This innovative practice WIP paper describes the “Pledge of the Computing Professional” (the Pledge), its need, its relationship to two major codes of ethics, and its role in inspiring computing students to act ethically and professionally. Professional ethics are the heart and soul of any profession. In developing products and services, professionals are to conduct themselves ethically, focusing on the public good. ACM has developed the Code of Ethics and Professional Conduct, and ACM and IEEE Computer Society have together published the Software Engineering Code of Ethics and Professional Practice. ACM considers ethics a core topic in curricular recommendations. In 2011, a team of computing professionals developed the Pledge of the Computing Professional. The purpose of the Pledge is to acknowledge the impact of computing and the attendant responsibilities and to nurture a desire to promote the public good. The Pledge has now been adopted by over 75 institutions in the US. This paper clarifies the relationship between the Pledge and the codes of ethics. It is hoped that emphasizing this connection and promoting a culture surrounding the Pledge can be instrumental in producing graduates with a well-rounded sense of ethics and professionalism, who become responsible citizens that can shape the profession's future. The Pledge may be viewed as both an introduction for the students to the profound impact and responsibilities they would have in society and a means to inspire them to reflect on its meaning as computing ethics become the center of public discourse. Future work includes making the Pledge international in scope.

Despite increasing demands for skilled workers within the technological domain, there is still a deficit in the number of graduates in computing fields (computer science, information technology, and computer engineering). Understanding the factors that contribute to students’ motivation and persistence is critical to helping educators, administrators, and industry professionals better focus efforts to improve academic outcomes and job placement. This article examines how experiences contribute to a student’s computing identity, which we define by their interest, recognition, sense of belonging, and competence/performance beliefs. In particular, we consider groups underrepresented in these disciplines, women and minoritized racial/ethnic groups (Black/African American and Hispanic/Latinx). To delve into these relationships, a survey of more than 1,600 students in computing fields was conducted at three metropolitan public universities in Florida. Regression was used to elucidate which experiences predict computing identity and how social identification (i.e., as female, Black/African American, and/or Hispanic/Latinx) may interact with these experiences. Our results suggest that several types of experiences positively predict a student’s computing identity, such as mentoring others, having a job, or having friends in computing. Moreover, certain experiences have a different effect on computing identity for female and Hispanic/Latinx students. More specifically, receiving academic advice from teaching assistants was more positive for female students, receiving advice from industry professionals was more negative for Hispanic/Latinx students, and receiving help on classwork from students in their class was more positive for Hispanic/Latinx students. Other experiences, while having the same effect on computing identity across students, were experienced at significantly different rates by females, Black/African American students, and Hispanic/Latinx students. The findings highlight experiential ways in which computing programs can foster computing identity development, particularly for underrepresented and marginalized groups in computing.

Increasingly companies assess a computing candidate's capabilities using technical interviews (TIs). Yet students struggle to code on demand, and there is already an insufficient amount of computing graduates to meet industry needs. Therefore, it is important to understand students' perceptions of TIs, and other professional experiences (e.g., computing jobs). We surveyed 740 undergraduate computing students at three universities to examine their experiences with the hiring process, as well as the impact of professional and cultural experiences (e.g., familial support) on computing identity. We considered the interactions between these experiences and social identity for groups underrepresented in computing - women, Black/African American, and Hispanic/Latinx students. Among other findings, we observed that students that did not have positive experiences with TIs had a reduced computing identity, but that facing discrimination during technical interviews had the opposite effect. Social support may play a role. Having friends in computing bolsters computing identity for Hispanic/Latinx students, as does a supportive home environment for women. Also, freelance computing jobs increase computing identity for Black/African American students. Our findings are intended to raise awareness of the best way for educators to help diverse groups of students to succeed, and to inform them of the experiences that may influence students' engagement, resilience, and computing identity development.

The lack of diversity in computing fields in the United States is a known issue. Students enter the computing fields with the intention of graduating; however, a large number leave and do not persist after enrolling, due to discrimination and biases. This particularly concerns groups already underrepresented in computing fields, such as women, Black/African American students, and Hispanic/Latinx students. However, there are various experiences that can make students feel more included or excluded in the field. Some of these experiences include internships, undergraduate research, capstone courses, and projects, etc. Drawing on Astin's I-E-O model and applying a random forest algorithm, we measure the feature importance of 14 distinct experiences on 1650 students' feelings of inclusivity in the computing field. We observe that there are gender and racial differences in terms of the opinions of computing fields' inclusivity. For example, tutoring experience, job offers, and job experience are considered some of the most important factors for female's perceived inclusiveness of women. However, men perceived women's inclusivity differently, based on the experiences they engaged in. We also looked at the perceived inclusiveness of computing fields for ethnically and racially underrepresented groups, such as Hispanic/Latinx students. Understanding the effect of different experiences on students of both genders with different races and ethnicities on the perceived inclusion could assist the computing community to provide more cohesive experiences that benefits all students and helps them to feel more welcome.

