Boris Nechaev

Netalyzr: illuminating the edge network

In this paper we present Netalyzr, a network measurement and debugging service that evaluates the functionality provided by peoples Internet connectivity. The design aims to prove both comprehensive in terms of the properties we measure and easy to employ and understand for users with little technical background. We structure Netalyzr as a signed Java applet (which users access via their Web browser) that communicates with a suite of measurement-specific servers. Traffic between the two then probes for a diverse set of network properties, including outbound port filtering, hidden in-network HTTP caches, DNS manipulations, NAT behavior, path MTU issues, IPv6 support, and access-modem buffer capacity. In addition to reporting results to the user, Netalyzr also forms the foundation for an extensive measurement of edge-network properties. To this end, along with describing Netalyzr s architecture and system implementation, we present a detailed study of 130,000 measurement sessions that the service has recorded since we made it publicly available in June 2009.

How penalty leads to improvement

Despite much theoretical work, different modifications of backoff protocols in 802.11 networks lack empirical evidence demonstrating their real-life performance. To fill the gap we have set out to experiment with performance of exponential backoff by varying its backoff factor. Despite the satisfactory results for throughput, we have witnessed poor fairness manifesting in severe capture effect. The design of standard backoff protocol allows already successful nodes to remain successful, giving little chance to those nodes that failed to capture the channel in the beginning. With this at hand, we ask a conceptual question: Can one improve the performance of wireless backoff by introducing a mechanism of self-penalty, when overly successful nodes are penalized with big contention windows? Our real-life measurements using commodity hardware demonstrate that in many settings such mechanism not only allows to achieve better throughput, but also assures nearly perfect fairness. We further corroborate these results with simulations and an analytical model. Finally, we present a backoff factor selection protocol which can be implemented in access points to enable deployment of the penalty backoff protocol to consumer devices.

Experiences from Netalyzr with engaging users in end-system measurement

Netalyzr is a widely used network diagnostic and debugging tool that has collected 259,000 measurement sessions to date. To use Netalyzr, users visit its website, download an applet that proceeds to conduct a suite of tests and measurements, and obtain a summary report detailing the findings. Along with the measurement data, for each session, we record the HTTP referrer that brought the user to the Netalyzr page, the level of trust the user bestowed upon the applet, and any feedback that the user voluntarily left via a form that we include at the bottom of the report page. These data sources illuminate how Netalyzrs users employ the tool, and can provide insights as to how other measurement tools or user surveys involving end-host measurement could effectively involve users. We find that even with little prompting, users leave explicit comments 3% of the time and answer one or more survey questions in 17% of the sessions, reaching up to 44% og sessions during bursts of activity. We also find that significant usage of the tool comes from four types of need: (i) to aid in troubleshooting performance for an on-line game, often via measurement sessions conducted when requested by more sophisticated users in a help forum; (ii) curiosity, often exacerbated by blog postings and other mentions on high-profile websites; (iii) repeat visitors who arrive via a search engine that they used to locate Netalyzrs website; and (iv) IPv6 deployment tests conducted or organized by specialists.

Demand-aware flow allocation in data center networks

In this work we consider a relatively large and highly dynamic data center network in which flows have small interarrival times and different demands for the network resources. Taking into account the properties and specifics of such networks we consider the problem of flow placement, i.e. assignment of an outgoing port for flows at each hop from source to destination. Using the characteristics of modern data centers from previous measurement studies, in this work we first simulate the flow allocation using several algorithms with and without global knowledge. We find that in all settings local forwarding decisions are almost as good as decisions made with global information at hand. This finding enables us to propose a fully distributed mechanism that relies only on local knowledge and allows to achieve fair and demand aware flow allocation in the data center network. The mechanism has low complexity and performs better than naive random flow allocation.

CR-Chord: Improving lookup availability in the presence of malicious DHT nodes

Distributed Hash Tables (DHTs) provide a useful key-to-value lookup service for many Internet applications. However, without additional mechanisms DHTs are vulnerable to attacks. In particular, previous research showed that Chord is not well resistant to malicious nodes that joined the DHT. We introduce the cyclic routing algorithm as an extension of Chord (CR-Chord). Using simulations we compare the lookup availability of Chord and CR-Chord. The results suggest that CR-Chord improves the lookup availability on the average by 1.4 times. When the number of malicious nodes is small, such as 5%, CR-Chord has almost twice lower lookup failure rate.

Cyclic routing: Generalizing look-ahead in peer-to-peer networks

Distributed Hash Tables (DHT) provide a lookup service in peer-to-peer overlay networks. Many valuable applications have been recently built on top of several available DHTs. However, they function poorly when no direct IP connectivity is available to some nodes (e.g., located behind a NAT or firewall) or in the presence of overloaded or malicious nodes. In this paper, we propose a new method for DHT-based routing called cyclic routing. It generalizes existing single-hop look-ahead approach (also known as “Know thy neighbors neighbor”) and supports multipath routing. The method provides a systematic way for collecting stable and efficient overlay paths. Cyclic routing has the same theoretical dependability and efficiency upper bounds as basic DHT routing but it is more resilient when IP connectivity is limited or when the overlay suffers from overloaded nodes.

A novel demand-aware fairness metric for IEEE 802.11 wireless networks

Even though literature that focuses on improving fairness among wireless stations in single hop IEEE 802.11 networks is rife, little is discussed on how to measure fairness in such networks when stations have unequal demands for resources. Typically, the performance of such protocols is measured assuming that all stations have equal resource demands under full channel saturation. But if these protocols are evaluated in real environments, where such assumption may not hold, measuring fairness with off the shelf metrics may give inaccurate and inadequate results. To account for these settings, we propose a demand-aware fairness index (DA-index). We argue that the suggested metric is a useful tool for investigating fairness in single-hop IEEE 802.11 networks where resource allocation is governed with backoff protocols. We demonstrate the merits of the proposed metrics with analysis and empirical evaluation.