D.C. Verma, "Guaranteed Performance Communication in High Speed Networks". Ph.D. Thesis, University of California, Berkeley, CS Report 91-663. 1991.  Home/Search   Document Not in Database   Summary   Related Articles   Check

.... the minimum envelope process [3] The maximum arrival function, on the other hand, as a deterministic upper bound on the arrival traffic, must be an upper approximation to e(I) There are some traffic models such as the ( p) leaky bucket model [5] the (Xmi, X, ve, r, m,x) model [10] and others [40, 17, 23] that follows this approximation. Among all the models, the D BIND model provides a good upper bound that can closely bound e(I) for a wide variety of sources and we can easily model the D BIND traffic by a series of leaky buckets. Furthermore, the D BIND maximum arrival function is a piecewise ....

D.C. Verma, "Guaranteed Performance Communication in High Speed Networks". Ph.D. Thesis, University of California, Berkeley, CS Report 91-663. 1991.

....real time traffic in order to meet performance requirements. Several researchers investigated the problem of supporting real time communication with performance guarantees for given worst case input traffic characteristics in wide area point to point networks [9] 12] 13] 14] 23] [27]. Among these, the concept of real time channel proposed by Ferrari and Verma [9] and refined by Kandlur, Shin, and Ferrari [13] is the most notable in explicitly addressing the problem of meeting delivery deadlines in wide area point to point networks. A real time channel is a unidirectional ....

....the data link network layer which is responsible for implementing network communication with performance guarantees. So, the application user layer is responsible for deriving traffic characteristics and performance requirements, and may choose to regulate user s input traffic [8] 14] 19] [27] in order to preserve existing performance guarantees. Although the exact distribution of incoming traffic may not always be available, an approximate distribution is usually not difficult to obtain off line for most real time applications. For example, during interactive playback of stored video, ....

[Article contains additional citation context not shown here]

D.C. Verma, "Guaranteed Performance Communication in High Speed Networks," PhD thesis, Univ. of California at Berkeley, 1991.

....calculations from the routing algorithm, thanks to its parallel search. 3 The Proposed Solution Approach We first describe the environment and the assumption under which our distributed route selection scheme will be developed. The underlying network is an arbitrary point to point network. As in [1, 7, 11, 12, 16], the generation of real time messages is assumed to be governed by the linear bounded model that is characterized by three parameters: maximum message size S max (bytes) maximum message rate R max (messages second) and maximum burst size B max (messages) In the linear bounded model, the number ....

D. C. Verma, Guaranteed Performance Communication in High Speed Networks, PhD thesis, University of California at Berkeley, 1991.

....if we only provide network access to the mobile users by any of the above architectures. A good wireless system should provide mobile users with all the services that are available to the users attached to the fixed network access points. One of the services, real time communications service [7, 12, 19], is of particular importance in today s high speed networks. It can be used to support many applications such as manufacturing systems management, remote monitoring, robotic control, voice and video transmission. In [1, 2, 21] the synchronous transmission capacity of the FDDI networks is shown ....

D. C. Verma, Guaranteed Performance Communications in High Speed Networks. PhD thesis, University of California at Berkeley, 1991.

....of the latter. DeSimone [7] studies a network of three queues in tandem with one cross traOEc stream and uses squared coeOEcients of variation as a measure of burstiness. He uses simulation and a simple analytic model to compare cell level and packet level FCFS and Round Robin (RR) Verma [8] has done multihop simulations, but his cross traOEc is composed of CBR streams of equal bandwidths (among other cross traOEc models) He observes bunching by measuring the change of the minimum inter cell spacing after a number of hops. He does not discuss, however, the implications for ....

D. C. Verma, Guaranteed Performance Communication in High Speed Networks. PhD thesis, University of California, Berkeley, Berkeley, CA 94720, December 1991. Report No. UCB/CSD 91/663.

.... by RCAP to the application may be interpreted as a contract: the network guarantees the performance bounds requested by the application, provided that the application obeys its traffic description at all times and that there are no failures in the network during the new channel s lifetime [20] [21]. This contract is valid until the channel is torn down. At any time, the sending application can request the status of the channel at each node. This status includes the channel state (e.g. partially established, fully established) the address of each node on the path, and the local resource ....

....marks each packet as ineligible when it arrives. If a packet was expected no earlier than time , but arrives at an earlier time, then the packet becomes eligible for transmission at time . A detailed explanation of the rate control module and the method to compute eligibility times can be found in [21]. After it becomes eligible, the packet is transferred to the RTIP scheduling module. The scheduling module is responsible for transmitting a packet within the interval after it receives the packet. Any service discipline that can provide such a guarantee can be used with RTIP. The RTIP entities ....