In this Work in Progress Innovative Practice paper, we describe a process for finding predictors for student success - and failure - for Computer Science and Computer Engineering students with a focus on the second programming course (CS2). We use readily available off-the-shelf statistical and data mining tools for generating summary statistics, calculating correlations, testing statistical significance, and creating decision trees. We analyze grade data from the first programming course (CS1), entry-level STEM courses (Calculus and Physics), and an English course to determine success predictors for CS2. Not surprisingly, the grade in CS1 is the best predictor for success in CS2. We also find that success in CS2 is independent of gender. Looking deeper into the data, we find characteristics of students who are very likely to pass or fail CS2. Being able to identify predictors for success is useful for calibrating admission criteria and designing appropriate interventions (e.g., requiring prereq classes, recitation sessions, and so on) to improve success probability for all students. A key contribution of this paper is a step-by-step process that can be used by other programs to find success predictors and design appropriate interventions.

This Research Full Paper presents the effects of computing identity sub-constructs on the persistence of computer science students. Computer science (CS) is one of the fastest growing disciplines in the world and an emerging critical field for all students to obtain vital skills to be successful in the 21st century. Despite the growing importance of computer science, many university and college programs suffer from low student persistence rates. Disciplinary identity is a theoretical framework that refers to how students see themselves with respect to a discipline and is related to long-term membership in a disciplinary community. The theory has been effectively applied in Science, Technology, Engineering, and Mathematics (STEM) to understand students' success and persistence. This study examines the effects of performance/competence, recognition, interest and sense of belonging on the academic persistence of computer science students. A survey of approximately 1,640 computing students as part of a National Science Foundation (NSF) funded project was developed and administered at three metropolitan public institutions. Confirmatory Factor Analysis (CFA) was performed to validate the sub-constructs of identity for use in a computing identity model. Then, a structural equation model (SEM) was constructed as a snapshot of the structural relationships for describing and quantifying the impact of the identity sub-constructs on persistence. The results indicated that our model for CS aligns with prior research on disciplinary identity but also adds the importance of sense of belonging. In addition, the findings indicate that students' academic persistence is directly influenced by their interest. A better understanding of these factors may leverage insight into students' academic persistence in computer science/engineering programs as well as a meaningful lens of analysis for further curriculum and extracurricular activities.

Power consumption and throughput as a function of utilization are key measures for understanding the performance-energy trade-offs of a server. Existing energy efficiency metrics use these measures in various ways. In this paper, we demonstrate the importance of taking workload characteristics into account such that the best server can be selected on a case by case basis. The best server will meet the performance demands of the workload with the lowest possible energy consumption. We rigorously evaluate existing metrics, with and without workload weighting, using publicly available SPECpower benchmark data for actual servers and a publicly available Google cluster usage data. We show that using unweighted metrics to select a server can result in selecting the wrong server ensuing in higher operational energy use, the difference being as high as 20% additional energy use. Weighting power consumption of a server by the workload distribution results in a metric that is linear and reliable in ranking of servers. An analysis of server rankings for two different existing metrics shows that our weighted metric is statistically significantly better than existing metrics. The results in this paper can help data center operators to quantify the importance of workload weighting to help them select the best server to reduce operational energy costs.

A Bloom filter is a data structure that provides probabilistic membership checking. Bloom filters have many applications in computing and communications systems. The performance of a Bloom filter is measured by false positive rate, memory size requirement, and query (or memory look-up) overhead. A recent paper by Qiao et al. proposes the Fast Bloom Filter, also called Bloom-1, which requires only a single memory look-up for a membership test. Bloom-1 achieves a reduced query overhead at the expense of a slightly higher false positive rate for a given memory size. The false positive rate of Bloom-1 has been analyzed theoretically by Qiao et al. relying on a well-known, but flawed, approximation for the false positive rate for a Bloom filter. In this comment paper we show that the Qiao et al. analysis of Bloom-1 under-estimates the false positive rate for low loads. We provide a correct analysis of Bloom-1 yielding an expression for the exact false positive rate.

