Sridhar Pingali. Protocol and Real-Time Scheduling Issues for Multimedia Applications. PhD thesis, University of Massachusetts Amherst, September 1994.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....reliable multicast services for different applications have been presented and can be expected to co exist in the future. The approaches differ, among others, by the various error control mechanisms used. Several taxonomies were presented to classify the different multicast protocols (see [1], 2] 3] 4] 5] With respect to participation of group members in multicast error recovery, protocols can be classified as: ffl Centralized error recovery (CER) allows retransmissions exclusively to be performed by the multicast source, referred to also as source based recovery. ffl ....

S. Pingali, Protocol and Real-Time Scheduling Issues for Multimedia Applications, Ph.D. thesis, UMass, Sept. 1994.

.... treats all applications equally, where the network sends out one packet from an application in each time slot (if there are packets that have not been sent out from the application) First come first served is a scheduling policy that the packets are sent out according to their order of arriving [9, 10, 14]. All of these schedule policies make no difference to the system performance in the dropping model. But in the keeping model, if we give penalties to packets not sent out in time, different scheduling policies make a difference in the meaning of weighted packet transferability. The more time a ....

S. Pingali. Protocol and Real Time Scheduling Issues for Multimedia Applications. PhD thesis, Dept. of Elec. and Comp. Engg., University of Massachusetts, Amherst, September 1994.

....the sender to determine when data can safely be released from memory, without having to process ACKs for every packet. NAK based protocols can potentially suffer from the NAK implosion problem if many receivers detect errors. To remedy this problem, a NAK suppression scheme can be developed [17]. In the NAK suppression scheme, when a receiver detects an error, it waits for a random period of time and then multicasts a NAK to the sender and other receivers. When a receiver obtains a NAK for a packet which it intends to send a NAK for, the receiver behaves as if it had sent the NAK. The ....

....limitation, however, is that other mechanisms must be incorporated into a NAK based protocol to offer reliable multicast service. Figure 2 illustrates the NAK based protocols. In our implementation, the NAK based protocols with polling utilize a different NAK suppression scheme than the one in [17]. Our NAK suppression scheme is implemented at the sender side. A retransmission will happen only after a designated period of time has passed since the previous transmission. Thus, the receivers may send multiple NAKs to the sender while the sender performs retransmission only once. Ring based ....

S. Pingali, "Protocol and Real--Time Scheduling Issues for Multimedia Applications." PhD thesis, University of Massachusetts, Amhest, 1994.

....in any given slot is p. A non real time connection has an infinite deadline. 2. Real Time NonMulticast Connection: Each arriving cell in the stream has a n associated deadline and is routed to exactly one output multiplexer. The cell stream is assumed to follow a classical On Off traffic source [1, 19] as shown in Figure 2. The cell stream for the connection consists of arrivals at fixed intervals T during an active period having a mean a Gamma1 . The number of cells in an active period is geometrically distributed with a mean of (aT ) Gamma1 . The idle period is exponentially distributed ....

....technique MC EDD with EDD and FCFS techniques as described in section 3.2. 4.1 Assumptions We assume that the ATM multiplexer has a T1 output link of 1.544 Mbps giving a cell transmission of one time slot of 0.275 ms. We simulate each real time connection by a classical On Off source model [19], as shown in Figure 2, where the mean active period a Gamma1 and the mean idle period b Gamma1 (as defined in section 3.1) are 352 ms and 650 ms respectively. The cell separation time T is 16 ms. Hence, as defined in Equations 3 and 4 2 and 3 become: 2 = Nm : 352 352 650 : 1 58 = ....

Sridhar Pingali. "Protocol and Real-Time Scheduling Issues For MultiMedia Applications ". In Ph.D. Dissertation, September 1994.

....multicasts all packets, giving priority to retransmissions, and a receiver sends a negative acknowledgment (nak) when it detects an error or a lost packet. The first comparative analysis of ideal sender initiated and receiver initiated reliable multicast protocols was presented by Pingali et al. [17, 18]. This analysis showed that receiver initiated protocols are far more scalable than sender initiated protocols, because the maximum throughput of sender initiated protocols is dependent on the number of receivers, while the maximum throughput of receiver initiated protocols becomes independent of ....

