Sanghwan Lee

Proactive vs reactive approaches to failure resilient routing

Dealing with network failures effectively is a major operational challenge for Internet service providers. Commonly deployed link state routing protocols such as OSPF react to link failures through global (i.e., network-wide) link state advertisements and routing table recomputations, causing significant forwarding discontinuity after a failure. The drawback with these protocols is that they need to trade off routing stability and forwarding continuity. To improve failure resiliency without jeopardizing routing stability, we propose a proactive local rerouting based approach called failure insensitive routing (FIR). The proposed approach prepares for failures using interface-specific forwarding, and upon a failure, suppresses the link state advertisement and instead triggers local rerouting using a backwarding table. In this paper, we prove that when no more than one link failure notification is suppressed, FIR always finds a loop-free path to a destination if one such path exists. We also formally analyze routing stability and network availability under both proactive and reactive approaches, and show that FIR provides better stability and availability than OSPF.

Fast local rerouting for handling transient link failures

Link failures are part of the day-to-day operation of a network due to many causes such as maintenance, faulty interfaces, and accidental fiber cuts. Commonly deployed link state routing protocols such as OSPF react to link failures through global link state advertisements and routing table recomputations causing significant forwarding discontinuity after a failure. Careful tuning of various parameters to accelerate routing convergence may cause instability when the majority of failures are transient. To enhance failure resiliency without jeopardizing routing stability, we propose a local rerouting based approach called failure insensitive routing. The proposed approach prepares for failures using interface-specific forwarding, and upon a failure, suppresses the link state advertisement and instead triggers local rerouting using a backwarding table. With this approach, when no more than one link failure notification is suppressed, a packet is guaranteed to be forwarded along a loop-free path to its destination if such a path exists. This paper demonstrates the feasibility, reliability, and stability of our approach.

Failure inferencing based fast rerouting for handling transient link and node failures

With the emergence of voice over IP and other real-time business applications, there is a growing demand for an IP network with high service availability. Unfortunately, in todays Internet, transient failures occur frequently due to faulty interfaces, router crashes, etc., and current IP networks lack the resiliency needed to provide high availability. To enhance availability, we proposed failure inferencing based fast rerouting (FIFR) approach that exploits the existence of a forwarding table per line-card, for lookup efficiency in current routers, to provide fast rerouting similar to MPLS, while adhering to the destination-based forwarding paradigm. In our previous work, we have shown that the FIFR approach can deal with single link failures. In this paper, we extend the FIFR approach to ensure loop-free packet delivery in case of single router failures also, thus mitigating the impact of many scenarios of failures. We demonstrate that the proposed approach not only provides high service availability but also incurs minimal routing overhead.

On suitability of Euclidean embedding of internet hosts

In this paper, we investigate the suitability of embedding Internet hosts into a Euclidean space given their pairwise distances (as measured by round-trip time). Using the classical scaling and matrix perturbation theories, we first establish the (sum of the) magnitude of negative eigenvalues of the (doubly-centered, squared) distance matrix as a measure of suitability of Euclidean embedding. We then show that the distance matrix among Internet hosts contains negative eigenvalues of large magnitude, implying that embedding the Internet hosts in a Euclidean space would incur relatively large errors. Motivated by earlier studies, we demonstrate that the inaccuracy of Euclidean embedding is caused by a large degree of triangle inequality violation (TIV) in the Internet distances, which leads to negative eigenvalues of large magnitude. Moreover, we show that the TIVs are likely to occur locally, hence, the distances among these close-by hosts cannot be estimated accurately using a global Euclidean embedding, in addition, increasing the dimension of embedding does not reduce the embedding errors. Based on these insights, we propose a new hybrid model for embedding the network nodes using only a 2-dimensional Euclidean coordinate system and small error adjustment terms. We show that the accuracy of the proposed embedding technique is as good as, if not better, than that of a 7-dimensional Euclidean embedding.

Blacklist-aided forwarding in static multihop wireless networks

Static broadband wireless networks, due to their ease of deployment, are likely to proliferate in the near future. The major stumbling block, however, is that wireless links are prone to external interference, channel fading, inclement weather, etc. Therefore scalable and reliable routing despite frequent link quality fluctuations is needed for accelerating the growth of these networks. Most of the wireless routing schemes proposed in the literature are less suitable for these networks, as they are designed primarily for mobile ad hoc networks with dynamic and unpredictable topologies. In this paper, we propose a novel link-state-based blacklist-aided forwarding (BAF) approach, that takes advantage of the fact that the nodes and therefore their adjacencies are relatively static, for scalable packet delivery in static wireless networks. Under BAF, each packet carries a blacklist, a minimal set of degraded-quality links encountered along its path, and the next hop is determined based on both its destination and blacklist. BAF provides loop-free delivery of packets to reachable destinations regardless of the number of degraded links in the network. We evaluate the performance of BAF and show that it is not only reliable but also scalable.