D. C. Verma, "Guaranteed Performance Communication in High Speed Networks," PhD dissertation, University of California at Berkeley, December 1991.

....properties possessed by Stop and Go and RCSP. For example, Jitter EDD provides per connection end to end delay and delay jitter bounds and allows the buffer space requirements to be uniformly distributed across the network. However, it is unclear how to implement an EDD scheduler efficiently. In [24], another rate controlled discipline is proposed with a First Come First Served scheduler. Though much simpler than EDD, a FCFS scheduler provides only one delay bound and thus cannot efficiently support the diverse QOS requirement in the future integrated services networks. Hierarchical Round ....

D. Verma. Guaranteed Performance Communication in High Speed Networks. PhD dissertation, University of California at Berkeley, November 1991.

....to enhance the utilization of the network by guaranteed performance traffic. If applications can tolerate certain losses of data without significantly affecting the quality, statistical services can be provided to achieve a higher average network utilization by exploiting statistical multiplexing [9, 17, 19]. Also, cooperative, consenting, high level multiplexing schemes can be used to address the tradeoffs between the quality of services offered to each individual clients and the overall utilization of the network [16] 7 Conclusion In this paper, we have showed that it is possible to provide ....

D. Verma. Guaranteed Performance Communication in High Speed Networks. PhD dissertation, University of California at Berkeley, November 1991.

....minimum cell inter arrival time. x ave The minimum average cell inter arrival time, over any time interval of length I. I The averaging interval for calculating x ave . Recent work has shown that this model accurately describes most types of rate controlled traffic expected in high speed networks [Ver91]. We assume that the user of the channel is bound to obey these restrictions on the input traffic, and the delay and buffer values calculated need only hold if these restrictions are not violated. 2.3. Fluid Approximation Note that each server is rate controlled and so will send cells from a ....

D. Verma, "Guaranteed Performance Communication in High-Speed Networks, " PhD Thesis, University of California at Berkeley, November 1991.

....by Stop and Go and RCSP. For example, Jitter EDD provides per connection end to end delay and delayjitter bounds and allows the buffer space requirements to be uniformly distributed across the network. However, it is unclear how to implement an EDD scheduler efficiently at high speeds. In [29], another rate controlled discipline is proposed with a First Come First Served scheduler. Though much simpler than EDD, a FCFS scheduler provides only one delay bound and thus cannot efficiently support the diverse QOS requirement in the future integrated services networks. Hierarchical Round ....

D. Verma. Guaranteed Performance Communication in High Speed Networks. PhD dissertation, University of CA at Berkeley, November 1991.

....systems. Another example is CSIM [89] by Schwetman, a library of C routines. It includes support for co routines, generators of standard probability distributions, and routines for managing a simulated clock. These packages can be used to study wide area network behavior, as Verma did with CSIM [97], but they are also applicable to many other performance evaluation problems. Other researchers have developed more specialized simulators. NEST, by Bacon, Dupuy, Schwartz, and Yemini [27] is tailored to communication networks. REAL, by Keshav [57] is 32 built upon NEST and simulates ....

D. C. Verma, Guaranteed Performance Communication in High Speed Networks, Technical Report UCB/CSD 91/663, Berkeley, California, December, 1991. (Ph.D. Dissertation).

....and for reasoning about timing behavior of programs. Gerber, Hong and Saksena [8] provided a comprehensive uni processor design methodology for guaranteeing application end to end timing requirements. There have been several papers on multi resource scheduling. Malcom and Zhao [15] Verma [19], and Sathaye [17] explored end to end analysis of multi hop network resources. Huang and Du [10] extended the uni processor real time consumer producer paradigm and provided a multi dimensional bin packing approach to resource allocation and task scheduling. Limitations included a target platform ....

D. Verma, "Guaranteed Performance Communication in High Speed Networks," Ph.D. Thesis, University of California at Berkeley, November 1991.

....timing requirements. There have been several papers on multi resource cpu E cpu F cpu A cpu B Audio Video (solid line: non ideal saturation) dashed line: ideal saturation) a) Resource Saturation (b) Latency Figure 7: Optimized Timing Results TR A scheduling. Malcom and Zhao [12] Verma [15], and Sathaye [14] explored end to end analysis of multi hop network resources. Huang and Du [9] extended the uniprocessor real time consumer producer paradigm and provided a multi dimensional bin packing approach to resource allocation and task scheduling. Limitations included a target platform ....

D. Verma, "Guaranteed Performance Communication in High Speed Networks," Ph.D. Thesis, University of California at Berkeley, November 1991.

