Che92b. Chen, B., G. Agrawal and W. Zhao, "Optimal Synchronous Capacity Allocation for Hard Real-Time Communications with the Timed Token Protocol", Proceedings of the 13th IEEE Real-Time Systems Symposium, pp. 198-207 (December 1992).  Home/Search   Document Details and Download   Summary   Related Articles   Check

....The timing properties of the timed token protocol were first formally analyzed by Johnson and Sevcik in [4, 5] where it is shown that the average token rotation time is bounded by the Target Token Rotation Time (TTRT ) and the maximum token rotation time cannot exceed twice the TTRT . Chen et al. [1, 6, 7, 8] made a detailed study on the timing behavior of the timed token protocol and generalized the upper bound derived by Johnson and Sevcik on the maximum token rotation time. That is, they extended the upper bound on the time possibly elapsed between any two successive token s arrivals at a node ....

....time) to between any v (v is a positive integer no less than two) successive token s arrivals at a node. Their general result is important for studies on real time communications in any network where the timed token protocol is employed, and has already been used extensively by many researchers [1, 8, 9, 10, 11, 12, 13, 14, 15, 16] in studying (analyzing) various kinds of synchronous bandwidth allocation (SBA) schemes. Unfortunately, this general upper bound derived by Chen et al., although very important, 1 may not keep tight when v grows large enough, and consequently the SBA schemes previously developed and analyzed ....

[Article contains additional citation context not shown here]

B. Chen, G.Agrawal, and W.Zhao, "Optimal synchronous capacity allocation for hard realtime communications with the timed token protocol," in Proc. of the 13th IEEE Real-Time Systems Symp., pp. 198--207, Dec. 1992.

.... synchronous messages before their deadlines, synchronous bandwidths must be carefully allocated to network nodes such that not only the protocol constraint (of the timed token network) can be met but also the deadline constraint (of the considered synchronous message set) can be satis ed as well [3]. Three optimal synchronous bandwidth allocation (SBA) schemes [3] 4] 5] have been proposed for guaranteeing synchronous messages with their deadlines equal to periods, using the timed token protocol. A SBA scheme is optimal if it can always guarantee a synchronous message set whenever ....

.... must be carefully allocated to network nodes such that not only the protocol constraint (of the timed token network) can be met but also the deadline constraint (of the considered synchronous message set) can be satis ed as well [3] Three optimal synchronous bandwidth allocation (SBA) schemes [3], 4] 5] have been proposed for guaranteeing synchronous messages with their deadlines equal to periods, using the timed token protocol. A SBA scheme is optimal if it can always guarantee a synchronous message set whenever there exists a SBA scheme that can do so [3] Based on their derived ....

[Article contains additional citation context not shown here]

B. Chen, G.Agrawal and W.Zhao, \Optimal synchronous capacity allocation for hard real-time communications with the timed token protocol," in Proc. IEEE RTSS'92, Dec. 1992, pp. 198-207.

....this assumption may not match what can really happen in reality in the worst case for a specific synchronous message set to be guaranteed. As a result of such an unrealistic worst case assumption, a pessimistic result on calculation of the worst case available transmission time has been derived [3, 6, 7], which may further lead to a wrong judgement on whether or not a given allocation of synchronous bandwidths can guarantee the synchronous message set considered [8, 9] and may even cause a previously claimed optimal synchronous bandwidth allocation (SBA) scheme to perform non optimally in ....

B. Chen, G.Agrawal, and W.Zhao, "Optimal synchronous capacity allocation for hard real-time communications with the timed token protocol," in Proc. IEEE Real-Time Systems Symp., pp. 198--207, Dec. 1992.

....then the node may not have enough network access time to transmit a message before its deadline. Therefore, guaranteeing synchronous message deadlines lies on the appropriate allocation of synchronous bandwidths to the nodes. Many synchronous bandwidth allocation (SBA) schemes have been proposed [3, 4, 5, 6, 7, 8, 9, 10, 11] in terms of guaranteeing synchronous message deadlines. In [3] four SBA schemes are proposed and analyzed by Agrawal et al. and a metric called the worst case achievable utilization (WCAU) is adopted as a means to compare and evaluate different schemes. The WCAU of a SBA scheme is defined as the ....

....by Zheng and Shin [12] More recently, an improved NPA normalized proportional allocation scheme is proposed by Zhang et al. [11] to solve a problem in testing the feasibility of overallocated synchronous bandwidths produced by NPA. Optimal allocation of synchronous bandwidths is addressed in [5, 6, 8]. All proposed SBA schemes are divided into two classes: local and global. A SBA scheme is local if it allocates the synchronous bandwidth to each node without using information of messages on other nodes. In contrast, a global scheme utilizes system wide information including the message ....

[Article contains additional citation context not shown here]

B. Chen, G.Agrawal, and W.Zhao, "Optimal synchronous capacity allocation for hard real-time communications with the timed token protocol," in Proc. IEEE RTSS'92, pp. 198--207.

....Zho92] with hard real time constraints, that might be analyzed using this methodology. For these, and other problems, it is important to characterize not only correctness but also timeliness. Perhaps the most useful application of these techniques lies in the analysis of communication protocols [CAZ92, MSST93] which generally have only informal claims of efficiency and even correctness. For many of these, especially the distributed group communication protocols, the correctness guarantees are not always clear, and only recently have there been attempts at stating these more formally [HT93, ....

