I. Stoica, S. Shenker, and H. Zhang, "Core-Stateless Fair Queueing: A Scalable Architecture to Approximate Fair Bandwidth Allocations in High-Speed Networks," IEEE/ACM Transactions on Networking, vol. 11, no. 1, pp. 33--46, 2003.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....than stateless schemes and unlikely to be implemented in a router. On the other hand, it is possible to provide fair bandwidth sharing without keeping any per flow states. Such designs are called the stateless architecture. However, the services provided with current stateless architectures [12][10] have little flexibility, low utilization, and or fairness level as compared to the services that can be provided with per flow mechanisms. As a result, some existing schemes [3] 4] have been proposed to approximate fairness among flows by keeping states for some flows only. In this paper, we ....

....not required to marked Identify the number of marks among the recaptured animals Identify the number of packets has a certain flow ID among the recaptured packets C. Adjustment mechanism An important component of CARE is the rate adjustment mechanism. One possible approach would be CSFQ s [10] adjustment mechanism. Under the ideal situation, the resultant throughput of each flow should be as fair as possible. Therefore:# # # # # # # # ## # where# is the sending rate for flow and, is the ideal dropping probability for flow . Upon each packet arrival we apply the ....

I. Stoica, S. Shenker, H. Zhang, "Core-Stateless Fair Queueing: A Scalable Architecture to Approximate Fair Bandwidth Allocations in High Speed Networks," ACM SIGCOMM 1998.

....in the range of seconds to minutes; and traffic engineering, including rerouting and capacity planning, operate on slower time scales on the order of hours to months. Significant progress has been made in the area of scheduling and flow control, e.g. dynamic packet state and its derivatives [11]) In the area of traffic engineering, solutions for circuit based networks has been widely investigated in literature [10] There has been recent progress on the application of these techniques to IP routed net works [5] In contrast, in the area of dynamic provisioning, most research effort ....

.... 4q (q (K 1)o) 2 otherwise (3) The performance of the proposed node algorithm depends on the measure ment of queue length ,q, packet loss Szoss , arrival rate and departure rate rdepart for each class. We use the same form of exponentially weighted moving average function proposed in [11] to smooth the measurement samples. 3.4 Control Actions The control conditions that invoke changes to the traffic intensity (i) are as follows: 1. If .q (i) Nj up (i) reduce traffic intensity to 3(i) by either increasing service weights or reducing arrival rate by applying multipliedrive ....

I. Stoica, S. Shenker, and H. Zhang. Core-Stateless Fair Queueing: A Scalable Architecture to Approximate Fair Bandwidth Allocations in High Speed Networks. In Proc. ACM SIGCOMM, September 1998.

....at the edge of the core network. In this case, the challenge becomes gaining e#ective control with coarse granularity control information resulting from the aggregation of flow states and control mechanisms. Most of current quantitative di#erentiated service results are based on packet scheduling [95, 96, 27, 71, 48] and admission control techniques [21, 17, 44, 59] In contrast, we present a dynamic bandwidth provisioning framework for quantitative service di#erentiation. Our scheme comprises a pair of node and core provisioning algorithms. The node provisioning algorithm prevents transient violations of ....

....constraints based on time scale and bandwidth granularity. Our Di#Serv provisioning work is the first to demonstrate quantitative service di#erentiation for tra#c aggregates. The state of the art at the time was either quantitative performance guarantee for individual flow (Stoica s CSFQ work [95, 96]) with the packet header carrying the control state or qualitative di#erentiation for tra#c aggregates (Dovrolis work of proportional Di# Serv [27] We realize this by dynamically changing the service weights of an o# the shelf per class weighted scheduler, with an extended virtual queue ....

[Article contains additional citation context not shown here]

I. Stoica, S. Shenker, and H. Zhang. Core-Stateless Fair Queueing: A Scalable Architecture to Approximate Fair Bandwidth Allocations in High Speed Networks. In Proc. ACM SIGCOMM, September 1998.

....Randomized algorithms are typically simple to implement and can often offer a simple alternative to deterministic algorithms. Indeed, in the last few years randomized algorithms have been discussed in the context of congestion avoidance (RED [FJ93] and approximation of bandwidth allocation (CSFQ [I. 98] Our focus in this work is randomized algorithms for fair bandwidth allocation which are simple, practical and robust. As we discuss later, we use randomization to ensure that on the average the bandwidth allocation is fair. Deterministic approaches to packet fair queuing have evolved from the ....

....fairness and delay guarantee, of the more complex fair queuing algorithms. In fact, only a few variants of WFQ were deployed due to the complexity of their implementation. Two alternative approaches to approximated fair bandwidth allocation are Core Stateless Fair Queueing (CSFQ, described in [I. 98] and Rainbow Fair Queueing (RFQ, which was discussed in [CWZ00] CSFQ attempts to achieve an approximate bandwidth allocation without maintaining a per flow state. It does so by dividing the network into regions and distinguishing between edge routers and core routers. Edge routers maintain per ....