On suitability of Euclidean embedding for host-based network coordinate systems

In this paper, we investigate the suitability of embedding Internet hosts into a Euclidean space given their pairwise distances (as measured by round-trip time). Using the classical scaling and matrix perturbation theories, we first establish the (sum of the) magnitude of negative eigenvalues of the (doubly centered, squared) distance matrix as a measure of suitability of Euclidean embedding. We then show that the distance matrix among Internet hosts contains negative eigenvalues of large magnitude, implying that embedding the Internet hosts in a Euclidean space would incur relatively large errors. Motivated by earlier studies, we demonstrate that the inaccuracy of Euclidean embedding is caused by a large degree of triangle inequality violation (TIV) in the Internet distances, which leads to negative eigenvalues of large magnitude. Moreover, we show that the TIVs are likely to occur locally; hence the distances among these close-by hosts cannot be estimated accurately using a global Euclidean embedding. In addition, increasing the dimension of embedding does not reduce the embedding errors. Based on these insights, we propose a new hybrid model for embedding the network nodes using only a two-dimensional Euclidean coordinate system and small error adjustment terms. We show that the accuracy of the proposed embedding technique is as good as, if not better than, that of a seven-dimensional Euclidean embedding.

Avoiding transient loops through interface-specific forwarding

Under link-state routing protocols such as OSPF and IS-IS, when there is a change in the topology, propagation of link-state announcements, path recomputation, and updating of forwarding tables (FIBs) will all incur some delay before traffic forwarding can resume on alternate paths. During this convergence period, routers may have inconsistent views of the network, resulting in transient forwarding loops. Previous remedies proposed to address this issue enforce a certain order among the nodes in which they update their FIBs. While such approaches succeed in avoiding transient loops, they incur additional message overhead and increased convergence delay. We propose an alternate approach, loopless interface-specific forwarding (LISF), that averts transient loops by forwarding a packet based on both its incoming interface and destination. LISF requires no modifications to the existing link-state routing mechanisms. It is easily deployable with current routers since they already maintain a FIB at each interface for lookup efficiency. This paper presents the LISF approach, proves its correctness, discusses three alternative implementations of it and evaluates their performance.

Efficient Server Consolidation Considering Intra-Cluster Traffic

Due to the advancement of virtual machine technologies, large server farms or data centers consolidate their servers into a set of more powerful machines to increase the flexibility, agility, and adaptability to the fluctuating resource demands. Furthermore, managing a small number of powerful machines rather than a large number of low-performing machines decreases the management costs. In such a consolidation procedure, the network traffic as well as the usage of CPU and memory should be taken into account because the network bandwidth is also a critical resource. Since the network traffic of two servers or applications can be eliminated when they are consolidated into the same machine, traditional binpacking algorithms for the consolidation may not work efficiently. In this paper, we propose a cluster based consolidation algorithm, which reduces the maximum resource usage ratio of the machines. Through extensive simulation analysis, we show that the network traffic based clustering approach achieves a near optimal maximum resource usage, that is, the usage is close to the simple lower bound that only considers CPU and memory usage.

Leopard: a locality aware peer-to-peer system with no hot spot

Recent research [7,12,2] has shown that Internet hosts can be efficiently (i.e., without excessive measurements) mapped to a virtual (Euclidean) coordinate system, where the geometric distance between any two nodes in this virtual space approximates their real IP network distance (latency). Based on this result, in this paper, we propose an alternative approach that inherently incorporates a virtual coordinate system into a P2P network. In our system, called Leopard, a node is assigned a coordinate in the so-called node geo space as it joins the network, and obtains neighbor relationships that reflects network proximity from the beginning. The object id space and the node geo space are then “weaved” together via a novel technique called geographically-scoped hashing. Through analysis and simulation, we show three major desirable properties of Leopard to exemplify the power of this paradigm shift: i) a constant routing stretch, i.e., IP level network latency of object look-up is proportional to the distance between a requesting node and the target object; ii) always locates a near-by copy when multiple copies exist; and iii) effectively handles “flash crowd” traffic with near optimal load balancing.

Fundamental effects of clustering on the Euclidean embedding of internet hosts

The network distance estimation schemes based on Euclidean embedding have been shown to provide reasonably good overall accuracy. While some recent studies have revealed that triangle inequality violations (TIVs) inherent in network distances among Internet hosts fundamentally limit their accuracy, these Euclidean embedding methods are nonetheless appealing and useful for many applications due to their simplicity and scalability. In this paper, we investigate why the Euclidean embedding shows reasonable accuracy despite the prevalence of TIVs, focusing in particular on the effect of clustering among Internet hosts. Through mathematical analysis and experiments, we demonstrate that clustering of Internet hosts reduces the effective dimension of the distances, hence low-dimension Euclidean embedding suffices to produce reasonable accuracy. Our findings also provide us with good guidelines as to how to select landmarks to improve the accuracy, and explains why random selection of a large number of landmarks improves the accuracy.