Local renewable DC electricity sources, such as solar, have become cost competitive. In this position paper, we argue that an Ethernet LAN with Power-over-Ethernet (PoE) is a nanogrid where the physical layer infrastructure is used for DC electricity distribution and the link-layer capabilities can be used to match electricity supply and demand. To be cost effective with respect to wiring, many of these nanogrids must be multi-drop where each drop may be a power socket. Open challenges to employing PoE as a multi-drop managed nanogrid include 1) extending LLDP for multi-drop, 2) adapting LLDP to support the use of price signals for modulating power demand, 3) achieving efficient PoE power flow in a multi-drop circuit, and 4) implementing low-cost and fair scheduling for packet transmission. We explore how multi-drop Ethernet can be implemented with fair access to all devices by employing at each drop a mini-switch with packet scheduling.

The Pledge of the Computing Professional (http://computing-professional.org), endorsed by SIGCAS and the ACM Committee on Professional Ethics, was developed to both support professional responsibility in computing and to raise awareness of one's ethical responsibilities. To date, 25 institutions across the US have conducted Pledge Ceremonies to welcome their graduates into the computing profession. In a recent Inroads article, Don Gotterbarn asserts that, if one's ethics is influenced by those around us, then "just being aware of the existence of this Pledge helps to refocus us all on our own responsibility and the profession's responsibility." So in what ways can this Pledge be used both inside and outside of the classroom to better promote such traits? This session is for those interested in computing ethics and professional responsibility to get together and collaborate on new initiatives featuring The Pledge of the Computing Professional as a means to this end.

Network protocols can be designed to enable a reduction in energy use of data servers. We architect a new HTTP timed redirection response for a GET request to be redirected to another server with a given delay. This redirection response can be used in a hybrid web server to coalesce GET requests and allow a server to periodically sleep. In such a hybrid web server, a small low-power ARM-based Assistant receives all incoming HTTP requests and redirects them to a Pentium-class Master server with a delay calculated to allow the Master to sleep in alternating intervals. There is a growing class of applications that are delay tolerant at short time scales for file downloads. Experimental evaluation of timed redirection shows significant savings for the Master server with added delay for some requests which may be acceptable for certain applications.

The number of applications (or apps) in the Android Market exceeded 450,000 in 2012 with more than 11 billion total downloads. The necessity to fix bugs and add new features leads to frequent app updates. For each update, a full new version of the app is downloaded to the user's smart phone; this generates significant traffic in the network. We propose to use delta encoding algorithms and to download only the difference between two versions of an app. We implement delta encoding for Android using the bsdiff and bspatch tools and evaluate its performance. We show that app update traffic can be reduced by about 50%, this can lead to significant cost and energy savings.

Assessment of engineering and computer science programs involves a significant effort in data collection, analysis, and formatting of results for presentation to an ABET evaluation team. At the University of South Florida, we have developed a web-based system to unify program-level data collection, evaluation, and presentation for all programs in the College of Engineering. The goal of our web-based Accreditation Portal is to both improve the efficacy of assessment activities in the College and provide a clearer view of continuous improvement results to constituents and ABET evaluation teams. For each defined program educational objective and student outcome there are links to pages for Performance Indicators, Methods of Assessment, Assessment Data, Evaluation, Changes, and Closing the Loop. For each of these assessment steps Word, Excel, PDF, text, and/or image files can be uploaded. The portal also includes survey tools to enable data collection. We describe the development and deployment of the portal. We show a full sample assessment loop for a mock objective. Our evaluation of the portal is based on a trade study with comparison to several existing web-based assessment systems. Faculty members in the College are using the portal.

Web servers in small and medium enterprises (SME) consume a significant amount of energy. We consider how a hybrid SME web server based on two co-located platforms (one high performance and high power and the second low performance and low power) can be architected to appear as a single system image to clients. A prototype based on a Dell AMD Athlon x64-based PC and a Sheeva ARM-based plug computer is described and evaluated using the Apache ab benchmark. Using traffic traces from the KETI corporate web server, we show that for the majority of the time a low-performance platform can very likely meet the performance demands (during which time the high-power platform can sleep). We explore prediction of server load using network traffic analysis as a means to achieve a selection (or switching) policy between the two platforms. Using simple prediction methods, we show that we can achieve between 41% to 67% energy savings with minimal performance impact.

