Figueira, N., and Pasquale, J. Leave-in-time: A new service discipline for real-time communications in a packet-switching network. In ACM SIGCOMM '95 (1995).  Home/Search   Document Details and Download   Summary   ACM   TOC   Related Articles   Check

.... service in high speed networks, e.g. the VirtualClock [25] Fair Share Queueing [7] GPS [17] and RCSP [23] These models have also been used to derive statistical delay guarantees; in particular, within the framework of RCSP (in [24] and GPS (in [26] Related results can be found in [9] (for VitualClock) and in [22] for FCFS, with a variety of traffic distributions) In [14] statistical service quality objectives are achieved via proportional share queueing, in conjunction with server guided backoff, where servers dynamically Input Pulse [ I Q Equal. RangeDFT RCS 2OO Hz 36 ....

....gets reconfigured. When a resource fails, an executive process activates the redundant software components whose design was, after all, predetermined before the system was deployed. 13 CPU Tasks Derived From E[t] ms) Var[t] Min,Max] ms) NumSteps I Q Normal 59 100 [50,83] 33 EQ Normal 11 16 [9,24] 15 RDFT Exponential 103 [87 ,140] 53 CT Exponential 14 [12,30] 18 ADFT Exponential 187 [160,210] 50 KM Exponential 23 [20,35] 15 AIDFT Exponential 187 [160,210] 50 I Q,EQ Normal 70 144 [56,107] 51 KM, AIDFT Exponential 210 [180 , 245] 50 Network Tasks Derived From E[t] ms) Var[t] ....

Norival R. Figueira and Joseph Pasquale. Leave-in-Time: A New Service Discipline for Real-Time Communications in a Packet-Switching Network. In Proceedings of ACM SIGCOMM, pages 207 218. ACM Press, October 1995.

.... [4] Generalized Processor Sharing (or GPS) 23] and Rate Controlled Static Priority Queuing (or RCSP) 36] These mod els have also been used to derive statistical delay guarantees; in particular, within the framework of RCSP (in [37] and GPS (in [39] Related results can be found in [5] (using a policy like Virtual Clock [38] and in [34] for FCFS, with a variety of traffic distributions) In [14] statistical service quality objectives are achieved via proportional share queuing, in conjunction with server guided backoff, where servers dynamically adjust their rates to help ....

Norival R. Figueira and Joseph Pasquale. Leave-in-Time: A New Service Discipline for Real- Time Communications in a Packet-Switching Network. In Proceedings of ACM SIGCOMM, pages 207 218. ACM Press, October 1995.

.... include : D EDD (Delay Earliest Due Date) 1] J EDD (Jitter Earliest Due Date) 2] HRR (Hierarchical Round Robin) 3] S G (Stop and Go Queuing) 4] WFQ (Weighted Fair Queuing) 5] PGPS (Packet Generalized Processor Sharing) 6] RCSP (Rate Controlled Static Priority) 7] LIT (Leave In Time) [8]; MRTS (Multi Rate Traffic Shaping) 9] VC (Virtual Clock) 10] The traffic management scheme for real time applications investigated in this work may be viewed as an enhancement of the Rate Controlled Static Priority (RCSP) scheme proposed in [7] 2 The Proposed Dynamic R S Policy The ....

N. R. Figueira J. Pasquale. "Leave in Time: A New Service Discipline for Real Time Communications in a Packet Switching Nework". In Proceedings ACM SIGCOMM '95 Cambridge MA, pages 207--218, 1995.

....as well as available traffic policing mechanisms for enforcement. We discuss the advantages and disadvantages of each traffic characterization method. Section 3. 2, where we describe the operations and implementations of a number of packet schedulers and review their available delay bound tests [31, 32, 33, 40, 52, 63, 74, 82, 102, 107, 110, 111, 112]. We describe the properties of each packet scheduling discipline, emphasizing tradeoffs between achievable network utilization and implementation overhead costs. 3.1 Traffic Characterization A traffic characterization should conform to a parameterized traffic model that can be enforced by some ....