....multicast transport protocol (reliable multicast protocol, for short) can be designed that enjoys all the scaling properties of the ideal receiver initiated protocols, while still being able to operate correctly with finite memory. To address this question, the previous analysis by Pingali et al. [17, 18, 22] is extended to consider the maximum throughput of generic ring based protocols, which organize receivers into a ring, and two classes of tree based protocols, which organize receivers into acknowledgment trees. These classes are the other three known approaches that can be used to solve the ack ....

[Article contains additional citation context not shown here]

Pingali, S. Protocol and Real-Time Scheduling Issues for Multimedia Applications. PhD thesis, University of Massachusetts Amherst, September 1994.

....acknowledgment (NAK) when it detects an error or a lost packet. This work was supported in part by the Office of Naval Research under Grant N00014 94 1 0688. The first comparative analysis of sender initiated and receiverinitiated reliable multicast protocols was presented by Pingali et al. [15, 16]. This analysis showed that receiver initiated protocols are far more scalable than sender initiated protocols, because the maximum throughput of sender initiated protocols is dependent on the number of receivers, while the maximum throughput of receiver initiated protocols is independent of the ....

....This paper addresses the question of whether a reliable multicast protocol can be designed that enjoys all the scaling properties of receiver initiated protocols, while still being able to operate correctly with finite memory. To address this question, the previous analysis by Pingali et al. [15, 16] is extended to consider the maximum throughput of generic ring based protocols, and two classes of tree based protocols. These classes are the other three known approaches that can be used to solve the ACK implosion problem. Our analysis shows that tree and ring based protocols can work ....

[Article contains additional citation context not shown here]

S. Pingali. Protocol and Real-Time Scheduling Issues for Multimedia Applications. PhD thesis, University of Massachusetts Amherst, September 1994.

....multicasts all packets, giving priority to retransmissions, and a receiver sends a negative acknowledgement (nak) when it detects an error or a lost packet. The first comparative analysis of ideal sender initiated and receiver initiated reliable multicast protocols was presented by Pingali et al. [17, 18]. This analysis showed that receiver initiated protocols are far more scalable than sender initiated protocols, because the maximum throughput of sender initiated protocols is dependent on the number of receivers, while the maximum throughput of receiverinitiated protocols becomes independent of ....

....multicast transport protocol (reliable multicast protocol, for short) can be designed that enjoys all the scaling properties of the ideal receiver initiated protocols, while still being able to operate correctly with finite memory. To address this question, the previous analysis by Pingali et al. [17, 18, 22] is extended to consider the maximum throughput of generic ring based protocols, which organize receivers into a ring, and two classes of tree based protocols, which organize receivers into ack trees. These classes are the other three known approaches that can be used to solve the ack implosion ....

[Article contains additional citation context not shown here]

Pingali S (1994) Protocol and Real-Time Scheduling Issues for Multimedia Applications. PhD thesis, University of Massachusetts, Amherst, Mass.

....by the term . The details of the proof are beyond the scope of this paper and may be found in [10] If the range of periods is small, then our result gives a much higher utilization bound compared to[20] Ours is a worst case result, and when we simulated a standard multimedia task set taken from [18], we observed that the available utilization under RMDP was almost the same as for RM. More importantly, we saw a 30 reduction in number of pre emptions under RMDP and more details can be found in [10] Our analysis of delayed pre emption is independent of the RM priority scheme itself and so is ....

Pingali, S., "Protocol and Real Time Scheduling Issues for Multimedia Applications," Ph.D dissertation, Dept. of Elec. and Comp. Engg., University of Massachusetts, Amherst, September 1994.

....an earliest deadline first policy. We refer to this as the mixed policy. We show that any set of tasks that are schedulable by the rm policy, are schedulable by the mixed policy. An interesting feature of the mixed policy is that it has fewer context switches compared to rm. In fact it is shown in [12] that edf has fewer context switches than the rm for periodic tasks. One can see that when a preemption occurs in the edf policy, the period of the arriving task must be smaller than the period of the running task, which implies that a preemption should occur in the rm policy as well. However, ....

