During the past decade, network and storage devices have undergone rapid performance improvements, delivering ultra-low latency and several Gbps of bandwidth. Nevertheless, current network and storage stacks fail to deliver this hardware performance to the applications, often due to the loss of I/O efficiency from stalled CPU performance. While many efforts attempt to address this issue solely on either the network or the storage stack, achieving high-performance for networked-storage applications requires a holistic approach that considers both. In this article, we present FlashNet, a software I/O stack that unifies high-performance network properties with flash storage access and management. FlashNet builds on RDMA principles and abstractions to provide a direct, asynchronous, end-to-end data path between a client and remote flash storage. The key insight behind FlashNet is to co-design the stack’s components (an RDMA controller, a flash controller, and a file system) to enable cross-stack optimizations and maximize I/O efficiency. In micro-benchmarks, FlashNet improves 4kB network I/O operations per second (IOPS by 38.6% to 1.22M, decreases access latency by 43.5% to 50.4μs, and prolongs the flash lifetime by 1.6-5.9× for writes. We illustrate the capabilities of FlashNet by building a Key-Value store and porting a distributed data store that uses RDMA on it. The use of FlashNet’s RDMA API improves the performance of KV store by 2× and requires minimum changes for the ported data store to access remote flash devices.

During the past decade, network and storage devices have undergone rapid performance improvements, delivering ultra-low latency and several Gbps of bandwidth. Nevertheless, current network and storage stacks fail to deliver this hardware performance to the applications, often due to the loss of IO efficiency from stalled CPU performance. While many efforts attempt to address this issue solely on either the network or the storage stack, achieving high-performance for networked-storage applications requires a holistic approach that considers both.

Distributed DRAM stores have become an attractive option for providing fast data accesses to analytics applications. To accelerate the performance of these stores, researchers have proposed using RDMA technology. RDMA offers high bandwidth and low latency data access by carefully separating resource setup from IO operations, and making IO operations fast by using rich network semantics and offloading. Despite recent interest, leveraging the full potential of RDMA in a distributed environment remains a challenging task. In this paper, we present RDMA Store or RStore, a DRAM-based data store that delivers high performance by extending RDMA's separation philosophy to a distributed setting. RStore achieves high aggregate bandwidth (705 Gb/s) and close-to-hardware latency on our 12-machine testbed. We developed a distributed graph processing framework and a Key-Value sorter using RStore's unique memory-like API. The graph processing framework, which relies on RStore for low-latency graph access, outperforms state-of-the-art systems by margins of 2.6 -- 4.2× when calculating Page Rank. The Key-Value sorter can sort 256 GB of data in 31.7 sec, which is 8× better than Hadoop TeraSort in a similar setting.

DMA-capable interconnects, providing ultra-low latency and high bandwidth, are increasingly being used in the context of distributed storage and data processing systems. However, the deployment of such systems in virtualized data centers is currently inhibited by the lack of a flexible and high-performance virtualization solution for RDMA network interfaces.

Remote Procedure Call (RPC) has been the cornerstone of distributed systems since the early 80s. Recently, new classes of large-scale distributed systems running in data centers are posing extra challenges for RPC systems in terms of scaling and latency. We find that existing RPC systems make very poor usage of resources (CPU, memory, network) and are not ready to handle these upcoming workloads.

Network latency has become increasingly important for data center applications. Accordingly, several efforts at both hardware and software level have been made to reduce the latency in data centers. Limited attention, however, has been paid to network latencies of distributed systems running inside an application container such as the Java Virtual Machine (JVM) or the .NET runtime.

The performance of large-scale data-intensive applications running on thousands of machines depends considerably on the performance of the network. To deliver better application performance on rapidly evolving high-bandwidth, low-latency interconnects, researchers have proposed the use of network accelerator devices. However, despite the initial enthusiasm, translating network accelerator's capabilities into high application performance remains a challenging issue.