....3 The arriving packet is inserted into the sorted transmission queue according to F k i . Delay guarantees for VC schedulers were developed in [32, 107] for connections that conform to the (oe; ae) model. Two variants of the VC scheduling discipline are Burst Scheduling [63] and Leave inTime [33]. In Burst Scheduling, traffic is not assumed to be a sequence of packet arrivals but rather a sequence of packet bursts. While VC assigns a virtual finishing time to each individual packet, all packets in the same burst receive the same virtual finishing time in a Burst Scheduler. Bursts of ....

N. R. Figueira and J. Pasquale. Leave-in-Time: A New Service Discipline for RealTime Communications in a Packet-Switching Network. In Proc. ACM Sigcomm, pages 207--218, August 1995.

....(SIP) servers defined below. Definition 3 A server belongs to the Service rate Independent Priority (SIP) class if it belongs to the priority class, and in addition its packet priority computation is independent of the work done by the server. Virtual Clock [23] Delay EDD [7] and Leave in Time [8] servers belong to the SIP class. So do all the packet fair queueing servers that approximate a hypothetical constant rate Fluid model Fair Queueing (FFQ) server, which include WFQ [6] PGPS [18] and VSFQ [1] FIFO servers also belong to the SIP class since they in effect use the arrival time of ....

Norival R. Figueira and Joseph Pasquale. Leave-in-time: A new service discipline for real-time communications in a packet-switching network. In Proceedings of ACM SIGCOMM '95, pages 207--218, August 1995.

....should be chosen to run. Our reservation model decouples the delay and bandwidth provided to a flow by allowing a flow to specify a delay d different from T (as per the reservation parameters in section 3.2. 1) This can be done using a simple modification to the standard fair queueing formulation [15]. Besides per flow isolation, the second issue that the scheduler needs to address is that of guaranteeing a minimum level of performance to flows that do not go through the computation path. We mentioned earlier that core datapath functions, viz classification, forwarding and scheduling are ....

N. Figueira, J. Pasquale; "Leave-in-Time: A New Service Discipline for Real-Time Communications in a PacketSwitching Network"; Proc. ACM SIGCOMM 1995.

.... Scheduling an output link to ensure that the performance guarantees of ows contending for the link are satis ed, is a well studied problem in the literature [7] A well accepted scheduling model is based on the notion of weighted uid fairness [15] with decoupled delay and bandwidth guarantees [8]. Emulation of a uid fair scheduler is computationally expensive and may seriously limit the router s throughput, since it requires computation of a virtual deadline for each incoming packet and a sorting operation on these deadlines to select a packet to output. A class of self clocked schemes ....

N. Figueira, J. Pasquale; "Leave-in-Time: A New Service Discipline for Real-Time Communications in a Packet-Switching Network"; Proc. ACM SIGCOMM 1995.

.... an output link to ensure that the performance guarantees of flows contending for the link are satisfied, is a well studied problem in the literature [7] A well accepted scheduling model is based on the notion of weighted fluid fairness [15] with decoupled delay and bandwidth guarantees [8]. Emulation of a fluid fair scheduler is computationally expensive and may seriously limit the router s throughput, since it requires computation of a virtual deadline for each incoming packet and a sorting operation on these deadlines to select a packet to output. A class of self clocked schemes ....

N. Figueira, J. Pasquale; "Leave-in-Time: A New Service Discipline for Real-Time Communications in a Packet-Switching Network"; Proc. ACM SIGCOMM 1995.

....process a packet. Like the original FQ, DDR cant provide different bandwidth shares to active flows and, therefore, can only be used in best effort networks, where users are all treated equally. Leave in time (LiT) The LiT service discipline was proposed by Norival Figueira and Joseph Pasquale [Figueira 95] LiT can be regarded as a merge between virtual clock and jitter EDD, exploiting the good properties of each. LiT has a dual operation mode: as in J EDD, the discipline can operate in a non work conserving mode providing jitter upper bound for jitter sensitive traffic, and; as in VC, the ....