One of the most urgent challenges of the 21st century is to investigate new technologies that can enable a transition towards a more sustainable society with reduced energy use and a smaller CO2 footprint. ICT is a direct contributor to more than 2% of global CO2 emissions, and this is expected to double in the near future. It has been shown that PCs (which consume about 2% of the electricity in the US and a growing amount in developing countries) are idle most of the time and the majority of them do not have power management enabled. In this talk, I will describe ongoing standards work by IEEE 802.3az to address an energy efficient ethernet and by Ecma to address reduction of induced energy use of network-connected hosts, such as PCs. I will also describe possible future directions for further reducing energy consumption of both ICT and non-ICT equipment, systems, and processes focusing on the role of networks. Networks can play a significant role in reducing global energy consumption and achieving a level of comfortable conservation needed for a sustainable society.

Billions of dollars of electricity are being used to keep idle or unused network hosts fully powered-on only to maintain their network presence. We investigate how a network connectivity proxy (NCP) could enable significant energy savings by allowing idle hosts to enter a low-power sleep state and still maintain full network presence. An NCP must handle ARP, ICMP, DHCP, and other low-level network presence tasks for a network host. An NCP must also be able to maintain TCP connections and UDP data flows and to respond to application messages. The focus of this paper is on how TCP connections can be kept alive during periods of host sleep by using a SOCKS-based approach called green SOCKS (gSOCKS) as part of an NCP. The gSOCKS includes awareness of the power state of a host. A prototype implementation of gSOCKS in a Linksys router shows that TCP connections can be preserved.

Energy consumption has become a major factor in the total cost of ownership (TCO) of IT equipment. The power state of IT equipment is effectively ldquoinvisiblerdquo to the network making it difficult to measure or control energy use. In this short paper we propose, prototype, and evaluate a new SNMP power state MIB and its agent to expose equipment power state to the network. The power state includes all supported power management (PM) capabilities, current PM settings, total and current active, inactive, and sleep times, and statistics on wakeup and sleep events. With knowledge of the power state of network devices, a network manager could remotely audit the energy consumption of IT equipment and make changes to PM settings. We have implemented a subset of the power state MIB for Microsoft Windows Vista desktop and server computers.

Bloom filters are a well known data structure for approximate set membership. Bloom filters are space efficient but require many independent hashes and consecutive memory accesses for an element test. In this paper, we develop a hash table data structure that stores string signatures in an array. This new signature array hash table (SAHT) supports faster element testing than a bloom filter and requires less memory than a standard hash table that uses linked-list chains. The SAHT also supports removal of elements (which a Bloom filter does not) and addition of elements at the expense of requiring about 1.5x more memory than a bloom filter with same false positive rate.

The rapidly increasing energy consumption by computing and communications equipment is a significant economic and environmental problem that needs to be addressed. Ethernet network interface controllers (NICs) in the US alone consume hundreds of millions of US dollars in electricity per year. Most Ethernet links are underutilized and link energy consumption can be reduced by operating at a lower data rate. In this paper, we investigate adaptive link rate (ALR) as a means of reducing the energy consumption of a typical Ethernet link by adaptively varying the link data rate in response to utilization. Policies to determine when to change the link data rate are studied. Simple policies that use output buffer queue length thresholds and fine-grain utilization monitoring are shown to be effective. A Markov model of a state-dependent service rate queue with rate transitions only at service completion is used to evaluate the performance of ALR with respect to the mean packet delay, the time spent in an energy-saving low link data rate, and the oscillation of link data rates. Simulation experiments using actual and synthetic traffic traces show that an Ethernet link with ALR can operate at a lower data rate for over 80 percent of the time, yielding significant energy savings with only a very small increase in packet delay.

Bloom filters are a probabilistic data structure used to evaluate set membership. A group of hash functions are used to map elements into a bloom filter and to test elements for membership. In this paper, we propose using multiple groups of hash functions and selecting the group that generates the bloom filter instance with the smallest number of bits set to I. We evaluate the performance of this new Best-of-N method using order statistics and an actual implementation. Our analysis shows that significant reduction in the probability of a false positive can be achieved. We also propose and evaluate a new method that uses a random number generator (RNG) to generate multiple hashes from one initial "seed" hash. This RNG method (motivated by a method from Kirsch and Mitzenmacher) makes the computational expense of the Best-of-N method very modest. The target application is a power management proxy for P2P applications executing in a resource-constrained "SmartNIC".

The IEEE 802.3 energy efficient Ethernet (EEE) study group is considering rapid PHY Selection (RPS) as a mechanism to quickly switch the data rate of an Ethernet link to match link data rate with link utilization. When switching the data rate, RPS causes a momentary disruption of the link. This disruption may cause packet loss due to buffer overflow in upstream switches. We emulate RPS using PAUSE flow control and experimentally study the possible effects of RPS on TCP and UDP file transfer. We show that RPS has little or no perceivable effect on performance, but has some subtle effects on TCP throughput if PAUSE flow control is enabled in the file server.

