Today’s Internet Service Providers (ISPs) serve two roles: managing their network infrastructure and providing (arguably limited) services to end users. We argue that coupling these roles impedes the deployment of new protocols and architectures, and that the future Internet should support two separate entities: infrastructure providers (who manage the physical infrastructure) and service providers (who deploy network protocols and offer end-to-end services). We present a high-level design for Cabo, an architecture that enables this separation; we also describe challenges associated with realizing this architecture.

WiFi-based Long Distance (WiLD) networks with links as long as 50–100 km have the potential to provide connectivity at substantially lower costs than traditional approaches. However, real-world deployments of such networks yield very poor end-to-end performance. First, the current 802.11 MAC protocol has fundamental shortcomings when used over long distances. Second, WiLD networks can exhibit high and variable loss characteristics, thereby severely limiting end-to-end throughput. This paper describes the design, implementation and evaluation of WiLDNet, a system that overcomes these two problems and provides enhanced end-to-end performance in WiLD networks. To address the protocol shortcomings, WiLDNet makes several essential changes to the 802.11 MAC protocol, but continues to exploit standard (low-cost) WiFi network cards. To better handle losses and improve link utilization, WiLDNet uses an adaptive loss-recovery mechanism using FEC and bulk acknowledgments. Based on a real-world deployment, WiLDNet provides a 2–5 fold improvement in TCP/UDP throughput (along with significantly reduced loss rates) in comparison to the best throughput achievable by conventional 802.11. WiLDNet can also be configured to adapt to a range of end-to-end performance requirements (bandwidth, delay, loss). 1

The ever increasing demand of new applications has led researchers to propose new network architectures that address limitations of the current Internet. Given the rigidity of the Internet today, overlay networks are used to implement such architectures, in the hope of gaining a large user base. Despite sustained efforts to test and deploy new network architectures (on testbeds such as Planetlab), few of these efforts have attracted a significant number of users. We believe that chances of user acceptance of overlays, and eventually new network architectures, will be substantially improved by enabling users to leverage their functionality without any modifications to their applications and operating systems. In this paper, we present our design, implementation, and experience with OCALA, an overlay convergence architecture that achieves this goal. OCALA interposes an overlay convergence layer below the transport layer, that is composed of an overlay independent sub-layer that interfaces with legacy applications, and an overlay dependent sub-layer that delivers packets to the overlay. Unlike previous efforts, this design enables: (a) simultaneous access to multiple overlays (b) communication between hosts in different overlays (c) communication between overlay hosts and legacy hosts (d) extensibility, allowing researchers to incorporate their overlays into OCALA. We currently support three overlays, i3 [29], RON [1] and HIP [17], on Linux and Windows XP/2000. We (and a few other research groups and end-users) have used OCALA for over a year with many legacy applications ranging from web browsers to remote desktop applications.

The cost savings and novel features associated with Voice over IP (VoIP) are driving its adoption by service providers. Such a transition however can successfully happen only if the quality and reliability offered is comparable to the existing PSTN. Unfortunately, the Internet’s best effort service model provides no inherent quality of service guarantees. Because low latency and jitter is the key requirement for supporting high quality interactive conversations, VoIP applications use UDP to transfer data, thereby subjecting themselves to performance degradations caused by packet loss and network failures. In this paper we describe two algorithms to improve the performance of such VoIP applications. These mechanisms are used for localized packet loss recovery and rapid rerouting in the event of network failures. The algorithms are deployed on the routers of an application-level overlay network and require no changes to the underlying infrastructure. Initial experimental results indicate that these two approaches can be composed to yield voice quality on par with the PSTN.

known bottleneck links like enterprise access links or wireless links. A more comprehensive, end-to-end QoS deployment, for instance across large enterprise networks or the global Internet, remains elusive. There is growing interest in the idea of using overlay networks to provide differential QoS services (improve performance for some flows at the expense of other flows). A necessary building block is the ability to provide differential service over a single overlay link that traverses many IP router hops. This paper presents MPAT, the first truly scalable algorithm for fairly providing differential services to TCP flows that share a bottleneck link. Unlike known schemes, our approach preserves the cumulative fair share of the aggregated flows even where the number of flows in the aggregate is large. Specifically we demonstrate, primarily through experiments on the real Internet, that congestion state can be shared across more than 100 TCP flows with throughput differentials of 95:1. This is up to five times better than differentials achievable by known techniques. Indeed, MPAT scalability is limited only by the delay-bandwidth product of the aggregated flows. With this tool, it is now possible to seriously explore the viability of network QoS through overlay network services. 1