Pingali, S., "Protocol and Real Time Scheduling Issues for Multimedia Applications," PhD Thesis, University of Massachusetts, Amherst, Sep 1994.

....approaches will coexist. A large number of protocols providing reliable multicast services have been presented which feature, among other differences, a large variety of error control mechanisms. Several taxonomies were presented to classify the large number of different multicast protocols (see [1, 2, 3, 4, 5]) Multicast error recovery can be classified, dependent on the participation of group members, as: 1 Copyright 1998 IEEE. Published in the Proceedings of INFOCOM 98, March 29 April 2 1998, San Francisco, USA. Personal use of this material is permitted. However, permission to reprint republish ....

S. Pingali, Protocol and Real-Time Scheduling Issues for Multimedia Applications, Ph.D. thesis, UMass, Sept. 1994.

....sender multicasts all packets, giving priority to retransmissions, and a receiver sends a negative acknowledgment (nak) when it detects an error or a lost packet. The first comparative analysis of sender initiated and receiver initiated reliable multicast protocols was presented by Pingali et al. [2, 3]. This analysis showed that receiver initiated protocols are far more scalable than sender initiated protocols, because the maximum throughput of sender initiated protocols is dependent on the number of receivers, while the maximum throughput of receiver initiated protocols is independent of the ....

....finite memory. This thesis addresses the question of whether a reliable multicast protocol can be 2 designed that enjoys all the scaling properties of receiver initiated protocols, while still being able to operate correctly with finite memory. To address this question, the previous analysis [2, 3] is extended to consider the maximum throughput of generic ring based protocols, and two classes of tree based protocols. These classes are the other three known approaches that can be used to solve the ack implosion problem. Our analysis shows that tree and ring based protocols can work ....

[Article contains additional citation context not shown here]

Sridhar Pingali. Protocol and Real-Time Scheduling Issues for Multimedia Applications. PhD thesis, University of Massachusetts Amherst, September 1994.

....[10] we derive the value of M , given that the source has a local receiver subset of size B from which to collect NAKs and retransmit packets to. The expected number of transmissions per packet is [2, 8] E[M ] B X i=1 B i ( Gamma1) i 1 1 (1 Gamma p i ) 3) It is shown in [9] that HB Gamma ln p E[M ] 1 HB Gamma ln p ; where HB = P B i=1 1=i, the harmonic numbers. From the known inequality ln(1 p) p 1 p , it follows that Gamma ln p p Gamma1 p . Using this result, assuming all operations (e.g. Xf and Xp ) are of constant cost, and taking into ....

....longer path that retransmissions must traverse to an expecting receiver. Figure 3(b) shows the number of supportable receivers by each of the different classes, relative to processor speed requirements. This number is obtained by normalizing all classes to a baseline processor. As described in [8, 9], the baseline uses a sender initiated protocol and can support exactly one receiver. As in [8, 9] let [R] be the speed of the processor that can support at most R receivers under protocol , where 2 fA; N1;N2; R; H1;H2g representing sender initiated, receiver initiated, RINA, ....

[Article contains additional citation context not shown here]

S. Pingali, Protocol and Real-Time Scheduling Issues for Multimedia Applications. PhD thesis, University of Massachusetts Amherst, September 1994.

No context found.

Sridhar Pingali. Protocol and Real-Time Scheduling Issues for Multimedia Applications. PhD thesis, University of Massachusetts Amherst, September 1994.

No context found.

Pingali S (1994) Protocol and Real-Time Scheduling Issues for Multimedia Applications. PhD thesis, University of Massachusetts, Amherst, Mass.

No context found.

Sridhar Pingali. Protocol and Real-Time Scheduling Issues for Multimedia Applications. PhD thesis, University of Massachusetts, Amherst, Mass, 1994.

No context found.

Pingali, S., "Protocol and Real Time Scheduling Issues for Multimedia Applications," PhD Thesis, University of Massachusetts, Amherst, Sep 1994.

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