Baio Chen, Gopal Agrawal, and Wei Zhao. Optimal synchronous capacity allocation for hard real-time communications with the timed token protocol. In Proceedings of the Thirteenth Real-Time Systems Symposium, pages 198-- 207, Phoenix, AZ, December 1992. IEEE Computer Society Press.

....4 experimental evaluation of the performance of the algorithms is discussed in Section 6. Finally, the paper concludes with a summary of the work and conclusions in Section 7. 2. Previous Work Past efforts in real time communication research have focused primarily on static message scheduling [ACZD92, BM91, CAZ92, Fer92, FV90, KZ93, KMZ93, MZ95, MKZ96, Sath93, SLS88a, TBW95]. The fundamental thesis of these efforts is that the communication behavior of the application can be made to be deterministic through extensive a priori knowledge about communication load parameters, message latencies, and precedence relationships. The a priori information is then used to ....

B. Chen, G. Agrawal, and W. Zhao, "Optimal Synchronous Capacity Allocation for Hard Real-Time Communications with the Timed Token Media Access Control Protocol," Proceedings of The IEEE Real-Time Systems Symposium, December 1992.

....executing on multiple computers require time constrained message transmission. These messages, known as synchronous messages, are periodically generated and are subject to real time constraints. In recent years significant advances have been made in the area of real time communication scheduling [2, 1, 4, 10, 12, 13, 15, 14]. In many cases, the approach taken is to emulate centralized scheduling and exploit well known results from uniprocessor scheduling. However, the scheduling of hard real time messages in a local area network is inherently different due to its distributed nature. As a result, significant overheads ....

....critical that the real time communication schemes operate with minimal global state information. The protocols developed in recent years may be classified into three classes [6] global priority arbitration protocols [13, 15, 14] global reservation protocols [1] and bounded access time schemes [10, 2]. A survey of various real time communication protocols can be found in [6] In this paper, we consider the problem of guaranteeing timing constraints for synchronous traffic on slotted ring and DQDB networks, both of which use slots as unit of transmission. We propose a bandwidth allocation ....

B. Chen, G. Agrawal, and W. Zhao. Optimal Synchronous Capacity Allocation for Hard Real-Time Communicatins with the Timed Token Protocol. In Proceedings, IEEE Real-Time Systems Symposium, pages 198--207, December 1992.

....real time systems [23, 31] Unfortunately, due to their intrinsic randomness in medium access, CSMA CD and its variants are unable to provide a deterministic guarantee on message delivery time. Afterward, fiber distributed data interface (FDDI) was shown to be suitable for real time communications [1, 8]. Specifically, the timed token protocol used in the FDDI can guarantee the deadlines of periodic messages. Not surprisingly, FDDI has become one of the most popular networks for distributed real time systems. For the past several years, sales of FDDI networks have been growing at an annual rate ....

....Token Cell Cell Cell Cell Cell Channel Figure 1: An architecture for FDDI based mobile networks and a BS is a topic of many papers (e.g. 7, 11, 15, 24] and will not be discussed here. Supporting real time communications in regular (non mobile) FDDI networks has been studied in [1, 2, 8, 33], where the major concern is the amount of FDDI synchronous bandwidth that is needed for a given real time connection. Several schemes have been proposed in attempts to minimize the amount of synchronous bandwidth allocated to a connection. Most technical issues that do not exist in a regular FDDI ....

[Article contains additional citation context not shown here]

B. Chen, G. Agrawal, and W. Zhao, "Optimal synchronous capacity allocation for hard realtime communications with the timed token protocol," in Proc. Real-Time Systems Symposium, December 1992.

....excess amount of synchronous capacities to the nodes increases T TR , which may also cause message deadlines to be missed. Therefore, synchronous capacities must be properly allocated to individual nodes. As a consequence, synchronous capacities allocated to the nodes must satisfy two constraints [Agr92, Che92, Mal93]: a protocol constraint and a deadline constraint. The protocol constraint states that the total sum of the allocated synchronous capacities should not be greater that the available portion of T TR , i.e. n i TTRT H 1 t (3.1) Theoretically, the total available time to transmit ....

....to the nodes should be such that the synchronous messages are always guaranteed to be transmitted before their deadlines. A message set can be guaranteed by an allocation scheme once the protocol and the deadline constraints are satisfied. Several allocation schemes have been proposed in [Agr92, Che92, Mal93]. In [Vas94] we can find a first analysis on the message schedulability in Profibus networks. Based on the Timed Token Protocol, these results were later improved and presented in [Vas96] The work here by described is a step forward in the analysis of message schedulability in Profibus networks. ....

Chen, B., G. Agrawal and W. Zhao, "Optimal Synchronous Capacity Allocation for Hard Real-Time Communications with Timed Token Protocol", Proceedings of the 13 th IEEE Real-Time Systems Symposium, pp.198-207 (December 1992).

....a token arrives late, a node transmits only synchronous messages. The performance of FDDI depends on protocol parameters such as TTRT and synchronous bandwidths. Thus, previous works focused on the synchronous bandwidth allocation to meet the timing constraints of messages. Agrawal et al. 1992)(Chen et al. 1992)(Agrawal et al. 1993) Chen et al. 1992) proposed optimal allocation schemes. However, depending on the periodic message streams, the network utilization may drop below 33 . 2.3 Overview of FC AL FC AL is a serial data channel that provides a logical bidirectional link between two L Ports. ....