- Norival Figueira and Joseph Pasquale. Leave-in-Time: A New Service Discipline for Real-Time Communications in Packet Switching Networks. In Proceedings of the SigComm95 Conference, ACM, pp. 207-218, Cambridge, MA, USA, August 1995.

.... : D EDD (Delay Earliest Due Date) 1] J EDD (Jitter Earliest Due Date) 2] 3] HRR (Hierarchical Round Robin) 4] S G (Stop and Go Queuing) 5] WFQ (Weighted Fair Queuing) 6] PGPS (Packet Generalized Processor Sharing) 7] RCSP (Rate Controlled Static Priority) 8] LIT (Leave In Time) [9]; MRTS (Multi Rate Traffic Shaping) 10] VC (Virtual Clock) 11] The traffic management scheme for real time applications investigated in this work may be viewed as an enhancement of the Rate Controlled Static Priority (RCSP) scheme proposed in [8] Under the RCSP scheme each traffic stream ....

N.R.Figueira J.Pasquale. "Leave in Time: A New Service Discipline for Real Time Communications in a Packet Switching Nework". In Proceedings ACM SIGCOMM '95 Cambridge MA, pages 207--218, 1995.

.... include : D EDD (Delay Earliest Due Date) 1] J EDD (Jitter Earliest Due Date) 2] HRR (Hierarchical Round Robin) 3] S G (Stop and Go Queuing) 4] WFQ (Weighted Fair Queuing) 5] PGPS (Packet Generalized Processor Sharing) 6] RCSP (Rate Controlled Static Priority) 7] LIT (Leave In Time) [8]; MRTS (Multi Rate Traffic Shaping) 9] VC (Virtual Clock) 10] The traffic management scheme for real time applications investigated in this work may be viewed as an enhancement of the Rate Controlled Static Priority (RCSP) scheme proposed in [7] 2 The Proposed Dynamic R S Policy The typical ....

N. R. Figueira J. Pasquale. "Leave in Time: A New Service Discipline for Real Time Communications in a Packet Switching Nework". In Proceedings ACM SIGCOMM '95 Cambridge MA, pages 207--218, 1995.

....issue of non work conserving service disciplines to reduce burstiness of traffic inside the network has been addressed in scheduling mechanisms like jitter EDD [8] HRR [9] Stop and Go [7] and RCSP [10] Among these, decoupling of delay and bandwidth bounds can be achieved in jitterEDD and RCSP. [12] also addresses the problem of delay bandwidth decoupling by incorporating a delay parameter in the priority calculation equations of WFQ. The priority of transmission is determined based on an eligibility time and a delay bound. Eligibility times are controlled through the bandwdith reservation ....

....architecture but is beyond the scope of this paper. The above scheduling and MUDEM algorithms still work in this case, except that the receiver s connection ID output queue map should allow multiple link layer connections to be mapped to the same per network connection output queue. The work in [12] assigns connections to delay classes, where each delay class is characterized by a rate, which is the maximum amount of bandwidth that can be allocated to connections in that class, and a base delay, which is a measure of the delay that lower priority classes may suffer because of the higher ....

[Article contains additional citation context not shown here]

Norival, R.F. and Pasquale, J., Leave-In-Time : A New Service Discipline for Real Time Communications in a Packet Switching Network, In Proceedings of ACM SIGCOMM 95, Cambridge, September 1995.

....(GD) server, together with an appropriate admission control mechanism. Each packet arrival from a real time flow to a GD server is given a deadline, or priority value, and the server ensures that the packet departs by its deadline. 1 Many GD service disciplines have been proposed [1, 2, 3, 4, 10, 11, 15]. In this paper, we assume that packets in the same real time flow are served in FIFO order. 2 At a GD server, there is a packet scheduler which repeatedly searches for the smallest value in a set of priority values, one for each flow. The priority value of a flow is the priority value of its ....

