Sylvia Ratnasamy

Making gnutella-like P2P systems scalable

Napster pioneered the idea of peer-to-peer file sharing, and supported it with a centralized file search facility. Subsequent P2P systems like Gnutella adopted decentralized search algorithms. However, Gnutellas notoriously poor scaling led some to propose distributed hash table solutions to the wide-area file search problem. Contrary to that trend, we advocate retaining Gnutellas simplicity while proposing new mechanisms that greatly improve its scalability. Building upon prior research [1, 12, 22], we propose several modifications to Gnutellas design that dynamically adapt the overlay topology and the search algorithms in order to accommodate the natural heterogeneity present in most peer-to-peer systems. We test our design through simulations and the results show three to five orders of magnitude improvement in total system capacity. We also report on a prototype implementation and its deployment on a testbed.

GHT: a geographic hash table for data-centric storage

Making effective use of the vast amounts of data gathered by large-scale sensor networks will require scalable, self-organizing, and energy-efficient data dissemination algorithms. Previous work has identified data-centric routing as one such method. In an asso-ciated position paper [23], we argue that a companion method, data-centric storage (DCS), is also a useful approach. Under DCS, sensed data are stored at a node determined by the name associated with the sensed data. In this paper, we describe GHT, a Geographic Hash Table system for DCS on sensornets. GHT hashes keys into geographic coordi-nates, and stores a key-value pair at the sensor node geographically nearest the hash of its key. The system replicates stored data lo-cally to ensure persistence when nodes fail. It uses an efficient consistency protocol to ensure that key-value pairs are stored at the appropriate nodes after topological changes. And it distributes load throughout the network using a geographic hierarchy. We evaluate the performance of GHT as a DCS system in simulation against two other dissemination approaches. Our results demonstrate that GHT is the preferable approach for the application workloads predicted in [23], offers high data availability, and scales to large sensornet deployments, even when nodes fail or are mobile.

Topologically-aware overlay construction and server selection

A number of large-scale distributed Internet applications could potentially benefit from some level of knowledge about the relative proximity between its participating host nodes. For example, the performance of large overlay networks could be improved if the application-level connectivity between the nodes in these networks is congruent with the underlying IP-level topology. Similarly, in the case of replicated Web content, client nodes could use topological information in selecting one of multiple available servers. For such applications, one need not find the optimal solution in order to achieve significant practical benefits. Thus, these applications, and presumably others like them, do not require exact topological information and can instead use sufficiently informative hints about the relative positions of Internet hosts. In this paper, we present a binning scheme whereby nodes partition themselves into bins such that nodes that fall within a given bin are relatively close to one another in terms of network latency. Our binning strategy is simple (requiring minimal support from any measurement infrastructure), scalable (requiring no form of global knowledge, each node only needs knowledge of a small number of well-known landmark nodes) and completely distributed (requiring no communication or cooperation between the nodes being binned). We apply this binning strategy to the two applications mentioned above: overlay network construction and server selection. We test our binning strategy and its application using simulation and Internet measurement traces. Our results indicate that the performance of these applications can be significantly improved by even the rather coarse-grained knowledge of topology offered by our binning scheme.

Geographic routing without location information

For many years, scalable routing for wireless communication systems was a compelling but elusive goal. Recently, several routing algorithms that exploit geographic information (e.g. GPSR) have been proposed to achieve this goal. These algorithms refer to nodes by their location, not address, and use those coordinates to route greedily, when possible, towards the destination. However, there are many situations where location information is not available at the nodes, and so geographic methods cannot be used. In this paper we define a scalable coordinate-based routing algorithm that does not rely on location information, and thus can be used in a wide variety of ad hoc and sensornet environments.

Application-Level Multicast Using Content-Addressable Networks

Most currently proposed solutions to application-level multicast organise the group members into an application-level mesh over which a Distance-Vector routing protocol, or a similar algorithm, is used to construct source-rooted distribution trees. The use of a global routing protocol limits the scalability of these systems. Other proposed solutions that scale to larger numbers of receivers do so by restricting the multicast service model to be single-sourced. In this paper, we propose an application-level multicast scheme capable of scaling to large group sizes without restricting the service model to a single source. Our scheme builds on recent work on Content-Addressable Networks (CANs). Extending the CAN framework to support multicast comes at trivial additional cost and, because of the structured nature of CAN topologies, obviates the need for a multicast routingalg orithm. Given the deployment of a distributed infrastructure such as a CAN, we believe our CAN-based multicast scheme offers the dual advantages of simplicity and scalability.