....synchronous messages. The performance of FDDI depends on protocol parameters such as TTRT and synchronous bandwidths. Thus, previous works focused on the synchronous bandwidth allocation to meet the timing constraints of messages. Agrawal et al. 1992) Chen et al. 1992) Agrawal et al. 1993) (Chen et al. 1992) proposed optimal allocation schemes. However, depending on the periodic message streams, the network utilization may drop below 33 . 2.3 Overview of FC AL FC AL is a serial data channel that provides a logical bidirectional link between two L Ports. Every node in FC AL is connected to the ....

Chen, B., G. Agrawal and W. Zhao (1992). Optimal synchronous capacity allocation for hard real-time communications with the timed token protocol. In: Proceedings of Real-Time Systems Symposium. pp. 198--207.

....to the nodes should be such that synchronous messages are always transmitted before their deadlines. As a result, a message set can be guaranteed by an allocation scheme once the protocol and the deadline constraints are satisfied. Several allocation schemes have been proposed in the literature [17 19]. Both FDDI and IEEE802.4 are examples of network protocols based on the timed token protocol. Upper bounds for the time elapsed between two consecutive token arrivals can be found in [13] and [14] These results cannot however be applied to PROFIBUS, as significant differences to the timed token ....

Chen, B., Agrawal, G. and Zhao, W.: "Optimal Synchronous Capacity Allocation for Hard RealTime Communications with the Timed-Token Protocol", Proceedings of the 13 th IEEE Real-Time Systems Symposium, pp. 198-207, December 1992.

.... R b prec#m# (10) R w m = # c max m Hm # # #T ,Hm# C max m R w prec#m# (11) where T denotes the time interval for which a node gains transmission right again. T is fixed to one frame time in TDMA network. On the contrary, it is varying but bounded in other MAC protocol such as FDDI [2]. In this case, Eq. 10) and (11) should be modified according to the MAC layer. In priority arbitration based approach, we analyze the CAN (controller area network) bus [17] The CAN bus supports fixed priority driven media access protocol. Thus, the message transmission on the CAN bus is ....

B. Chen, G. Agrawal, and W. Zhao. Optimal synchronous capacity allocation for hard real-time communications with the timed token protocol. In Proceedings of Real-Time Systems Symposium, pages 198--207, Dec. 1992.

....[3] As a result, each node is guaranteed a certain bandwidth for transmitting synchronous messages. Using this property, several different schemes have been proposed in the literature for selecting TTRT and synchronous capacities to guarantee the hard deadlines of a given set of periodic messages [3, 7, 18, 43]. These schemes can be characterized using a metric called the 19 P1 P3 P2 P4 P5 P6 P7 Figure 2.4: Example of a token ring architecture Worst Case Achievable Utilization (WCAU) The WCAU of a parameter selection scheme is defined as the largest utilization U such that the scheme can guarantee ....

B. Chen, G. Agrawal, and W. Zhao, "Optimal synchronous capacity allocation for hard real-time communications with the timed token protocol," in Proceedings of Real-Time Systems Symposium, pp. 198--207, 1992.

....proposed [1 3] Local SBA schemes use only information available locally to each node, and are thus preferred to global schemes because of their lower network management overhead. Unfortunately, it has been formally proved in [4] that there does not exist any optimal local SBA scheme. Chen et al. [2] proposed the only known optimal global SBA scheme which is based on an iterative approach. However, their algorithm may not terminate theoretically. In this paper, we present an optimal global SBA scheme of polynomial time worst case complexity. 1 Introduction The problem of guaranteeing the ....

....less than TTRT. Many researchers studied the access time bounds and other timing properties of the timed token protocol. In particular, Johnson et al. 5, 6] proved that the average token cycle time is bounded by TTRT, and the maximum token cycle time is bounded by 2 Theta TTRT. Agrawal et al. [1, 2] extended Johnson s result and proved that the time elapsed between k consecutive token s visits to a node is bounded by k Theta TTRT. They also formulated a synchronous bandwidth allocation (SBA) problem and attempted to calculate the synchronous bandwidth H i that should be allocated to node ....

[Article contains additional citation context not shown here]

B. Chen, G. Agrawal, and W. Zhao, "Optimal synchronous capacity allocation for hard real-time communications with the timed token protocol," in Proc. of the 13th Real-Time Systems Symposium, (Phoenix, Arizona), pp. 198--207, December 1992.

....to ensure that the bandwidth guaranteed to each station for synchronous messages is no lower than the traffic generated for that service class. Agrawal et al. proposed and analyzed several schemes for selecting the synchronous capacities to guarantee the deadlines of real time messages [2 4]. Hamdaoui and Ramanathan extended this work to include the selection of both TTRT and synchronous capacities [8,9] In [15] Malcolm and Zhao accounted for arbitrary deadlines in selecting the parameters. Given a set of real time streams, the above schemes can be used to select suitable protocol ....