I. Stoica and S. Shenker and H. Zhang. Core-Stateless Fair Queueing: A scalable architecture to approximate fair bandwidth allocations in high speed networks. In ACM SIGCOMM, August 1998.

....extend to active nodes since allocation of resources in active networks involves more than one resource such as, CPU, bandwidth, and memory. Moreover, the allocation of the three is interdependent. Thus fair allocation of one does not guarantee fair allocation of other. Some work has been done [5,6,7,8] on scheduling network resources but those allocate resource for a single resource. This does not directly extend to active networks. It has been identified that CPU requirement of active packets on multiple platforms cannot be determined a priori and it is a major obstacle in managing CPU ....

Ion Stoica, Scott Shenker, and Hui Zhang. Core-stateless Fair queueing: A Scalable Architecture to Approximate Fair Bandwidth Allocations in High Speed Networks.", SIGCOMM '98.

....to their reception TCP friendly reception rates computed using Eq. 1. For both cases, the source tcp friendly throughput rate can be included in the IP packet header by the multicast source or the source s Designed Router. This technique is largely used by many other mechanisms such as CSFQ [15] for different purposes. Thus, a SBQ intermediate router gets the rates from the IP multicast packet headers and computes their aggregated value according to Eq. 2. 3.3 Weights Updating Time In this sub section, we try to answer the question: How often we update the weights In other words, ....

I. Stoics, S. Shenker, and H. Zhang, Core-Stateless Fair Queueing: A Scalable Architecture to Approximate Fair Bandwidth Allocations in High Speed Networks, In Proc. of SIGCOMM'98.

....conditions necessary for obtaining accurate and high confidence inferences. I. INTRODUCTION Both research and commercial networks are increasingly able to provide minimum quality of service levels to traffic classes, e.g. 20] Example components of such networks include QoS schedulers [10] [18], diffserv style service level agreements [4] 8] 14] 21] 25] edge based traffic shaping and prioritizing devices, and novel architectures and algorithms for scalable QoS management [6] 7] 18] 19] 25] However, even as the network s infrastructure and services become increasingly ....

.... traffic classes, e.g. 20] Example components of such networks include QoS schedulers [10] 18] diffserv style service level agreements [4] 8] 14] 21] 25] edge based traffic shaping and prioritizing devices, and novel architectures and algorithms for scalable QoS management [6] 7] [18], 19] 25] However, even as the network s infrastructure and services become increasingly sophisticated, the network s clients lack reciprocal tools for validation and monitoring of the network s QoS capabilities. Clients of Service Level Agreements (SLAs) will have monitoring requirements ....

I. Stoica, S. Shenker, and H. Zhang. Core-Stateless Fair Queueing: A scalable architecture to approximate fair bandwidth allocations in high speed networks. In Proceedings of ACM SIGCOMM '98, Vancouver, British Columbia, September 1998.

....measurement, multiclass, statistical envelopes. 1INTRODUCTION OTH research and commercial networks and Web servers are increasingly able to provide minimum quality of service levels to traffic and application classes, e.g. 1] Example components of such networks include QoS schedulers [2] [3], diffserv style service level agreements [4] 5] 6] 7] edge based traffic shaping and prioritizing devices, and novel architectures and algorithms for scalable QoS management [8] 9] 10] Similar resource management mechanisms, request scheduling policies, and algorithms are also ....

I. Stoica, S. Shenker, and H. Zhang, "Core-Stateless Fair Queueing: A Scalable Architecture to Approximate Fair Bandwidth Allocations in High Speed Networks," Proc. ACM SIGCOMM '98, Sept. 1998.

....sharing at both the flow level and the class level. For example, if a class is provided a circuit like service without sharing among traffic classes, the service envelope will measure a simple linear function. In contrast, if the black box performs weighted fair queueing like scheduling [24] [25], the service envelope will reflect the available capacity beyond the minimum guaranteed rate which can be exploited by the class, i.e. the excess capacity which is available due to fluctuating resource demands of cross traffic and other traffic classes. Finally, by limiting a class traffic by ....

....experiments of Section V. In particular, for a link with 10 Mb s capacity and the same 50 Pareto on off sources as in Fig. 2, the figure depicts the minimum empirical service versus interval length, that is, versus . For the figure, the Core Stateless Fair Queuing (CSFQ) service discipline [25] is employed in the backbone network. Notice that the envelope has a roughly convex shape indicating that the service rate is increasing with interval length. Moreover, the variation in the envelope s slope is due to variations in available capacity due to the burstiness of other traffic flows. In ....

[Article contains additional citation context not shown here]

I. Stoica, S. Shenker, and H. Zhang, "Core-stateless fair queueing: A scalable architecture to approximate fair bandwidth allocations in high speed networks," in Proc. ACM SIGCOMM '98, Vancouver, BC, Canada, September 1998.