In order to be part of a peer-to-peer (P2P) file sharing network a host must be fully powered-on all of the time. In addition to providing a user interface, a P2P host handles query messages and serves requested files. In this paper, we describe the development of a prototype Gnutella-like P2P power management proxy sub-system that handles query messages. This can allow desktop PCs acting as P2P hosts to enter a low-power sleep state for most of the time and be woken-up by the proxy only when needed to serve files. TCP connections with neighbors are maintained by the host when it is awake and by the proxy when the host is sleeping. Experiments show that a low-cost Freescale ColdFire processor can effectively proxy for a P2P host. This suggests that a controller for a Gnutella P2P proxy could be co-located on an Ethernet NIC at low cost. This could lead to significant energy savings by allowing P2P hosts to power manage into a low-power sleep state when not in active use.

Rapidly increasing energy use by computing and communications equipment is a significant problem that needs to be addressed. Ethernet network interface controllers (NICs) consume hundreds of millions of US$ in electricity per year. Most Ethernet links are underutilized and link power consumption can be reduced by operating at lower data rates. An output buffer threshold policy to change link data rate in response to utilization is investigated. Analytical and simulation models are developed to evaluate the performance of Adaptive Link Rate (ALR) with respect to mean packet delay and time spent in low data rate with Poisson traffic and 100 Mb/s network traces as inputs. A Markov model of a state-dependent service rate queue with rate transitions only at service completion is developed. For the traffic traces, it is found that a link can operate at 10 Mb/s for over 99% of the time yielding energy savings with no user-perceivable increase in packet delay.

The Internet and the devices that connect to it consume a growing and significant amount of electricity. The utilization of desktop-to-switch Ethernet links is generally very low and thus there is a potential for energy savings by using an adaptive link rate (ALR) protocol that matches link rate to utilization. In this paper, we design and evaluate a new ALR policy suitable for both bursty and smooth traffic. The policy uses output buffer thresholds and fine-grain utilization monitoring to determine when to switch link data rate. We develop a new traffic model for generating synthetic 1 and 10 Gb/s bursty traffic traces. Using this traffic model and simulation, we show that the new ALR policy is suitable for smooth traffic and also does not degrade performance for bursty traffic. Performance is measured in packet delay versus time in low (and energy saving) data rate

Service-learning enables computer science and engineering students to apply and refine their technical skills, learn new soft skills, and make a contribution to their community. At the University of South Florida we have organized a group of computer science and engineering students, called the CSE Volunteers, to provide IT support to the Hillsborough County School District. In spring 2005 over 7% of our undergraduate students voluntarily served in this program. While the work is fairly low-level, the experience for the students and school district has been very positive. Future directions include quantitative assessment, tackling projects of greater complexity, seeking corporate funding, and applying this program to help recruit new students to the major.

In many applications-including peer-to-peer (P2P) file sharing, content distribution networks, and grid computing-a single object will be searched for in multiple servers. In this paper, we find the provably optimal search method for such applications and develop analytical models for search time and cost. A client node searching for objects maintains statistics on where (in which servers) it has previously found objects. Using these statistics to target future searches to popular servers is provably optimal. For object location and request distributions that are non-uniform, which has been shown to be the case in P2P file sharing networks, this method of targeted searching is found to be more cost effective (i.e., use less server resources) than broadcast-based searching. Our targeted search method is implemented in a prototype Gnutella servent called Ditella. Ditella can improve the scalability of file sharing in P2P networks and reduce the amount of traffic in the Internet by reducing file search query traffic.

Most input buffered packet switches internally segment variable-length packets into fixed-length cells. The last cell in a segmented packet contains overhead bytes if the packet length is not evenly divisible by the cell length. Switch speed-up is used to compensate for this overhead. In this paper, we develop an analytical model of a single-server queue where an input stream of packets is segmented into cells for service. Analytical models are developed for M/M/1, M/H/sub 2//1, and M/E/sub 2//1 queues with a discretized (or quantized) service time. These models and simulation using real packet traces are used to evaluate the effect of speed-up on mean queue length. We propose and evaluate a new method of segmenting a packet trailer and subsequent packet header into a single cell. This cell merging method reduces the required speed-up. No changes to switch-matrix scheduling algorithms are needed. Simulation with a packet trace shows a reduction in the needed speed-up for an iSLIP scheduled input buffered switch.

Building large-scale video surveillance systems is of importance to national security. To support economical installation of video cameras, there is a need for new shared-medium protocols. This paper describes the new spatial reuse FireWire protocol (SFP). SFP is a bus arbitration protocol for an acyclic daisy-chained network topology. SFP is an extension of IEEE 1394b FireWire. SFP preserves the simple repeat path functionality of FireWire while offering two significant advantages: (1) SFP supports spatial reuse of bandwidth in order to increase effective throughput; and (2) SFP provides support for priority traffic to be able to support real-time applications (e.g., video) and data traffic. Simulation results show that for a uniform traffic pattern, SFP improves upon the throughput of IEEE 1394b by a factor of 1.7. For a traffic pattern typical of video surveillance applications, throughput increases by a factor of 6.8.

The combined input and crossbar queued (CICQ) switch is an input buffered switch suitable for very high-speed networks. The implementation feasibility of the CICQ switch architecture for 24 ports and 10-Gbps link speed is shown in this paper with an FPGA-based design (estimated cost of $30,000 in mid-2002). The bottleneck of a CICQ switch with RR scheduling is the RR poller. We develop a priority encoder based RR poller that uses feedback masking. This design has lower delay than any known design for an FPGA implementation.

Input buffered switches with virtual output queueing (VOQ) can be unstable when presented with unbalanced loads. Existing scheduling algorithms, including iSLIP for input queued (IQ) switches and round robin (RR) for combined input and crossbar queued (CICQ) switches, exhibit instability for some schedulable loads. We investigate the use of a queue length threshold and bursting mechanism to achieve stability without requiring internal speed-up. An analytical model is developed to prove that the burst stabilization protocol achieves stability and to predict the minimum burst value needed as a function of offered load. The analytical model is shown to have very good agreement with simulation results. These results show the advantage of the RR/RR CICQ switch as a contender for the next generation of high-speed switches.

Round robin (RR) arbitration is commonly used for scheduling of cells in high-speed packet switches. In this paper, we present an overlapped RR (ORR) arbiter design that fully overlaps RR polling and cell scheduling. The ORR arbiter achieves 100% throughput even when a cell transfer time is less than a worst case polling, or scheduling, cycle. This is done by scheduling blocks of celts during a cell transfer time.

Parallel packet switches (PPS) use internal, parallel switch planes that operate at less than line speed. A PPS can scale-up to faster line speeds than a single-plane switch can. Load balancing between planes and providing QoS to flows are open problems. We investigate the performance of a new "VIQ PPS" that distributes cells or variable-length packets to planes and uses multiplexors with virtual input queues (VIQ). A VIQ at an output multiplexor consists of one FIFO queue for each input. Simulation is used to evaluate the performance of a 10-Gbps VIQ PPS that contains ten 1-Gbps switch planes. It is found that at high offered loads the mean delay of a VIQ PPS switch is lower than that of a single-plane iSLIP switch. For unbalanced loads, the VIQ PPS demonstrates stability where an iSLIP switch is unstable. Especially promising results are shown for VIQ PPS native switching of variable-length Ethernet packets.

High-bandwidth sensor applications such as video surveillance give rise to the need for a wired sensor network (WSN). We show IEEE 1394b FireWire to be a suitable candidate for a shared-medium WSN. We also propose a hybrid location-centric routing protocol for future WSN with store and forward nodes.

We investigate methods of using CRC32 for compressing Web URL strings and sharing of URL lists between servers, caches, and URL switches. Using trace-based evaluation, we compare our new CRC32 digesting method against the existing Bloom filter and incremental CRC19 methods. Our CRC32 method requires less CPU resources, generates equal or smaller size digests, achieves equal collision rates, and simplifies switching.

Input-buffered switches with virtual output queueing require crossbar switch matrix scheduling algorithms. Existing scheduling algorithms are non-deterministic and are based on parallel and iterative request-grant-accept arbitration schemes. This presents challenges to flow-level scheduling for guaranteed throughput and bounded delay services and also to scalability. In this paper, the parallel-polled virtual output queued (PP-VOQ) switch is presented. Using parallel token passing, the PP-VOQ switch has deterministic and bounded scheduling delay and is implementable for 16 or 32 ports of 10-Gigabit Ethernet. The PP-VOQ switch is shown, via simulation, to perform very similar to an iSLIP switch. The PP-VOQ switch is extended to a "cube switch" design that decouples input port VOQ selection from output port selection to reduce the scheduling delay and improve scalability. This decoupled parallel polling is very scalable and is shown to result in better performance than an iSLIP switch.