....assigned to stream S ij . The overall synchronous capacity assigned to station i is H i = P p i j=1 h ij . We assume that the synchronous capacities are selected in such a way that the deadline of every real time message in the system is guaranteed. Algorithms for this purpose can be found in [2 4,8,9,15]. The basic idea of these algorithms is to assign capacities such that the transmission time available to a station between the arrival of a message and its deadline is sufficient to fully transmit the message. Messages may have to be fragmented and transmitted in multiple token visits in order to ....

B. Chen, G. Agrawal, and W. Zhao, "Optimal synchronous capacity allocation for hard real-time communications with the timed token protocol," in Proceedings of Real-Time Systems Symposium, pp. 198--207, December 1992.

....would again require hardware support, in particular global synchronized clocks. A lot of work has been done in the area of timed token protocols since it was first proposed by Grow [13] in 1982. Protocols for guaranteeing synchronous deadlines based on timed token protocols can be found in [5] [8], 16] 17] These protocols assume an underlying token bus MAC protocol. Our work, however, is the software implementation of such a protocol over commodity network hardware, with enhancements to cater to applications with periodic channel access requirements. Also, RETHER addresses reliability ....

Chen, B., Agrawal, G., and Zhao, W. Optimal synchronous capacity allocation for hard real-time communications with timed token protocol. Proc. IEEE Real-time Systems Symp., Dec. 1992.

....require any changes in the current DQDB standards. 1 Introduction The problem of guaranteeing the timely delivery of messages has drawn considerable attention, especially in the areas of voice video data transmission over a data network, and message communication in an embedded real time system [2 5]. In the former area, a certain amount of bandwidth must be pre allocated in order to guarantee the bounds on delay and delay jitter in delivering video audio frames consistent with human perception. In the latter area, embedded real time applications are usually realized by executing a number of ....

B. Chen, G. Agrawal, and W. Zhao, "Optimal synchronous capacity allocation for hard real-time communications with the timed token protocol," in Proc. of the 13th Real-Time Systems Symposium, (Phoenix, Arizona) , pp. 198--207, December 1992.

....for FDDI and is applicable to Token bus. This bounds the network access latency for each node and is used to guarantee jitter bounds for multimedia applications. A lot of work has been done to develop schemes for efficient synchronous bandwidth allocation to nodes on a token ring network [ACZD92, CAZ92, DOC93] There has also been work that addresses the problem of supporting real time traffic on collision based networks. In [GW85] and [SSC91] the authors propose hybrid token CSMA CD CHAPTER 2. RELATED WORK 15 protocols to support real time and non real time traffic. Real time data is ....

B. Chen, G. Agrawal, and W. Zhao. Optimal synchronous capacity allocation for hard real-time communications with timed token protocol. Proc. IEEE Real-time Systems Symp., Dec. 1992.

....computers require deterministic guarantees on the worst case message transmission delays. We propose to enhance the DQDB MAC protocol to offer services that support such quality of service requirements. There have been several studies on bandwidth allocation for hard real time communication [2, 7, 9, 10]. However, most of them are not applicable to the dualbus architecture. In [9] a global priority scheme is used to guarantee deadlines of messages in dualbus networks. This scheme is not very suitable for DQDB since it requires a large number of priority levels (which is not supported in DQDB) to ....

B. Chen, G. Agrawal, and W. Zhao. Optimal Synchronous Capacity Allocation for Hard RealTime Communicatins with the Timed Token Protocol. In Proceedings, IEEE Real-Time Systems Symposium, pages 198--207, December 1992.

....channels and providing guarantees of maximum delivery delays. For local area network, IEEE 802.4 token bus network [7] IEEE 802.5 token ring network [8] and FDDI [9] adopt the timed token medium access control (MAC) protocol for providing bounded medium access times. Agrawal et al. [10, 11] and Han et al. 12, 13] attempted to solve the synchronous bandwidth allocation problem for FDDI networks to meet the protocol constraint while transmitting all synchronous messages before their deadlines. Another protocol which aims to provide timeconstrained communication services is the DQDB ....

B. Chen, G. Agrawal, and W. Zhao, "Optimal synchronous capacity allocation for hard real-time communications with the timed token protocol," in Proc. of Real-Time Systems Symposium, Dec. 1992.

....aim of the real time communication sub system is to guarantee timely delivery of synchronous messages. In addition, it is desirable to have low response times for the asynchronous messages. In recent years significant advances have been made in the area of real time communication scheduling [3, 1, 2, 8, 15, 17, 18, 23, 22]. In many cases, the approach taken is to emulate centralized scheduling and exploit well known results from uniprocessor scheduling. However, the scheduling of hard real time messages in a local area network is inherently different due to its distributed nature. As a result, significant overheads ....

....critical that the real time communication schemes operate with minimal global state information. The protocols developed in recent years may be classified into three classes [10] global priority arbitration protocols [18, 23, 22] global reservation protocols [2] and bounded access time schemes [15, 1, 3]. A survey of various real time communication protocols can be found in [10] In this paper, we consider the problem of guaranteeing timing constraints for synchronous traffic on a slotted ring network 1 . The allocation scheme is developed on top of a recently proposed media access control ....

B. Chen, G. Agrawal, and W. Zhao. Optimal Synchronous Capacity Allocation for Hard RealTime Communicatins with the Timed Token Pro- tocol. In Proceedings, IEEE Real-Time Systems Symposium, pages 198--207, December 1992.

