process and self-similarity through an analysis of high quality SMTP traces collected from one of the largest Web portal sites in South Korea over a period of almost nine months. First, we show that, at small (several-second) time scales, interarrival times of SMTP session arrivals are exponentially distributed and independent of each other, which makes it possible to model SMTP session arrivals as a Poisson process. On the contrary, at large (several-month) time scales, SMTP session arrivals exhibit self-similarity. They are strongly autocorrelated across time scales of days and their Hurst parameters are estimated to be somewhere between 0.85 to 0.97. In addition, we find that SMTP traffic consists of many individual ON/OFF sources whose distributions of OFF-period lengths are heavy-tailed, which confirms self-similarity of SMTP traffic. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.