....cell inter arrival time. x ave The minimum average cell inter arrival time, over any time interval of length I. I The averaging interval for calculating x ave . Recent work has shown that this model accurately describes many types of rate controlled traffic expected in high speed networks [14]. We assume that the user of the channel is required to obey these restrictions on the input traffic, and the delay and buffer values calculated need only hold if these restrictions are not violated. 3.2. Definitions Our notation draws from, and extends, that used in Reference [11] We define a ....

D. Verma, "Guaranteed Performance Communication in High-Speed Networks," PhD Thesis, University of California at Berkeley, November 1991.

....till they become eligible. Eligibility times for packets from different connections are computed using the same formula (as described in Section 3. 2) with different parameters; holding packets is equivalent to managing a set of timers; the mechanism for managing timers, which is a calendar queue [2, 25], can be shared by all regulators. Figure 3 shows the proposed implementation. Each of the real time queues is implemented as a linked list. The rate controller is implemented using a modified version of a calendar queue. A calendar queue consists of a clock and a calendar, which is a pointer ....

D. Verma. Guaranteed Performance Communication in High Speed Networks. PhD dissertation, University of California at Berkeley, November 1991.

.... and do not depend on the node s position along the route of a channel (except for the first node, whose input jitter is zero) The number C max ,n of real time channels node n can support in this case, when the scheduling policy is J FCFS (i.e. FCFS preceded by a jitter control regulator [Verm91]) satifies the inequalities C max ,n # # # x min 2d max ,n dp n ############ # # # 1 C max ,n C max ,n . 15) If we are not allowed to make all of the above assumptions, the admission criteria and the expression of d max ,n are more complicated, but it is still true that the ....

Verma, D. C., "Guaranteed Performance Communication in High Speed Networks," Ph.D. Dissertation, University of California at Berkeley, November 1991. - 22 -

....cell inter arrival time. x ave The minimum average cell inter arrival time, over any time interval of length I. I The averaging interval for calculating x ave . Recent work has shown that this model accurately describes many types of rate controlled traffic expected in high speed networks [14]. We assume that the user of the channel is required to obey these restrictions on the input traffic, and the delay and buffer values calculated need only hold if these restrictions are not violated. 3.2. Definitions Our notation draws from, and extends, that used in Reference [11] We define a ....

D. Verma, "Guaranteed Performance Communication in High-Speed Networks," PhD Thesis, University of California at Berkeley, November 1991.

....accepts the client s request, it guarantees that the specified performance requirements will be met provided that the client obeys its traffic specification. In this model, the guaranteed performance service defines a contractual relationship between the communication client and the network [13] [55], 15] the network promises to fulfill its obligation (guaranteeing the performance for the client s traffic) only if the client honors its own part of the contact (not sending more data than declared) In addition, the network may reject the client s request due to lack of resources or ....

D. Verma. Guaranteed Performance Communication in High Speed Networks. PhD dissertation, University of California at Berkeley, November 1991.

....multiservice packet switching networks must provide guarantees on throughput, delay, delay jitter, and error rate. In network architectures that provide service guarantees [3, 8, 12] the relationship between network clients and the network provider can be defined in terms of a traffic contract [23]. When requesting a connection with service guarantees, the client submits a specification of its traffic together with a set of desired guarantees. The network performs admission control tests to verify that the requested service can be given without violating any previously given guarantees. ....

D. C. Verma. Guaranteed Performance Communication in High Speed Networks. PhD thesis, University of California - Berkeley, November 1991.

....problem can be solved by a simple table look up at the source node. 3 The Proposed Solution Approach We first describe the environment and the assumption under which our distributed routeselection scheme will be developed. The underlying network is an arbitrary point to point network. As in [1, 7, 10, 11, 15, 18], the generation of real time messages is assumed to be governed by a linear bounded model that is characterized by three parameters: maximum message size Smax (bytes) maximum message rate R max (messages second) and maximum burst size B max (messages) In the linear bounded model, there are two ....

D. C. Verma, Guaranteed Performance Communication in High Speed Networks, PhD thesis, University of California at Berkeley, 1991.

.... of a channel is analogous to the signing of a contract: the network guarantees the performance bounds requested by the application, provided that the application obeys its traffic description at all times and that there are no failures in the network during the channel s lifetime [22] [23]. This contract is valid until the channel is torn down. Finally, either the source or the destination of a given channel can request that it be closed, using the close request forward and close request reverse messages. They are transmitted hop by hop along the channel route and cause the ....

D. C. Verma, "Guaranteed Performance Communication in High Speed Networks," PhD dissertation, University of California at Berkeley, December 1991.

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