....should be chosen small enough to satisfy the most stringent response time requirements of all nodes. Each node is assigned a fraction of the TTRT, known as its synchronous bandwidth, which is the maximum time the node is allowed to transmit its synchronous messages every time it receives the token[5, 3]. After synchronous message transmission, asynchronous messages can be initiated (if any) but only if the token has arrived at the node earlier than expected. That is, synchronous traffic is assigned a guaranteed bandwidth while the leftover bandwidth (unallocated, unused or both) is dynamically ....

....synchronous message deadlines with the timed token protocol, the protocol parametres (TTRT and the synchronous bandwidths) have to be properly selected. Much work on a proper selection of these parametres has been reported in the literature, with the focus on synchronous bandwidth allocation (SBA) [7, 12, 5, 2, 14, 15, 21]. Hamdaoui and Ramanathan[7] address the problem of setting both TTRT and the synchronous bandwidth of each node so as to guarantee sets of periodic message streams. Similar work was conducted by Lim et al.[12] who studied the deadline guarantee of timedependent multimedia data in an FDDI network. ....

[Article contains additional citation context not shown here]

B. Chen, G. Agrawal and W. Zhao, "Optimal Synchronous Capacity Allocation for Hard Real-Time Communications with the Timed Token Protocol", Proceedings of the 13th IEEE Real-Time Systems Symposium, pp. 198-207 (December 1992).

....calendar activate deactivate Calendars Pending calendar de activations Results Table Real Time Maruti Hard EU done Modified Mach 3.0 Kernel Maruti Run Time Dispatcher Application Figure 4. 10 Maruti II Run Time Organization nication protocols based on time division multiplexing schemes (e.g. [3, 9, 10]) provide better control over the network resources and allow for decoupling of communication medium from the processing nodes. The design of Maruti permits both approaches for communication scheduling. 4.5.5 Task Allocation and Global Scheduling In a distributed real time system, the problem of ....

B. Chen, G. Agrawal, and W. Zhao. Optimal Synchronous Capacity Allocation for Hard Real-Time Communications with the Timed Token Protocol. In Proceedings, IEEE Real-Time Systems Symposium, pages 198--207, December 1992.

....to ensure that the bandwidth guaranteed to each station for synchronous messages is no lower than the traffic generated for that service class. Agrawal et al. proposed and analyzed several schemes for selecting the synchronous capacities to guarantee the deadlines of real time messages [2, 3]. Hamdaoui and Ramanathan extended this work to include the selection of both TTRT and synchronous capacities [7, 8] Given a set of real time streams, the above schemes can be used to select suitable protocol parameters to guarantee the deadline constraint of each real time message. The second ....

....assigned to stream S ij . The overall synchronous capacity assigned to station i is H i = P p i j=1 h ij . We assume that the synchronous capacities are selected in such a way that the deadline of every real time message in the system is guaranteed. Algorithms for this purpose can be found in [2, 3, 7, 8]. The basic idea of these algorithms is to assign capacities such that the transmission time available to a station between the arrival of a message and its deadline is sufficient to fully transmit the message. Messages may have to be fragmented and transmitted in multiple token visits in order to ....

B. Chen, G. Agrawal, and W. Zhao, "Optimal synchronous capacity allocation for hard realtime communications with the timed token protocol, " in Proceedings of Real-Time Systems Symposium, pp. 198--207, December 1992.

....than expected. Many researchers studied the access time bounds and other timing properties of the timed token protocol. In particular, Johnson et al. 5, 6] prove that the average token cycle time is bounded by TTRT, and the maximum token cycle time is bounded by 2 Theta TTRT. Agrawal et al. [1, 2] extend Johnson s result and prove that the time elapsed between k consecutive token s visits to a node is bounded by k Theta TTRT. They also formulated a synchronous bandwidth allocation (SBA) problem and attempted to calculate the synchronous bandwidth H i that should be allocated to node i, ....

....because they were calculated independently of node i s message stream parameters. As the global SBA schemes use global information to allocate synchronous bandwidths, they are naturally expected to achieve better performance. To our best knowledge, there are only two optimal global SBA schemes [2, 3] and several non optimal local SBA schemes [1, 4] reported in the open literature. By an optimal SBA scheme, we mean a SBA scheme that finds a feasible set of H i s subject to the protocol and deadline constraints whenever such a set exists. Whether to choose a non optimal local scheme or an ....

[Article contains additional citation context not shown here]

B. Chen, G. Agrawal, and W. Zhao, "Optimal synchronous capacity allocation for hard real-time communications with the timed token protocol," in Proc. of the 13th Real-Time Systems Symposium, (Phoenix, Arizona), pp. 198--207, December 1992.

....bandwidth to transmit a particular message may not be available between the time it arrives and its deadline. For guaranteeing the deadlines of periodic message streams, parameter selection schemes in the literature have mainly focussed on choosing suitable synchronous capacity allocations [2 4,28]. In [2,3] several synchronous capacity allocation schemes are compared using a metric called the Worst Case Achievable Utilization (WCAU) Informally, the WCAU of a synchronous capacity allocation scheme is defined as the largest utilization U such that the scheme can guarantee the deadlines of ....

