Mikhail Yakimov

Scalable blind search and broadcasting over Distributed Hash Tables

Typical blind search algorithms in P2P networks generate a significant amount of duplicate query messages in order to increase the success rate. We present a novel framework, named Recursive Partitioning Search (RPS), for blind search over structured peer-to-peer (P2P) networks, by which the query message duplication can be avoided. Two realizations of the framework for Chord and Pastry are presented. By simulation, we compare success rate, lookup delay and overlay network load of RPS with various well-known blind search algorithms, and illustrate RPS being a superior blind search algorithm running over DHTs. The algorithm guarantees that with high probability the lookup delay to visit every peer is of O(logN) hops, comparable to the delay of the exact-match search over the DHTs, which is proved for two example DHTs, Chord and Pastry in the paper. RPS is a simple and intuitive method for blind search over DHTs compared to other complex approaches like those building sophisticated index structures or requiring analysis of the words in the stored documents, yet a lot more efficient than known simple methods like Flooding and Random Walk. With RPS, every node in the overlay network is visited not more than once by design. These characteristics qualify the Recursive Partitioning Search over DHT as an efficient broadcasting algorithm. We investigate RPS scalability and propose a formula to choose an appropriate Time-to-Live (TTL) parameter value to maintain the balance between high success rate and reasonable network load. Active peer churn degrades the performance of RPS as a broadcasting mechanism proportionally to the churn rate. But the success rate of blind search using RPS may be affected negligibly if proper replications exist as in most P2P file sharing networks.

Analytical Study of QoS-Oriented Multicast in Wireless Networks

Multicast is a very popular bandwidth-conserving technology exploited in many multimedia applications. However, existing standards of high rate wireless networks provide no error recovery mechanism (ARQ) for multicast traffic. ARQ absence in wireless networks unreliable by their nature leads to frequent packet losses, which is inappropriate for most of multimedia applications. In this paper, we study new reliable multicast mechanism proposed recently to support multimedia QoS (packet loss ratio, latency, and throughput) with various wireless technologies. This mechanism is based on the concept of multiple ACK-leaders, that is, multicast recipients responsible for acknowledging data packets. We develop analytical models of the mechanism with various leader selection schemes and use the models to study the schemes efficiency and to optimize them. Numerical results show that the novel multicast mechanism with multiple ACK-leaders can be easily tuned to meet specific QoS requirements of multimedia or any other multicast applications.

Scalable Blind Search and Broadcasting in Peer-to-Peer Networks

Typical blind search algorithms in P2P networks generate a significant amount of duplicate query messages to increase the success rate. We present a framework, named recursive partitioning search (RPS), for blind search over structured peer-to-peer (P2P) networks, by which the query message duplication can be avoided. Two realizations of the framework for Chord and Pastry are presented. With simulation model, we compare success rate, lookup delay and overlay network load of RPS and various well-known blind search algorithms, and show RPS is the superior blind search algorithm running over a DHT. The algorithm guarantees that with high probability the lookup delay to visit every peer is of O(log N) hops, comparable to the delay of the exact match search over the DHTs. With RPS, every node in the overlay network is visited not more than once by design. These characteristics qualify the recursive partitioning search over DHT as an efficient broadcasting algorithm. We investigate RPS scalability and propose the way to choose an appropriate time-to-live (TTL) parameter value to maintain the balance between high success rate and reasonable network load

Multicast QoS Support in IEEE 802.11 WLANs

Existing IEEE 802.11 spec, provides no error recovery mechanism (ARQ) for multicast traffic. ARQ absence in wireless networks unreliable by their nature leads to frequent packet losses, which is inappropriate for most of multimedia applications. In the paper, we propose new reliable multicast mechanism supporting multimedia QoS (packet loss ratio, delay and throughput), in IEEE 802.11 Wireless local Area Networks (WLANs). This mechanism is based on the concept of multiple ACK-leaders, i.e. multicast recipients responsible for data packets acknowledging. We develop an analytical model of the mechanism and use the model to study and to optimize the mechanism.

Starvation effect study in IEEE 802.11 mesh networks

Initially, IEEE 802.11 standard was not designed for multihop networks, but its market success gave it a powerful incentive, and nowadays IEEE 802.11 Working Group is preparing various amendments and looking for new application scenarios to ensure future success in the market. In this paper, we analyze the performance of an IEEE 802.11s mesh network and show that the effect of starvation caused by hidden terminals interference makes the distribution of channel capacity between links very much disproportional and unpredictable. Some links may be suppressed completely by other links and their throughput is close to zero. The effect of starvation was discovered by simulations in many papers. We investigate the roots of this phenomenon in a mesh network with an analytical model. Also, this paper reports the results of an outdoor testbed which was deployed to find out the scale of the starvation problem in a real network with Voice-over-IP, Video-over-IP and FTP traffic. While the results for VoIP may be considered as satisfactory, FTP traffic suffers dramatically from interference. So, at the end of the paper, we outline possible solutions to address starvation problem in mesh networks.

Tag Routing for Efficient Blind Search in Peer-to-Peer Networks

A variety of algorithms have been developed for Blind Search (BS) over peer-to-peer (P2P) overlay networks, i.e., search of resources without any routing information for the exact location of the target resource. All these algorithms generate significant amount of duplicate query messages to increase the success rate. In this paper, we introduce tag routing search - a novel framework for BS over P2P overlay networks structured by Distributed Hash Tables (DHTs). A query message in this framework contains a tag that affect the selection of next hop nodes. We propose two specific algorithms: Zigzag Modified Breadth First Search with Local Indices (ZMBFS+L) and Recursive Partitioning Search (RPS). ZMBFS+L achieves high success rate with less duplications compared to other known BS algorithms. RPS, taking advantage of the DHT structure, does not generate any message duplication by design, thus is a very efficient broadcasting algorithm, and can achieve 100% success rate efficiently.

Direct transmission protection in IEEE 802.11 networks

In this paper, we discuss some problems related to direct transmissions between QSTAs in an infrastructure 802.11 WLAN with a legacy access point (AP). In the infrastructure mode, collisions of data frames caused by hidden terminals interference can be avoided by the RTS/CTS exchange anticipating the data frames, if and only if all data communications between stations are performed via the AP. The interference problem is solved because all terminals are within the transmission range (TX range) of the AP. However, the RTS/CTS exchange is not enough to protect direct transmissions between QSTAs. In this paper, we analyze the state-of-the-art direct transmissions protection approaches, show their limitations, propose and simulate our own protection schemes.

An Efficient Packet Transmission Scheme for Ultra High Rate WPAN

In the paper we propose and study a novel packet transmission scheme aimed to overcome a fundamental overhead of WPAN MAC/PHY protocols and to improve ARQ performance. Specifically, we propose to combine two well- known techniques: fragmentation and aggregation by adding multiple Frame Control Sums (FCS) to each fragment. We show analytically and by simulation that the proposed scheme increases WiMedia MAC efficiency greatly. Although the WiMedia MAC has been used as a design base, this novel packet transmission scheme can be extended to other high rate wireless communication systems to cover mmWave.