On Local CAC Schemes for Scalability of High-speed Networks (original) (raw)
Related papers
An Efficient Reservation Connection Control Protocol for Gigabit Networks
Computer Networks and Isdn Systems, 1998
. The Efficient ReserÕation Virtual Circuit protocol or ERVC is a novel connection control protocol designed for constant-rate, delay-insensitive traffic in gigabit networks. We explain the operation of the protocol, discuss its features and advantages, and present its performance characteristics. The ERVC protocol is appropriate for sessions that require an explicit reservation of capacity and can tolerate the round-trip delay associated with the reservations. In the ERVC protocol, the durations of the sessions are recorded, and every node keeps track of the utilization profile of each outgoing link, which describes the amount of residual capacity available on the link as a function of time. This feature allows capacity to be reserved only for the duration of the session, starting at the time it is actually needed. Therefore, the protocol utilizes capacity considerably more efficiently than regular reservation schemes do and results in markedly lower blocking probability for new sessions. The ERVC protocol also has the ''reservation ahead'' feature, which allows a node to calculate the time at which the requested capacity will be available and reserve it in advance, avoiding in this way the wasteful repetition of the call setup phase. q 1998 Elsevier Science B.V. All rights reserved. . changed communication environment. As Partridge w x 2 in his series of thought questions on the challenges of gigabit networking rightly points out, high bandwidth-delay product and increased propagation latency are two crucial factors that will differentiate gigabit networks from most present day networks, and that will impact not only the performance of such networks, but also the protocols designed to manage them and the applications designed to use them.
Fast Admission Control for Short TCP Flows
2009
Over the last decade, numerous admission control schemes have been studied to allocate network resources. Although per-flow control schemes can provide guaranteed QoS, such schemes face scalability issues in large networks due to the tremendous number of flows present. While aggregation-based approaches such as Differentiated Services relieve the storage of state in the core, admission control of flows, especially short-lived flows, is still a serious bottleneck. To that end, we propose an admission control scheme, Fast Admission for Short Flows (FASF), that enables accelerated admission control at the edge rather than via centralized or in-path mechanisms. FASF not only reduces the burden on admission control by largely distributing the dominant resource requests (i.e. short-lived flows), but also improves flow completion time and hence network goodput.
Distributed admission control without knowledge of the capacity region
2013 Proceedings IEEE INFOCOM, 2013
We consider the problem of distributed admission control without knowledge of the capacity region in single-hop wireless networks, for flows that require a pre-specified bandwidth from the network. We present an optimization framework that allows us to design a scheduler and resource allocator, and by properly choosing a suitable utility function in the resource allocator, we prove that existing flows can be served with a prespecified bandwidth, while the link requesting admission can determine the largest rate that it can get such that it does not interfere with the allocation to the existing flows.
Analysis of a rate-based access control mechanism for high-speed networks
IEEE Transactions on Communications, 1993
In this paper we study a rate-based access control mechanism for high-speed networks, based on the "buffered leaky bucket" scheme. The study is carried out in discrete time and assumes an ATM environment, where only fixed size packets or cells are transmitted over the network. Since cell arrivals often correspond to the segmentation of large user packets into many cells, or the superposition of a number of cells coming from different logical connections routed over the same path in the network, we consider a batch arrival process to represent such arrival patterns. With batch arrival processes, the impact of other factors such as burstiness of the source and packet length variations can be effectively captured. The distribution of batch sizes is arbitrary, while the time between arrivals of successive batches is taken to be geometrically distributed. The analysis is based on matrix analytic techniques, but the particular structure of the system allows for major simplifications. In particular, we show that the problem can be partitioned, hence greatly reducing its dimensionality. We also show that the standard iterative approach required by the matrix analytic solution method can be replaced by an exact recursive procedure. This greatly increases the range of systems that can be handled by the method. From these results, we obtain both the queue length distribution at the access point and the distribution of the interdeparture time from the leaky bucket. A number of numerical examples are provided that illustrate the influences of the various system parameters.