....that the normalized proportional allocation scheme (formally defined in Section 4) has a WCAU of 1=3 if the TTRT is selected as recommended in the FDDI standard. More recently, Chen et al. proposed an optimal allocation scheme that uses an iterative method to compute the synchronous capacities [4]. The WCAU of this scheme is shown to lie between 1=3 and 2=5. In [28] the authors devised a synchronous capacity allocation scheme when the deadlines are not necessary equal to the corresponding periods. The above parameter selection algorithms do not address the selection of TTRT. They assume ....

[Article contains additional citation context not shown here]

B. Chen, G. Agrawal, and W. Zhao, "Optimal synchronous capacity allocation for hard real-time communications with the timed token protocol," in Proceedings of Real-Time Systems Symposium, pp. 198--207, December 1992.

....systems. Liu and Layland first proposed the concept of worst case achievable utilization in the context of rate monotonic processor scheduling [13] More recently, the worst case achievable utilization has been calculated for various bandwidth allocation schemes for the timed token protocol [1, 2]. A lower bound on the worst case achievable utilization has also been developed for bandwidth allocation in slotted ring networks [16] This paper, however, is the first to consider the worst case achievable utilization for ATM networks. To further evaluate the system capability for supporting ....

B. Chen, G. Agrawal, and W. Zhao. Optimal synchronous capacity allocation for hard real-time communications with the timed token protocol. In Proceedings of the IEEE Real-Time Systems Symposium, pages 198--207, Dec. 1992. 19

....always guarantee a message set whenever there exists an allocation that can do so. While local SBA schemes are simple to implement, it has been proved that no local SBA schemes are optimal [8] An optimal global SBA scheme, called the minimum capacity allocation scheme (MCA) was first proposed in [5]. However, it was pointed out that besides being restricted to the case where the minimum message deadline is greater than or equal to 2 Delta TTRT , MCA does not always generate the optimal allocation. An enhanced version of MCA, called the enhanced minimum capacity allocation scheme (EMCA) was ....

....second approach is for the SBA scheme to partition the time length required to send a message, C i , for each synchronous message stream among a certain number of token rotating cycles. An example for this approach is the LA scheme (H i = C i b P i TTRT c Gamma1 ) 2] It has been proved in [5] that, assuming at time t a synchronous message with deadline D i arrives at node i(1 i n) then with P i = D i , in time interval (t; t D i ] the minimum amount of time X i available for node i to transmit its synchronous messages is given by: X i = q i Gamma 1) Delta H i max ( 0; ....

[Article contains additional citation context not shown here]

B. Chen, G. Agrawal, and W. Zhao. Optimal synchronous capacity allocation for hard real-time communications with the timed token protocol. Proc. Real-Time Syst. Symp., pages 198--207, 1992.

....The synchronous bandwidth allocation schemes discussed above are not optimal. An optimal synchronous bandwidth allocation scheme is able to guarantee that the message deadlines will be met whenever it is possible to do so. An optimal synchronous bandwidth allocation scheme has been developed (Chen et al. 1992), and is known to have a worst case achievable utilization between 33 to 40 . However, the optimal scheme is more complicated than the nonoptimal schemes discussed here, and might be difficult to implement. Given that some of the nonoptimal schemes have a worst case achievable utilization close ....

....(Agrawal et al. 1992, 1993) We use the term synchronous bandwidth in order to be consistent with the current FDDI standard. 5. TTRT can be no more than half of the minimum message deadline because the maximum time between consecutive token visits to a node can approach 2TTRT (Agrawal et al. 1992; Chen et al. 1992; Johnson 1987) 6. The synchronous bandwidth allocation schemes listed below assume that there is only one stream of synchronous messages originating at each node. This assumption involves no loss of generality, because a network in which several streams may originate at each node can be ....

Chen, B., G. Agrawal, and W. Zhao. 1992. Optimal synchronous capacity allocation for hard real-time communications with the timed token protocol. In Proceedings of the IEEE Real-Time Systems Symposium, pages 198--207.

....token departure from the node is less than the value of TTRT, i.e. only if the token arrives earlier than expected. In order to guarantee that the deadlines of synchronous messages are met, network parameters such 1 Some other synonymous terms that researchers use are synchronous capacity [1, 2, 4], bandwidth allocation [18] synchronous allocation [7] synchronous bandwidth assignments [10] and high priority token holding time [13] This paper uses the term synchronous bandwidth, in accordance with the most recent version of the FDDI standard. as the synchronous bandwidth, the target ....

....systems. The objective in much of the previous work was to maximize the throughout and or to minimize the delay, rather than to guarantee individual message deadlines [5, 9, 13, 14, 20] There has recently been much progress in synchronous bandwidth allocation for real time applications. In [1, 2, 4], a real time system is considered where synchronous messages on different nodes can have different periods. However, the deadline of a message is assumed to be equal to its period. For this case it was shown that the worst case achievable utilization of the network can be 33 ; if a message set ....

[Article contains additional citation context not shown here]

B. Chen, G. Agrawal, and W. Zhao. Optimal synchronous capacity allocation for hard real-time communications with the timed token protocol. In Proceedings of the Real-Time Systems Symposium, pages 198--207, Phoenix, Arizona, Dec. 1992.