Providing support for legacy applications is a crucial component of many overlay networks, as it allows end-users to instantly benefit from the functionality introduced by these overlays. This paper presents the design and implementation of a proxy-based solution to support legacy applications in the context of the i3 overlay [24]. The proxy design relies on an address virtualization technique which allows the proxy to tunnel the legacy traffic over the overlay transparently. Our solution can preserve IP packet headers on an end-to-end basis, even when end-host IP addresses change, or when endhosts live in different address spaces (e.g., behind NATs). In addition, our solution allows the use of human-readable names to refer to hosts or services, and requires no changes to applications or operating systems. To illustrate how the proxy enables legacy applications to take advantage of the overlay (i.e., i3) functionality, we present four examples: enabling access to machines behind NAT boxes, secure Intranet access, routing legacy traffic through Bro, an intrusion detection system, and anonymous web download. We have implemented the proxy on Linux and Windows XP/2000 platforms, and used it over the i3 service on PlanetLab over a three month period with a variety of legacy applications ranging from web browsers to operating system-specific file sharing.

Running multiple virtual networks, customized for different performance objectives, is a promising way to support diverse applications over a shared substrate. Despite being simple, a static division of resources between virtual networks can be highly inefficient, while dynamic resource allocation runs the risk of instability. This paper uses optimization theory to show that adaptive resource allocation can be stable and can maximize the aggregate performance across the virtual networks. In the DaVinci architecture, each substrate link periodically reassigns bandwidth shares between its virtual links; while at a smaller timescale, each virtual network runs a distributed protocol that maximizes its own performance objective independently. Numerical experiments with a mix of delay-sensitive and throughputsensitive traffic show that the bandwidth shares converge quickly to the optimal values. We demonstrate that running several custom protocols in parallel and allocating resource adaptively can be more efficient, more flexible, and easier to manage than a compromise “one-size-fits-all ” design. 1.

Abstract — The cost savings and novel features associated with Voice over IP (VoIP) are driving its adoption by service providers. Unfortunately, the Internet’s best effort service model provides no quality of service guarantees. Because low latency and jitter is the key requirement for supporting high quality interactive conversations, VoIP applications use UDP to transfer data, thereby subjecting themselves to quality degradations caused by packet loss and network failures. In this paper we describe an architecture to improve the performance of such VoIP applications. Two protocols are used for localized packet loss recovery and rapid rerouting in the event of network failures. The protocols are deployed on the nodes of an application-level overlay network and require no changes to the underlying infrastructure. Experimental results indicate that the architecture and protocols can be combined to yield voice quality on par with the PSTN. I.

We propose to construct routing overlay networks using the following principle: that overlay edges should be based on mutual advantage between pairs of hosts. Upon this principle, we design, implement, and evaluate Peer-Wise, a latency-reducing overlay network. To show the feasibility of PeerWise, we must show first that mutual advantage exists in the Internet: perhaps contrary to expectation, that there are not only “haves ” and “have nots” of low-latency connectivity. Second, we must provide a scalable means of finding promising edges and overlay routes; we seek embedding error in network coordinates to expose both shorter-than-default “detour ” routes and longer-than-expected default routes. We evaluate the cost of limiting PeerWise to mutually advantageous links, then build the intelligent components that put PeerWise into practice. We design and evaluate “virtual ” network coordinates for destinations not participating in the overlay, neighbor selection algorithms to find promising relays, and relay selection algorithms to choose the neighbor to traverse for a good detour. Finally, we show that PeerWise is practical through a wide-area deployment and evaluation. 1

Data intensive applications require massive amount of data to be transported over shared wide area networks. Traditional network congestion control and routing schemes have proven inadequate for fully utilizing available network resource for high-bandwidth data transport. In this work, we explore the flexibility of control at the application layer and propose various application level data relay schemes to largely improve the data throughput by optimally integrating application level routing and transport layer control. The proposed algorithms can be easily implemented in overlay networks. Preliminary experiments have shown that our relay schemes are e#cient in utilizing network resource for high-bandwidth data transport. The impact of application level relays on the underlay network is also discussed.