Andrey Lukyanenko

Cooperative caching through routing control in information-centric networks

Information-centric network (ICN), which is one of the prominent Internet re-design architectures, relies on in-network caching for its fundamental operation. However, previous works argue that the performance of in-network caching is highly degraded with the current cache-along-default-path design, which makes popular objects to be cached redundantly in many places. Thus, it would be beneficial to have a distributed and uncoordinated design. Although cooperative caches could be an answer to this, previous research showed that they are generally unfeasible due to excessive signaling burden, protocol complexity, and a need for fault tolerance. In this work we illustrate the ICN caching problem, and propose a novel architecture to overcome the problem of uncooperative caches. Our design possesses the cooperation property intrinsically. We utilize controlled off-path caching to achieve almost 9-fold increase in cache efficiency, and around 20% increase in server load reduction when compared to the classic on-path caching used in ICN proposals.

Network coding based multipath TCP

Multipath TCP (MPTCP) suffers from the degradation of goodput in the presence of diverse network conditions on the available subflows. The goodput can even be worse than that of one regular TCP, undermining the advantage gained by using multipath transfer. In this work, we propose a new multipath TCP protocol, namely NC-MPTCP, which introduces network coding (NC) to some but not all subflows traveling from source to destination. At the core of our scheme is the mixed use of regular and NC subflows. Thus, the regular subflows deliver original data while the NC subflows deliver linear combinations of the original data. The idea is to take advantage of the redundant NC data to compensate for the lost or delayed data in order to avoid receive buffer becoming full. We design a packet scheduling algorithm and a redundancy estimation algorithm to allocate data among different subflows in order to optimize the overall goodput. We also give a guideline on how to choose the NC subflows among the available subflows. We evaluate the performance of NC-MPTCP through a NS-3 network simulator. The experiments show that NC-MPTCP achieves higher goodput compared to MPTCP in the presence of different subflow qualities. And in the worst case, the performance of NC-MPTCP is close to that of one regular TCP.

Hi3: An efficient and secure networking architecture for mobile hosts

The Host Identity Indirection Infrastructure (Hi3) is a networking architecture for mobile hosts, derived from the Internet Indirection Infrastructure (i3) and the Host Identity Protocol (HIP). Hi3 has efficient support for secure mobility and multihoming, which both are crucial for future Internet applications. In this paper, we describe and analyze Hi3 in detail. Compared to existing solutions, Hi3 achieves better resilience, scalability, and security. Both our analysis and early measurements support the notion that Hi3 preserves the best of both approaches while improving performance compared to i3 and enhancing flexibility and security compared to HIP.

Multipath Transmission for the Internet: A Survey

Smart devices equipped with multiple network interfaces are becoming commonplace. Nevertheless, even though multiple interfaces can be used to connect to the Internet, their capabilities have not been fully utilized yet because the default TCP/IP stack supports only a single interface for communication. This situation is now changing due to the emergence of multipath protocols on different network stack layers. For example, many IP level approaches have been proposed utilizing tunneling mechanisms for hiding multipath transmission from the transport protocols. Several working groups under IEEE and IETF are actively standardizing multipath transmission on the link layer and transport layer. Application level approaches enable multipath transmission capability by establishing multiple transport connections and distributing data over them. Given all these efforts, it is beneficial and timely to summarize the state-of-the-art, compare their pros and cons, and discuss about the future directions. To that end, we present a survey on multipath transmission and make several major contributions: 1) we present a complete taxonomy pertaining to multipath transmission, including link, network, transport, application, and cross layers; 2) we survey the state-of-the-art for each layer, investigate the problems that each layer aims to address, and make comprehensive assessment of the solutions; and 3) based on the comparison, we identify open issues and pinpoint future directions for multipath transmission research.

Tolerating Path Heterogeneity in Multipath TCP with Bounded Receive Buffers

Abstract With bounded receive buffers, the aggregate bandwidth of multipath transmission degrades significantly in the presence of path heterogeneity. The performance could even be worse than that of single-path TCP, undermining the advantage gained by using multipath transmit. Furthermore, multipath transmission also suffers from delay and jitter even with large receive buffers. In order to tolerate the path heterogeneity when the receive buffer is bounded, we propose a new multipath TCP protocol, namely SC-MPTCP, by integrating linear systematic coding into MPTCP. In SC-MPTCP, we make use of coded packets as redundancy to counter against expensive retransmissions. The redundancy is provisioned into both proactive and reactive data. Specifically, to send a generation of packets, SC-MPTCP transmits proactive redundancy first and then delivers the original packets, instead of encoding all sent-out packets as all the existing coding solutions have done. The proactive redundancy is continuously updated according to the estimated aggregate retransmission ratio. In order to avoid the proactive redundancy being underestimated, the pre-blocking warning mechanism is utilized to retrieve the reactive redundancy from the sender. We use an NS-3 network simulator to evaluate the performance of SC-MPTCP with and without the coupled congestion control option. The results show that with bounded receive buffers, MPTCP achieves less than 20% of the optimal goodput with diverse packet losses, whereas SC-MPTCP approaches the optimal performance with significantly smaller receive buffers. With the help of systematic coding, SC-MPTCP reduces the average buffer delay of MPTCP by at least 80% in different test scenarios. We also demonstrate that the use of systematic coding could significantly reduce the arithmetic complexity compared with the use of non-systematic coding.