Work on delay tolerant networks (DTN) and peer-to-peer replication systems has made rapid advances with the similar goal of permitting reliable message delivery in challenged communication environments. Techniques developed for both types of systems also bear some similarity. Both exploit opportunistic connectivity to route messages and updates to their desired destinations while making minimal assumptions about end-to-end connectivity. However, they also have some unique characteristics. DTNs have recently incorporated forwarding algorithms that make use of historical information on past encounters between participants and predictions of future connectivity. Modern replication systems utilize protocols with low overheads that guarantee eventual consistency and at-most-once delivery while supporting content-based filters. In this paper, we show how a DTN-like messaging system can be readily built as a simple application on top of a peer-to-peer replication platform. To reduce delivery delays while retaining the desirable replication guarantees, we then extend the replication substrate to permit pluggable DTN routing protocols. We describe the implementation of four representative DTN schemes as replication policies and evaluate these extensions with emulations driven by traces of e-mail messaging and vehicular mobility. We conclude that DTNs and replication systems can benefit substantially from a cross-fertilization of ideas.

Docx2Go is a new framework to support editing of shared documents on mobile devices. Three high-level requirements influenced its design -- namely, the need to adapt content, especially textual content, on the fly according to the quality of the network connection and the form factor of each device; support for concurrent, uncoordinated editing on different devices, whose effects will later be merged on all devices in a convergent and consistent manner without sacrificing the semantics of the edits; and a flexible replication architecture that accommodates both device-to-device and cloud-mediated synchronization. Docx2Go supports on-the-go editing for XML documents, such as documents in Microsoft Word and other commonly used formats. It combines the best practices from content adaptation systems, weakly consistent replication systems, and collaborative editing systems, while extending the state of the art in each of these fields. The implementation of Docx2Go has been evaluated based on a workload drawn from Wikipedia.

This demonstration presents AdSocial, a software platform supporting social network applications in ad hoc networks. AdSocial targets small-scale scenarios such as friends playing a game on the train or co-workers sharing calendar information. Moreover, AdSocial is specifically designed to run on resource-constrained mobile devices, such as mobile phones. By using a simple and efficient data piggybacking mechanism, AdSocial applications share data by using any of the many existing routing protocols for ad hoc networks and without requiring any modification to the protocols themselves. The goal of this demonstration is to show the functionality that AdSocial can support with a very low overhead in an ad hoc network of 10--15 Nokia N810 handhelds. Conference participants will be able to establish voice-video calls, chat, or play games while moving around thus configuring a mobile and multi-hop ad hoc network.

In an effort to better understand connectivity and capacity in wireless networks, the log-normal shadowing radio propagation model is used to capture radio irregularities and obstacles in the transmission path. Existing results indicate that log-normal shadowing results in higher connectivity and interference levels as shadowing (i.e., the radio irregularity) increases. In this paper we demonstrate that such a behavior is mainly caused by an unnatural bias of the log-normal shadowing radio propagation model that results in a larger transmission range as shadowing increases. To avoid this effect, we analyze connectivity and interference under log-normal shadowing using a normalization that compensates for the enlarged radio transmission range. Our analysis shows that log-normal shadowing still improves the connectivity of a wireless network and even reduces interference. We explain this behavior by studying in detail what network parameters are affected by shadowing. Our results indicate that, when it comes to connectivity and interference, an analysis based on a circular transmission range leads to worst case results.

Recent work on QoS in Mobile Ad Hoc Networks (MANETs) has shown that bandwidth needs to be reserved in a distributed manner by also including interfering nodes. However, in practice, QoS protocols have difficulties to locally determine the set of nodes that actually interfere with a given transmission. To solve this problem, it is not uncommon to consider all nodes within a distance of k hops (k-neighbors) as interfering nodes. In this paper we use Monte-Carlo methods to study the correlation between the k-neighborhood of a node and the set of interfering nodes. We compute expected values for both reservation recall -the fraction of interfering k-neighbors to all interferers -and reservation precision -the fraction of interfering k-neighbors to all k-neighbors. The two metrics reflect the quality of a reservation and the loss of resources due to over-reservation. We further investigate the impact of different physical layer properties (e.g., fading) and network settings (e.g., network density) on the quality of the reservations. Our results indicate that existing reservation techniques to ensure QoS are inadequate and that new techniques are needed to efficiently implement QoS in MANETs.

Capacity is an important property for QoS support in Mobile Ad Hoc Networks (MANETs) and has been extensively studied. However, most approaches rely on simplified models (isotropic radio propagation, unidirectional links, perfect scheduling, perfect routing, etc.) and either provide asymptotic bounds or are based on integer linear programming solvers. In this paper we present a probabilistic approach to capacity calculation by linking the normalized throughput of a communication pair in an ad hoc network to the connection probability of the two nodes in a so called schedule graph GT(N,E). The effective throughput of a random network is modelled as a random variable and expected values of it are computed using Monte-Carlo methods. A schedule graph GT(N,E) for a given network directly emerges from the physical properties of the network, like node distribution, radio propagation or channel assignment. The modularity of the approach allows for capacity analysis under more realistic network models. In the paper throughput capacity is computed for various forms of network configurations and the results are compared to simulation results obtained with ns-2.

Most research on mobile ad hoc networks assumes a medium access (MAC) protocol similar to 802.11. Nevertheless, such networks can be built on top of any medium access scheme. An example are Bluetooth multi-hop networks (so called scatternets). In practice, many devices support several interfaces although the protocol stack used in each interface tends to be interface specific at the lower layers. This limits the ability of a device to switch back and forth between networks as need and opportunity dictates. As a step towards providing a more flexible handover infrastructure, in this paper we address the issue of integrating heterogeneous, mobile ad-hoc networks that use different MAC layer protocols. The goal is to provide an end-to-end communication abstraction that hides heterogeneity. In the paper, we discuss different possible designs with regard to application transparency, performance and mobility. Using these ideas, we describe an IP based heterogeneous mobile ad hoc testbed using Bluetooth and IEEE 802.11 that implements a virtual interface approach as the end-to-end abstraction. Furthermore, we also provide a detailed evaluation of the system in terms of throughput and handover-time.

Connectivity is an important property for QoS support in mobile ad hoc networks (MANETs). Recently, there has been a big effort in exploring the critical transmission range (CTR) analytically, based on different network models. While most of these studies rely on a geometric model and come up with asymptotic bounds, their significance regarding finite 802.11 based MANETs is unclear. In this paper, we investigate connectivity in MANETs from a layered perspective. We first point out how the transmission range affects the end-to-end connection probability in a log-normal shadowing model and compare the results to theoretical bounds and measurements in the path loss model. We then show how connectivity issues behave in 802.11 and IP based networks if the fading effect increases. The paper concludes with an analytical model for the link probability in log-normal shadowing environments as a function of the number of nodes, network area, transmission range, path loss exponent and shadowing deviation.

With the increasing widespread use of wireless technologies, the need arises for QoS provisioning mechanisms for multimedia applications in wireless networks. However, to support QoS in mobile ad hoc networks (MANETs) is more challenging than in fixed and last-hop wireless access networks. QoS protocols designed for fixed network, e.g. RSVP, are not applicable in MANETs, because they cannot cope with MANET's highly dynamic topology. Existing QoS frameworks explicitly built for MANETs are not as flexible and efficient as needed. In this paper, we propose a new QoS framework for MANETs-adaptive reservation and preallocation protocol (ASAP). By using two signaling messages, ASAP provides fast and efficient QoS support while maintaining adaptation flexibility and minimizing wasted reservations. Simulation of this framework using AODV C. Perkins, E. Royer (1999) illustrates the features and performance of ASAP, and demonstrates the design concepts.