....as at the hops that the connection will traverse. An acceptance triggers actual configuration of edge device, but only resource state updates of core routers interfaces in the Service Broker database. Other works that propose guaranteed services without per flow provisioning at core are: SZ99] SSZ98] CK00] ZDGH00] However, all of them consider short lived flows while VPN connections in our case are usually rate controlled long lived flows that are often provisioned for larger time scale. The centralized SB in its role as a global network manager maintains information about all the ....

....our virtual core provisioning approach does not su#er from the same problem since capacity reservation states are actually stored in a Service Broker based inventory and not in the core routers. Other works that propose guaranteed services without per flow provisioning at core are: SZ99] SSZ98] CK00] ZDGH00] BV01] LC01] CB01] However, all of them consider short lived flows while VPN connections in our case are usually rate controlled long lived flows that are often provisioned for larger time scale. Some notable works [FBP 01] JD02] QK01] KS99] on bandwidth ....

Ion Stoica, Scott Shenker, and Hui Zhang. Core-Stateless Fair Queueing: A Scalable Architecture to Approximate Fair Bandwidth Allocations in High Speed Networks. SIGCOMM'98 Conference, 1998.

No context found.

I. Stoica, S. Shenker, and H. Zhang, "Core-Stateless Fair Queueing: A Scalable Architecture to Approximate Fair Bandwidth Allocations in High-Speed Networks," IEEE/ACM Transactions on Networking, vol. 11, no. 1, pp. 33--46, 2003.

No context found.

Stoica, I., Shenker, S., Zhang, H.: Core-Stateless Fair Queueing: a Scalable Architecture to Approximate Fair Bandwidth Allocations in High-speed Networks. IEEE/ACM Trans. on Networking 11 (2003)

No context found.

I. Stoica and S. Shenker and H. Zhang, "Core-Stateless Fair Queueing: A scalable architecture to approximate fair bandwidth allocations in high speed networks," in ACM SIGCOMM, August 1998.

No context found.

I. Stoica, S. Shenker, and H. Zhang. Core-Stateless Fair Queueing: a Scalable Architecture to Approximate Fair Bandwidth Allocations in High Speed Networks. In Proc. ACM SIGCOMM, 1998. 26

No context found.

Ion Stoica, Scott Shenker, and Hui Zhang. Corestateless fair queueing: A scalable architecture to approximate fair bandwidth allocations in high speed networks. In SIGCOMM '98, 1998.

No context found.

I. Stoica, S. Shenker, and H. Zhang, "Core-Stateless Fair Queueing: A Scalable Architecture to Approximate Fair Bandwidth Allocations in High Speed Networks," in Proceedings of ACM SIGCOMM, September 1998.

No context found.

I. Stoica, S. Shenker, and H. Zhang, "Core-Stateless Fair Queueing: A Scalable Architecture to Approximate Fair Bandwidth Allocations in High Speed Networks," in Proceedings of ACM SIGCOMM, September 1998.

No context found.

I. Stoica, S. Shenker, and H. Zhang, Core-Stateless Fair Queueing: A Scalable Architecture to Approximate Fair Bandwidth Allocations in High Speed Networks, In Proc. of SIGCOMM'98.

No context found.

I. Stoica, S. Shenker, and H. Zhang. Core-Stateless Fair Queueing: A Scalable Architecture to Approximate Fair Bandwidth Allocations in High Speed Networks. In Proceedings ACM SIGCOMM'98, September 1998.

No context found.

I. Stoica, S. Shenker, and H. Zhang. Core-Stateless Fair Queueing: a Scalable Architecture to Approximate Fair Bandwidth Allocations in High-speed Networks. IEEE/ACM Trans. on Networking, 11(1), 2003.

No context found.

I. Stoica, S. Shenker, and H. Zhang, "Core-Stateless Fair Queueing: A Scalable Architecture to Approximate Fair Bandwidth Allocations in High Speed Networks," in Proc. Annual Conference of the ACM Special Interest Group on Data Commun. (SIGCOMM), September 1998.

No context found.

I. Stoica, S. Shenker, and H. Zhang. Core-stateless fair queueing: A scalable architecture to approximate fair bandwidth allocations in high-speed networks. IEEE/ACM Transactions on Networking, 11(1):33--46, January 2003.

No context found.

I. Stoica, S. Shenker, and H. Zhang. Core-Stateless Fair Queueing: A scalable architecture to approximate fair bandwidth allocations in high speed networks. In Proceedings of ACM SIGCOMM '98, Vancouver, British Columbia, September 1998.

No context found.

I. Stoica, S. Shenker, and H. Zhang. Core-Stateless Fair Queueing: A Scalable Architecture to Approximate Fair Bandwidth Allocations in High Speed Networks. In Proc. ACM SIGCOMM, September 1998. This article was processed using the L A T E X macro package with LLNCS style

No context found.

Ion Stoica, Scott Shenker, Hui Zhang, "Core-Stateless Fair Queueing: A Scalable Architecture to Approximate Fair Bandwidth Allocations in High Speed Networks", SIGCOMM, 1998.

First 50 documents  Next 50

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC