Aniket Mahanti

A tale of the tails: Power-laws in internet measurements

Power-laws are ubiquitous in the Internet and its applications. This tutorial presents a review of power-laws with emphasis on observations from Internet measurements. First, we introduce power-laws and describe two commonly observed power-law distributions, the Pareto and Zipf distributions. Two frequently occurring terms associated with these distributions, specifically heavy tails and long tails, are also discussed. Second, the preferential attachment model, which is a widely used model for generating power-law graph structures, is reviewed. Subsequently, we present several examples of Internet workload properties that exhibit power-law behavior. Finally, we explore several implications of power-laws in computer networks. Using examples from past and present, we review how researchers have studied and exploited power-law properties. We observe that despite the challenges posed, power-laws have been effectively leveraged by researchers to improve the design and performance of Internet-based systems.

Comparative performance analysis of high-speed transfer protocols for big data

Researchers working in diverse fields such as astronomy, experimental physics, genomics, and meteorology have to frequently deal with analyzing voluminous amounts of complex data. Such data is often referred to as big data. These researchers work in teams and have to transfer this data over long distances. Efficiently transferring big data over long distances requires the use of appropriate transfer protocols. Several TCP-based and UDP-based protocols have been proposed in the literature, however, a comparative analysis of such protocols is lacking in the literature. This paper presents a comparative performance analysis of four well-known high-speed data transfer protocols for long fat networks, namely, GridFTP, FDT, UDT, and Tsunami. We performed extensive experiments to measure the effectiveness of each protocol in terms of its throughput for various roundtrip times, and against increasing levels of congestion inducing TCP or UDP background traffic on a 10 Gb/s network. Our results show that without much tuning, TCP based protocols are able to achieve throughputs of more than 2 Gb/s. In presence of background traffic, UDP protocols perform better.

FOG-Engine: Towards Big Data Analytics in the Fog

Existing platforms fall short in providing effective solutions for big data analytics while the demands for processing large quantities of data in real-time are increasing. Moving data analytics towards where the data is generated and stored could be a solution for addressing this issue. In this paper, we propose a solution referred as FOG-engine, which is integrated into IoTs near the ground and facilitates data analytics before offloading large amounts of data to a central location. In this work, we introduce a model for data analytic using FOG-engines and discuss our plan for evaluating its efficacy in terms of several performance metrics such as processing speed, network bandwidth, and data transfer size.

Performance and Fairness Issues in Big Data Transfers

We present performance and fairness analysis of two TCP- based (GridFTP and FDT) and one UDP-based (UDT) big data transfer protocols. We perform long-haul performance experiments using a 10 Gb/s national network, and conduct fairness tests in our 10 Gb/s local network. Our results show that GridFTP with jumbo frames provides fast data transfers. GridFTP is also fair in sharing bandwidth with competing background TCP flows.

Comparative analysis of big data transfer protocols in an international high-speed network

Large-scale scientific installations generate voluminous amounts of data (or big data) every day. These data often need to be transferred using high-speed links (typically with 10 Gb/s or more link capacity) to researchers located around the globe for storage and analysis. Efficiently transferring big data across countries or continents requires specialized big data transfer protocols. Several big data transfer protocols have been proposed in the literature, however, a comparative analysis of these protocols over a long distance international network is lacking in the literature. We present a comparative performance and fairness study of three open-source big data transfer protocols, namely, GridFTP, FDT, and UDT, using a 10 Gb/s high-speed link between New Zealand and Sweden. We find that there is limited performance difference between GridFTP and FDT. GridFTP is stable in terms of handling file system and TCP socket buffer. UDT has an implementation issue that limits its performance. FDT has issues with small buffer size limiting its performance, however, this problem is overcome by using multiple flows. Our work indicates that faster file systems and larger TCP socket buffers in both the operating system and application are useful in improving data transfer rates.

Profiling Energy Consumption in a Residential Campus

We present preliminary analysis of energy consumption in a residential academic campus in India. We deployed over 200 smart electric meters across the campus, collecting over 5 million data points every day. The analysis is based on data collected over a period of five months. Our results show skews in energy consumption for different electrical loads on campus. For example, a large proportion of the energy consumed in the campus is due to fixed electrical appliance infrastructure such as HVAC and UPS. We classify and group the locations based on their energy consumption profile during the day and week. Our work is a first step towards understanding energy consumption of an academic community. The results can be useful in implementing energy conservation strategies in the campus.

Characterizing cyberlocker traffic flows

Cyberlockers have recently become a very popular means of distributing content. Today, cyberlocker traffic accounts for a non-negligible fraction of the total Internet traffic volume, and is forecasted to grow significantly in the future. The underlying protocol used in cyberlockers is HTTP, and increased usage of these services could drastically alter the characteristics of Web traffic. In light of the evolving nature of Web traffic, updated traffic models are required to capture this change. Despite their popularity, there has been limited work on understanding the characteristics of traffic flows originating from cyberlockers. Using a year-long trace collected from a large campus network, we present a comprehensive characterization study of cyberlocker traffic at the transport layer. We use a combination of flow-level and host-level characteristics to provide insights into the behavior of cyberlockers and their impact on networks. We also develop statistical models that capture the salient features of cyberlocker traffic. Studying the transport-layer interaction is important for analyzing reliability, congestion, flow control, and impact on other layers as well as Internet hosts. Our results can be used in developing improved traffic simulation models that can aid in capacity planning and network traffic management.

Identifying User Actions from HTTP(S) Traffic

When understanding modern web usage and providing optimized personalized service, it is important to identify the HTTP(S) requests directly caused by user actions like clicks and typing web addresses. With a majority of HTTP(S) requests being due to content that has not been explicitly requested by a user, the problem of identifying user actions at proxies or middleboxes becomes non-trivial. We present an automated evaluation framework for identifying user actions while also automatically providing a ground truth of the user actions. We utilize the framework to compare the performance of timing-based and HTTP-aware request classifiers, including timing-based classifiers operating on both per-request and per-connection basis to identify user actions. We emphasize the value of diverse information used by the classifiers when comparing identification accuracy both among classifiers and relative to the browser-based ground truth. Our classifiers can be useful to better understand users web usage and connection prioritization.

Generation and Risk Analysis of Network Attack Graph

Attack graph describes how an attacker can compromise with network security. To generate the attack graph, we required system as well as vulnerability information. The system information contains scanned data of a network, which is to be analyzed. The vulnerability data contain information about, how exploits can be generated due to multiple vulnerabilities and what effects can be of such exploitation. Multihost multistage vulnerability analysis (MulVAL) tool is used for generating attack graph in this work. MulVAL generated graphs are logical attack graphs based on logical programming and based on dependencies among attack goal and configuration information. The risk of network attack graph is measured through graph topology theoretic properties (connectivity, cycles, and depth), and analysis of possible attacks paths is carried out in this paper.

Characterizing performance and fairness of big data transfer protocols on long-haul networks

This paper presents a characterization study of big data transfer protocols on a long-haul network. We analyzed the performance and fairness of three well-known open-source protocols, namely, GridFTP, FDT, and UDT. Using a real-world 10 Gb/s network link between New Zealand and Sweden, we studied data transfer rates (in terms of goodput) and fairness (in terms of impact on round trip time) of the protocols. We performed extensive experiments using single and multiple data flows to comprehend how these protocols behave in real-world situations. We found that GridFTP has the fastest data transfer rates when using a single flow. UDT suffered from poor performance due to implementation issues. A small buffer size limited FDTs performance, however, this drawback can be overcome by using multiple flows in lieu of fairness.