....Sha [22] However, no analysis of such a scheme has been reported so far. The timed token protocol was first proposed by Grow in [6] This protocol has the important property of bounded access time for the nodes on the ring. The real time performance of this protocol has been recently studied in [2, 3, 4, 9]. It was shown that with a proper choice of protocol parameters, this protocol can guarantee a synchronous traffic of up to 33 in the worst case. However, the issue of average case performance of this protocol has not been addressed so far. Our objective here is to derive precise criteria for ....

.... Subsequently, Strosnider and Katcher[23] have used this approach to evaluate the performance of the rate monotonic algorithm under non ideal conditions by accounting for the preemption overheads for various operating system im 1 also referred as Worst Case Achievable Utilization in literature [2, 3, 4, 12]. plementation policies. They authors have reported that the average breakdown utilization of the rate monotonic algorithm can be significantly lower than the ideal due to preemption overheads and possibilities of blocking or priority inversion. Our analysis of the two protocols extends the ....

[Article contains additional citation context not shown here]

B. Chen, G. Agrawal, and W. Zhao, "Optimal Synchronous Capacity Allocation for Hard Real-Time Communications with the Timed Token Media Access Control Protocol", Proc. IEEE Real-Time Syst. Symp., 1992.

....ensure a bounded access delay for every node in the network [16] The high bandwidth supported by the fiber optic medium, and the guaranteed response time provided by the timed token protocol, make the FDDI standard an attractive proposition for use in real time systems. In several recent studies [1, 2, 4, 13, 14, 19], the real time performance of FDDI networks has been analyzed in terms of their worst case achievable utilization. Message sets with utilization less than the worst case achievable utilization are always guaranteed to meet their deadlines. Knowledge of the worst case achievable utilization is ....

....will be relaxed later in the paper. 3.1 Synchronous Bandwidth Allocation The guarantee probability of an FDDI network depends on the synchronous bandwidth allocation scheme that is used. Several synchronous bandwidth allocation schemes have been proposed, both for the case of D i = P i [1, 2, 4] and for the case of D i 6= P i [13, 14, 19] These schemes all have a similar worst case performance. We choose the scheme from [14] which allocates the synchronous bandwidths as H i = D i P i Delta C i b D i TTRT Gamma 1c : 12) This scheme, which is an extension of that in [1] can ....

B. Chen, G. Agrawal, and W. Zhao. Optimal synchronous capacity allocation for hard real-time communications with the timed token protocol. In Proceedings of the Real-Time Systems Symposium, pages 198--207, Dec. 1992.

....transmit its synchronous messages. In this section, some protocol timing properties are enumerated. These will be useful in our analysis of the optimal allocation scheme. Due to space limitations, the proofs of these timing properties are not presented here. The interested reader should refer to [9] for the formal proofs. Let t i (v; t) l = 1; 2; denote the maximum time elapsed between time t and the time when the token makes its v th visit to node i after time t. LEMMA 5.1 For v 1, t i (v; t) t v Delta TTRT X h=1; n;h6=i H h : 13) THEOREM 5.1 Assume that at time ....

.... Gamma 1) Delta H i max(0; min(r i Gamma ( X h=1; n;h6=i H h Gamma ) H i ) 14) where q i = b P i TTRT c. From Lemma 5.1 and Theorem 5. 1, we can make the following observations: ffl There are two terms in the expression of X i ( H) given in (14) From the formal proof given in ([9]) it is obvious that the first term, i.e. q i Gamma 1) Delta H i , represents the minimum amount of time available for node i to transmit its synchronous messages in interval (t; t q i Delta TTRT ] Note that the length of time interval (t; t q i Delta TTRT ] is equal to q i Delta TTRT ....

B. Chen, G. Agrawal, W. Zhao, "Optimal Synchronous Capacity Allocation for Hard RealTime Communications with the Timed Token Media Access Control Protocol," Technical Report, Computer Science Dept., Texas A&M Univ., May 1992.

....bandwidths to the nodes could increase the token rotation time, which may also cause message deadlines to be missed. Over the past two years, extensive studies have been carried out on this subject. The first comprehensive study on synchronous bandwidth allocation for FDDI was reported in [1] In [5], an optimal bandwidth allocation algorithm was studied. An optimal algorithm can guarantee deadlines for all messages assigned to an FDDI ring whenever there is an allocation method that can do so. However, the optimal algorithm is complicated and may not be feasible for on line use. ....

B. Chen, G. Agrawal, and W. Zhao, "Optimal Synchronous Capacity Allocation for Hard Real-Time Communications with the Timed Token Media Access Control Protocol," Proc. IEEE Real-Time Syst. Symp., 1992.

....provides additional information about the sensitivity of the performance to system utilization. Being defined as a function of utilization, it provides a detailed picture of the system performance as the load is varied. 1 Also referred as Worst Case Achievable Utilization in the literature [2, 3, 4, 15]. Based on these measures, we evaluate and compare the performance of the priority driven protocol and the timed token protocol. We find that neither protocol is superior to the other under all operating conditions. The priority driven protocol allows implementation of an optimal scheduling ....

....token protocol was first proposed by Grow in [7] Sevcik and Johnson [23] formally demonstrated that this protocol ensures a bounded access time for all stations on the ring. The issue of guaranteeing the deadlines of synchronous real time messages with this protocol has been addressed recently in [2, 3, 4]. The average breakdown utilization measure was introduced by Lehoczky, Sha, and Ding [13] to evaluate the average case performance of the rate monotonic algorithm in a CPU scheduling environment. We extend the methodology adopted in [13] to the token ring environment for obtaining the average ....