MPTCP incast in data center networks

In recent years, dual-homed topologies have appeared in data centers in order to offer higher aggregate bandwidth by using multiple paths simultaneously. Multipath TCP (MPTCP) has been proposed as a replacement for TCP in those topologies as it can efficiently offer improved throughput and better fairness. However, we have found that MPTCP has a problem in terms of incast collapse where the receiver suffers a drastic goodput drop when it simultaneously requests data over multiple servers. In this paper, we investigate why the goodput collapses even if MPTCP is able to actively relieve hot spots. In order to address the problem, we propose an equally-weighted congestion control algorithm for MPTCP, namely EW-MPTCP, without need for centralized control, additional infrastructure and a hardware upgrade. In our scheme, in addition to the coupled congestion control performed on each subflow of an MPTCP connection, we allow each subflow to perform an additional congestion control operation by weighting the congestion window in reverse proportion to the number of servers. The goal is to mitigate incast collapse by allowing multiple MPTCP subflows to compete fairly with a single-TCP flow at the shared bottleneck. The simulation results show that our solution mitigates the incast problem and noticeably improves goodput in data centers.

isBF: Scalable in-packet bloom filter based multicast

Abstract Bloom filter (BF) based forwarding was proposed recently in several protocol alternatives to IP multicast. Some of these protocols avoid the state in intermediate routers and leave the burden of scalability management to the multicast source and end-hosts. Still, the existing BF-based protocols have scalability limitations and require explicit network management as well as non-trivial functionality from the network components. In this work we address the scalability limitations of the BF-based forwarding protocols by partitioning end-hosts into clusters. We propose several algorithms to do the partitioning so as to decrease the overall traffic in the network. We evaluate our algorithms in a real Internet topology, demonstrating the ability of the proposed design to save up to 70% of traffic volume in the large-scale topology for big groups of subscribers, and up to 30% for small groups.

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.

CombiHeader: Minimizing the number of shim headers in redundancy elimination systems

Redundancy elimination has been used in many places to improve network performance. The algorithms for doing this typically split data into chunks, fingerprint them, and compare the fingerprint with cache to identify similar chunks. Then these chunks are removed from the data and headers are inserted instead of them. However, this approach presents us with two crucial shortcomings. Depending on the size of chunks, either many headers need to be inserted, or probability of missing similar regions is increased. Algorithms that try to overcome missed similarity detection by expanding chunk boundary suffers from excessive memory access due to byte-by-byte comparison. This situation leads us to propose a novel algorithm, CombiHeader, that allows near maximum similarity detection using smaller chunks sizes while using chunk aggregation technique to transmit very few headers with few memory accesses. CombiHeader uses a specialized directed graph to track and merge adjacent popular chunks. By generating different generations of CombiNodes, CombiHeader can detect different lengths of similarity region, and uses the smallest number of headers possible. Experiments show that CombiHeader uses less than 25% headers than general elimination algorithms, and this number improves with the number of hits. The required memory access to detect maximal similarity region is in the range of 1%-5% of comparable algorithms for certain situations. CombiHeader is implemented as a pluggable module, which can be used with any existing redundancy elimination algorithm.

The Delayed ACK evolution in MPTCP

Multipath TCP (MPTCP) is a major extension of TCP that aims to offer higher aggregate bandwidth and robustness by pooling multiple paths within one transport connection. One of the current weaknesses of the protocol is the Delayed ACK scheme inherited from TCP. The Delayed ACK is an option of TCP that allows the receiver to delay sending an ACK for every other packet within a window given by the Delayed ACK timer. At the sender, the RTO should be no less than the Minimum RTO to avoid spurious timeouts. This strategy can lead to significant performance degradation in the presence of timeouts, especially in high speed networks, where RTT is usually one- or two-order of magnitude smaller than the Minimum RTO. When a subflow occurs a timeout, the receiver has to buffer data from all the subflows until the missing packet is received. The data may overrun the receive buffer to cause flow control at the sender, which seriously impacts the overall performance. In order to avoid MPTCP performance degradation, we propose a new Delayed ACK aiming to remove the Minimum RTO constraint at the sender while to reserve the Delayed ACK function at the receiver. Our solution requires only minor modification to the legacy Delayed ACK scheme and it introduces negligible computational overhead and no extra traffic overhead. We use a NS-3 network simulator to evaluate the performance of our new Delayed ACK in three typical network environments. The results indicate that MPTCP using new Delayed ACK scheme requires much smaller aggregate buffer than MPTCP using the legacy Delayed ACK scheme, especially in high speed networks where the buffer requirement reduces one- or two-order of magnitude.