A Novel Scheme for Effective Number of Users in Intranet Congestion Control
An end-to-end congestion control, the current Internet suffers from two maladies: Congestion collapse from undelivered packets, and unfair allocations of bandwidth between competing traffic flows. The first malady — congestion collapse from undelivered packets. The second malady—unfair bandwidth allocation to competing network flows. Adaptive applications (e.g., TCP-based applications) that respond to congestion by rapidly reducing their transmission rates are likely to receive unfairly small bandwidth allocations when competing with unsympathetic applications. The Internet protocols themselves can commence unfairness. The TCP algorithm, for illustration, intrinsically causes every TCP flow to accept a bandwidth that is inversely proportional to its round-trip time. Hence, TCP connections with short round-trip times may receive unfairly large allocations of network bandwidth when compared to connections with longer round-trip times. In this paper, enhanced congestion control mechanism for packet retransmission is used to rectify collapse in network.
CLAMP: differentiated capacity allocation in access networks
Conference Proceedings of the 2003 IEEE International
Absffuct-This paper presents a solution for providing differentiated capacity allocation in an access network. The system is based on CLAMP, an algorithm that can differentiate between flows sharing the same FIFO queue. The system is suitable for access networks, such as those based on DSL and HFC modems and wireless LAN access points. The deployment of CLAMP is completely contained within the access network; no changes to the remainder of the network are required. CLAMP provides the opportunity to enforce local policies on TCP flows that originate from sources distributed globally. The performance of CLAMP is verified by both simulation and analysis.
Journal of the Brazilian Computer Society, 2001
Multimedia networks will support a wide range of applications with different requirements and traffic characteristics. Connection Admission Control (CAC) algorithms are used to decide whether an incoming connection should be accepted or rejected in order to maintain the quality of service (QoS) demanded by the applications. The objective of this work is to present an environment we developed useful for performance analysis, measurements and experimentation, in particular for testing resource usage based on different traffic characteristics. We demonstrate how a study of the effectiveness of different CAC algorithms can be performed from measurements collected using an ATM switch and the tools we developed. For the studies we selected and implemented two CAC algorithms, one for a non-regulated traffic, proposed by , and other for a leaky-bucket regulated traffic, proposed by . The environment tools include a traffic generator supporting IP and native-ATM as well as a CAC module that implements the algorithms above and can be used in conjunction with our test environment. These tools are currently part of the Tangram-II modeling environment and available to download.
LGC: An Active Congestion Control Mechanism
2000
This paper presents the design and evaluation of Limiting Greedy Connections (LGC), an active congestion control mechanism for minimizing the degradation in network performance caused by bandwidth greedy applications. The primary objectives of the LGC mechanism are to limit the impact of greedy connections on a congested node, to keep a loose upper bound on the packet queue occupancy at the intermediate nodes of the network, and to minimize packet loss. The LGC mechanism is evaluated for a variety of network topologies, transmitting sources, and node queue parameters, using a Java-based active network test bed.
Statistical Connection Admission Control Framework based on Achievable Capacity Estimation
2006 IEEE International Conference on Communications, 2006
Traditional traffic descriptor-based and measurementbased admission control schemes are typically combined with a node by node resource reservation scheme, rendering them unscalable. Although some Endpoint Admission Control schemes can resolve this problem, they impose significant signaling overhead. To cope with these two problems, this paper proposes a statistical connection admission control framework which can easily and efficiently estimate the network resource for a pair of ingress-egress nodes and make admission decision based on this estimated result. In this framework, the network is considered as a "black box." For a certain ingress-egress node pair, the egress node measures the QoS constraint violation ratio and feeds this information back to the ingress node periodically. With this information and the measured statistical characteristics of the existing aggregated traffic, the ingress node estimates the achievable capacity between the ingress-egress node pair, and makes the admission decision for a new traffic connection request. The signaling overhead of this framework is very small. Simulation results show the effective throughput is relatively high.