We study the characteristics of Internet1 and Internet2 traffic at a network access point at a major USA university. With the use of fluid-flow modeling, we show that Internet1 and Internet2 traffic have different queuing behavior and that a small percentage of traffic on both networks largely contributes to this overall queuing behavior. We also demonstrate that buffer sizing, as a method to reduce loss, is largely ineffective for Internet2 traffic. These finding have implications for quality of service of Internet applications.

Clusters of PCs are increasingly being used to build high-capacity Web sites. We develop and evaluate a method of building a single system image (SSI) cluster that does not require a central dispatcher. In our clone cluster method, which builds on the existing ONE-IP method, a group of mirrored server machines with identical MAC and IP addresses are configured on a shared-medium LAN. Each machine filters incoming IP packets based on a hash of IP source addresses. Unlike ONE-IP, our method does not require router configuration changes. We build and evaluate a Linux-based, two-PC cluster that uses an odd/even filter. Simulation results show that the effect of server selection method on mean response time is less significant for heavy-tailed interarrival times than for exponential interarrival times. Simulation results show that response time performance from a Clone Cluster can be very close to that of a centrally dispatched least-loaded cluster.

We investigate how to add rate control to full-duplex IEEE 802.3 Ethernet. We envision the need for rate control in order to enable bandwidth allocation in future Ethernet in the first mile (EFM) subscriber services. We propose a RATE mechanism that can be implemented with a simple byte-based leaky bucket. We demonstrate with simulation methods that RATE can isolate flows and provide lower loss and delay than no control or PAUSE control for misbehaving flows. We also show, using simulation models, that RATE and PAUSE can complement each other where a switch is RATE controlled and the sending source PAUSE controlled. An appendix gives a formal specification of RATE.

An overloaded web server will lose incoming requests resulting in a "404 error" appearing at a client browser. Front-end application-level switches can redirect requests to less loaded servers. However, there exist no native methods within common web servers to automatically redirect requests for reducing load. We develop and evaluate changes to the open source Apache HTTP server to automatically redirect requests during periods of high load to a pre-specified overload server. We call the modified server "Red-Apache". To reduce the double round-trip times of HTTP 301 redirection, a new "server tag" substitution method of redirection at a client browser is investigated. The server tag enables single round-trip redirection and is implemented by a JavaScript function in a returned page. Using the AB and SURGE benchmarks, we show that Red-Apache and server tag outperform the original Apache server and HTTP redirection, respectively.

This paper presents the new shortest best path tree (SBPT) algorithm for multicast trees. The SBPT algorithm establishes and maintains dynamic multicast trees which maximize the bandwidth to be shared by multiple receivers and simultaneously guarantee the shortest paths for each receiver node. The SBPT algorithm is a distributed algorithm with cost in the same order as the sum of the shortest path tree (SPT) and Greedy algorithms. The SBPT algorithm reduces bandwidth consumption by utilizing partial paths already established for other multicast receiver nodes. The SBPT algorithm finds such partial paths when multiple shortest paths exist. Simulation experiments comparing the SBPT and SPT algorithms show that the SBPT algorithm reduces bandwidth consumption by 5% to 17% when node utilization is greater than approximately 25% and always achieves the same shortest path lengths.

In a typical real-time voice application, voice packets are produced at deterministically-spaced time intervals. In the network they encounter a variable amount of delay that changes the deterministic time intervals. A receiving host can employ a buffer to delay the playout of the voice packets in order to reconstruct the original timing. Adaptive techniques can perform continuous estimations of the network delays and dynamically adjust the buffering delay at the beginning of each talkspurt. Such adjustments are usually undetectable by the human listener. This research develops a new, adaptive "gap-based" algorithm that can be tuned for both end-to-end delay and packet loss to satisfy a user-desired tolerance. This new gap based algorithm adapts the buffering delay based on historical information of arrival and playout times of received voice packers in the previous talkspurt. A simulation study shows that the new gap based algorithm can reduce delay by 10% when compared with existing methods.

The full-duplex repeater (FDR) has previously been proposed as an alternative to half-duplex operation using CSMA/CD for controlling shared access to Gigabit Ethernet. In this paper, the basic FDR architecture is described and two extensions for traffic control are introduced. Using simulation methods, the performance of the Gigabit FDR is studied under different topologies and population sizes for a range of offered load. It is shown that the FDR provides a dramatic performance improvement over CSMA/CD (using both BEB and BLAM arbitration) at high load. The Gigabit FDR is also compared with switched Ethernet in the context of medical image retrieval. It is shown that for medical image retrieval, the performance of the Gigabit FDR is much better than 100/100 or 1000/100-Mbps switched Ethernet, and equivalent to 1000/1000-Mbps switched Ethernet for low levels of non-image background traffic.