[Article contains additional citation context not shown here]

B. Chen, G. Agrawal, and W. Zhao. "Optimal Synchronous Capacity Allocation for Hard Real-Time Communications with the Timed Token Media Access Control Protocol". Proc. of the IEEE Real-Time Syst. Symp., Dec. 1992.

....minimum breakdown utilization. Thus, at the design stage, it is more appropriate to select a protocol based on its average case performance using measures such as the average breakdown utilization or the guarantee probability. 1 Also referred in the literature as worst case achievable utilization [2 4, 15]. Sec. 6.1 Introduction 118 The average breakdown utilization is a point measure and hence can be used as a single figure of merit to evaluate protocol performance. It represents how high the utilization can be, on average, without missing any deadlines. The guarantee probability provides ....

....token protocol was first proposed by Grow in [7] Sevcik and Johnson [22] formally demonstrated that this protocol ensures a bounded access time for all stations on the ring. The issue of guaranteeing the deadlines of synchronous real time messages with this protocol has been addressed recently in [2 4]. The average breakdown utilization measure was introduced by Lehoczky, Sha, and Ding [13] to evaluate the average case performance of the rate monotonic algorithm in a CPU scheduling environment. We extend the methodology adopted in [13] to the token ring environment for obtaining the average ....

[Article contains additional citation context not shown here]

B. Chen, G. Agrawal, and W. Zhao. "Optimal Synchronous Capacity Allocation for Hard Real-Time Communications with the Timed Token Media Access Control Protocol." Proc. of the IEEE Real-Time Systems Symp., Dec. 1992.

....for scheduling periodic real time tasks on a uniprocessor [15] and works by giving precedence to tasks with small periods. Other work has concentrated on how much network bandwidth needs to be dedicated to real time messages in order to avoid missing deadlines with the timed token protocol [1, 2, 5, 14, 17, 28]. Recently, there has also been much interest in guaranteeing that message deadlines will be met in wide area packet switched networks [7, 11, 13, 18, 22, 25, 26, 27] An excellent survey of the work in this area may be found in [4] This paper differs from the previous work in that this paper is ....

B. Chen, G. Agrawal, and W. Zhao. Optimal synchronous capacity allocation for hard real-time communications with the timed token protocol. In Proceedings of the IEEE Real-Time Systems Symposium, pages 198--207, Dec. 1992.

No context found.

Che92b. Chen, B., G. Agrawal and W. Zhao, "Optimal Synchronous Capacity Allocation for Hard Real-Time Communications with the Timed Token Protocol", Proceedings of the 13th IEEE Real-Time Systems Symposium, pp. 198-207 (December 1992).

No context found.

Che92a. Chen, B., G. Agrawal and W. Zhao, "Optimal Synchronous Capacity Allocation for Hard Real-Time Communications with the Timed Token Media Access Control Protocol ", Technical Report, Computer Science Department, Texas A&M University (May 1992).

No context found.

Che92b. Chen, B., G. Agrawal and W. Zhao, "Optimal Synchronous Capacity Allocation for Hard Real-Time Communications with the Timed Token Media Access Control Protocol ", Technical Report, Computer Science Department, Texas A&M University (May 1992).

No context found.

Che92a. Chen, B., G. Agrawal and W. Zhao, "Optimal Synchronous Capacity Allocation for Hard Real-Time Communications with the Timed Token Protocol", Proceedings of the 13th IEEE Real-Time Systems Symposium, pp. 198-207 (December 1992).

No context found.

B. Chen, G. Agrawal, and W. Zhao. Optimal synchronous capacity allocation for hard real-time communications with the timed token protocol. In Proceedings 13 IEEE Real-Time Systems Symposium, pages 198--207, Phoenix, AZ, USA, Dec. 1992. IEEE Computer Society Press. ISBN 0-8186-3195-3.

No context found.

Che92b. Chen, B., G. Agrawal and W. Zhao, "Optimal Synchronous Capacity Allocation for Hard Real-Time Communications with the Timed Token Protocol", Proceedings of the 13th IEEE Real-Time Systems Symposium, pp. 198-207 (December 1992).

No context found.

Che92a. Chen, B., G. Agrawal and W. Zhao, "Optimal Synchronous Capacity Allocation for Hard Real-Time Communications with the Timed Token Media Access Control Protocol ", Technical Report, Computer Science Department, Texas A&M University (May 1992).

No context found.

Che92b. Chen, B., G. Agrawal and W. Zhao, "Optimal Synchronous Capacity Allocation for Hard Real-Time Communications with the Timed Token Protocol", Proceedings of the 13th IEEE Real-Time Systems Symposium, pp. 198-207 (December 1992).

No context found.

Che92b. Chen, B., G. Agrawal and W. Zhao, "Optimal Synchronous Capacity Allocation for Hard Real-Time Communications with the Timed Token Media Access Control Protocol ", Technical Report, Computer Science Department, Texas A&M University (May 1992).

No context found.

Che92a. Chen, B., G. Agrawal and W. Zhao, "Optimal Synchronous Capacity Allocation for Hard Real-Time Communications with the Timed Token Protocol", Proceedings of the 13th IEEE Real-Time Systems Symposium, pp. 198-207 (December 1992).

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