The impact of DHT routing geometry on resilience and proximity

The various proposed DHT routing algorithms embody several different underlying routing geometries. These geometries include hypercubes, rings, tree-like structures, and butterfly networks. In this paper we focus on how these basic geometric approaches affect the resilience and proximity properties of DHTs. One factor that distinguishes these geometries is the degree of flexibility they provide in the selection of neighbors and routes. Flexibility is an important factor in achieving good static resilience and effective proximity neighbor and route selection. Our basic finding is that, despite our initial preference for more complex geometries, the ring geometry allows the greatest flexibility, and hence achieves the best resilience and proximity performance.

Data-centric storage in sensornets with GHT, a geographic hash table

Making effective use of the vast amounts of data gathered by large-scale sensor networks (sensornets) will require scalable, self-organizing, and energy-efficient data dissemination algorithms. For sensornets, where the content of the data is more important than the identity of the node that gathers them, researchers have found it useful to move away from the Internets point-to-point communication abstraction and instead adopt abstractions that are more data-centric. This approach entails naming the data and using communication abstractions that refer to those names rather than to node network addresses [1,11]. Previous work on data-centric routing has shown it to be an energy-efficient data dissemination method for sensornets [12]. Herein, we argue that a companion method, data-centric storage (DCS), is also a useful approach. Under DCS, sensed data are stored at a node determined by the name associated with the sensed data. In this paper, we first define DCS and predict analytically where it outperforms other data dissemination approaches. We then describe GHT, a Geographic Hash Table system for DCS on sensornets. GHT hashes keys into geographic coordinates, and stores a key–value pair at the sensor node geographically nearest the hash of its key. The system replicates stored data locally to ensure persistence when nodes fail. It uses an efficient consistency protocol to ensure that key–value pairs are stored at the appropriate nodes after topological changes. And it distributes load throughout the network using a geographic hierarchy. We evaluate the performance of GHT as a DCS system in simulation against two other dissemination approaches. Our results demonstrate that GHT is the preferable approach for the application workloads we analytically predict, offers high data availability, and scales to large sensornet deployments, even when nodes fail or are mobile.

RouteBricks: exploiting parallelism to scale software routers

We revisit the problem of scaling software routers, motivated by recent advances in server technology that enable high-speed parallel processing--a feature router workloads appear ideally suited to exploit. We propose a software router architecture that parallelizes router functionality both across multiple servers and across multiple cores within a single server. By carefully exploiting parallelism at every opportunity, we demonstrate a 35Gbps parallel router prototype; this router capacity can be linearly scaled through the use of additional servers. Our prototype router is fully programmable using the familiar Click/Linux environment and is built entirely from off-the-shelf, general-purpose server hardware.

Data-centric storage in sensornets

Sensornets are large-scale distributed sensing networks comprised of many small sensing devices equipped with memory, processors, and short-range wireless communication. Making effective use of sensornet data will require scalable, self-organizing, and energy-efficient data dissemination algorithms. Recent work has identified data-centric routing as one such method. In this paper we suggest that a companion method, data-centric storage, may also be a useful approach. While there are many ways to achieve data-centric storage, this paper proposes a mechanism that builds upon two recent advances; (1) the GPSR geographic routing algorithm and (2) a new generation of efficient peer-to-peer lookup systems (such as Chord, CAN, Pastry, Tapestry, etc.). We evaluate the performance of data-centric storage and two other dissemination approaches in several sensornet scenarios and identify the conditions under which the various approaches are preferable.

Making middleboxes someone else's problem: network processing as a cloud service

Modern enterprises almost ubiquitously deploy middlebox processing services to improve security and performance in their networks. Despite this, we find that todays middlebox infrastructure is expensive, complex to manage, and creates new failure modes for the networks that use them. Given the promise of cloud computing to decrease costs, ease management, and provide elasticity and fault-tolerance, we argue that middlebox processing can benefit from outsourcing the cloud. Arriving at a feasible implementation, however, is challenging due to the need to achieve functional equivalence with traditional middlebox deployments without sacrificing performance or increasing network complexity. In this paper, we motivate, design, and implement APLOMB, a practical service for outsourcing enterprise middlebox processing to the cloud. Our discussion of APLOMB is data-driven, guided by a survey of 57 enterprise networks, the first large-scale academic study of middlebox deployment. We show that APLOMB solves real problems faced by network administrators, can outsource over 90% of middlebox hardware in a typical large enterprise network, and, in a case study of a real enterprise, imposes an average latency penalty of 1.1ms and median bandwidth inflation of 3.8%.