Accurate traffic characterization by a packet source is needed to predict the network behavior and to properly allocate network resources to achieve a desired quality of service for all network users. As networks have become faster, the processing load required for complete packet sampling has also grown. In some cases, for example Gigabit Ethernet, the network can deliver packets faster than a network management subsystem can process them. In order to prevent inaccurate traffic statistics due to "clipping" of traffic peaks, Claffy et al. (1993) applied several static sampling strategies to network traffic characterization. As shown in this paper, static sampling may produce inaccurate traffic statistics. Adaptive sampling methods are developed and evaluated to address the inaccuracies of static sampling. In addition, the estimation of the Hurst parameter, a measure of traffic self-similarity, is studied for static and adaptive sampling. It is shown that adaptive sampling results in a more accurate estimation of the mean, variance, and Hurst parameter for packet counts.

Recent work has demonstrated that network traffic has self similar properties. These properties make short term control of traffic very difficult. Heavy tailed distributions of burst sizes contribute to traffic self similarity. The effects of heavy tailed file transfer traffic on queueing behavior are demonstrated using a simulated traffic source based on empirical Unix file size data. A method of application level traffic shaping, whereby selected large traffic bursts are shaped, is developed. This shaping method is shown to dramatically decrease ATM cell loss at a bottleneck queue. At the expense of a few large file transfers being increased in time duration, many smaller file transfers are decreased in time duration and cell loss is decreased for all file transfers.

The ATM Forum has adopted the rate-based scheme as the standard for ABR services. Effective ABR services are essential for ATM LAN emulation. The authors propose a new algorithm or scheme that has equivalent or better performance than existing schemes, but requires only O(1) computation. This O(1) computation is independent of both the number and rates of the connections. The scheme, called the "USF scheme", is compared via simulation to the existing EPRCA, MIT, and OSU schemes. Results show that the less complex USF scheme provides better throughput, delay, and fairness than the existing schemes.

The binary logarithmic arbitration method (BLAM) is currently under review by the IEEE 802.3 standards committee for possible standardization. BLAM is a new arbitration algorithm for Ethernet that is backwards compatible with the existing binary exponential backoff (BEB) algorithm. Simulation models of BEB and BLAM are developed. Using these models, four experiments evaluating BEB and BLAM under overload conditions are described. The experiments focus on a small population Ethernet, scaling to large population Ethernets, BLAM and BEB compatibility, and the ability to transmit 64-kbps voice streams. It is shown that BLAM supports higher throughput and lower delay than BEB. BLAM also demonstrates notably lower variability in delay than BEB. For the transmission of voice streams, it is shown that BLAM offers outstanding performance in the percentage of voice packets successfully transmitted within a hard deadline of 10 or 20 milliseconds.

Ethernet, token-ring, and FDDI are well established protocols for regulating the access to a common transmission medium among a large number of stations. All of the attached stations share this common medium and its bandwidth. Sharing of a common medium reduces the bandwidth that is available to an individual station. To increase bandwidth to the end user, emerging LAN topologies are departing from shared-media, shared-bandwidth methods in favor of dedicated-media and dedicated-bandwidth methods. Dedicated-bandwidth switched LANs can take advantage of full-duplex operation of attached stations. This is in contrast to the normal half-duplex operation of LAN stations on a shared-bandwidth LAN. This paper describes the evolution of shared-media, shared-bandwidth LANs into dedicated-media, dedicated-bandwidth switched LANs with full-duplex operation of the attached stations.>

The capabilities of a 16-Mb/s token ring to support multimedia applications, and simultaneously carry data and multimedia traffic are demonstrated. The token ring's priority mechanism makes it possible to effectively isolate multimedia and data traffic. Bandwidth management is then needed to allocate a guaranteed bandwidth to multimedia traffic. A simple form of bandwidth management limits the number of active multimedia sessions on a token ring segment. With bandwidth management implemented in a multimedia server, client/server multimedia applications on a single token ring segment can become a reality.

A multichannel LAN based on a ring topology is proposed. This LAN, the multichannel token ring (MCTR), is designed to meet the demand for additional bandwidth while preserving, as much as possible, existing hardware and software. An emerging technology, wavelength division multiplexing (WDM), is reviewed as a candidate technology for implementing the multiple channels of the MCTR over a single fiber optic cable. When compared to single-channel LANs, the MCTR is shown to offer increased reliability, bandwidth scalability, and price-performance advantages.>