....f being the same for all j. If certain schedulablity conditions are met, then a Delay EDD server provides the guaranteed deadline in (1) with fi k = 0 [2] By induction, it is easy to show that for all j 1 P f k (j) Gamma V f k (j) d f k Gamma v f (11) For a leave in time server [3], the P values of packets are computed as follows 5 for j 1, P f k (j) maxfA f k (j) V f k (j Gamma 1)g d f k (j) 12) V f k (j) maxfA f k (j) V f k (j Gamma 1)g v f (j) 13) where V f k (0) 0, d f k (j) is the local delay bound of packet j, and v f (j) s f ....

[Article contains additional citation context not shown here]

Norival R. Figueira and Joseph Pasquale. Leave-in-time: A new service discipline for real-time communications in a packet-switching network. In Proceedings of ACM SIGCOMM '95, pages 207--218, August 1995.

....Research supported in part by National Science Foundation grants no. NCR 9004464 and NCR9506048. An abbreviated version to appear in Proceedings IEEE INFOCOM 96. called priority) and the server ensures that the packet departs by the deadline. Many GD service disciplines have been proposed [1, 2, 3, 4, 8, 12]. For most applications, a flow is generated by its source as a sequence of messages, each of which is then segmented and transported as a sequence of packets by the network. 1 To these applications, the end to end delay of a message is a more important performance measure than the end to end ....

....and f being the same for all j. If certain schedulablity conditions are met, then a Delay EDD server provides the guaranteed deadline in (1) with fi k = 0 [2] By induction, it is easy to show that for all j 1 P f k (j) Gamma V f k (j) d f k Gamma v f (11) For a leave in time server [3], the P values of packets are computed as follows 2 for j 1, P f k (j) maxfA f k (j) V f k (j Gamma 1)g d f k (j) 12) V f k (j) maxfA f k (j) V f k (j Gamma 1)g v f (j) 13) where V f k (0) 0, d f k (j) is the local delay bound of packet j, and v f (j) s f ....

[Article contains additional citation context not shown here]

Norival R. Figueira and Joseph Pasquale. Leave-in-time: A new service discipline for real-time communications in a packet-switching network. In Proceedings of ACM SIGCOMM '95, pages 207--218, August 1995.

.... [4] Generalized Processor Sharing (or GPS) 23] and Rate Controlled Static Priority Queuing (or RCSP) 36] These models have also been used to derive statistical delay guarantees; in particular, within the framework of RCSP (in [37] and GPS (in [39] Related results can be found in [5] (using a policy like Virtual Clock [38] and in [34] for FCFS, with a variety of traffic distributions) In [14] statistical service quality objectives are achieved via proportional share queuing, in conjunction with server guided backoff, where servers dynamically adjust their rates to help ....

Norival R. Figueira and Joseph Pasquale. Leave-in-Time: A New Service Discipline for RealTime Communications in a Packet-Switching Network. In Proceedings of ACM SIGCOMM, pages 207--218. ACM Press, October 1995.

.... service in high speed networks, e.g. the VirtualClock [25] Fair Share Queueing [7] GPS [17] and RCSP [23] These models have also been used to derive statistical delay guarantees; in particular, within the framework of RCSP (in [24] and GPS (in [26] Related results can be found in [9] (for VitualClock) and in [22] for FCFS, with a variety of traffic distributions) In [14] statistical service quality objectives are achieved via proportional share queueing, in conjunction with server guided backoff, where servers dynamically Corner Turn Azi. DFT 114 MFLOPS 14 MFLOPS Azi. ....

....it gets reconfigured. When a resource fails, an executive process activates the redundant software components whose design was, after all, predetermined before the system was deployed. CPU Tasks Derived From E[t] ms) Var[t] Min,Max] ms) NumSteps I Q Normal 59 100 [50,83] 33 EQ Normal 11 16 [9,24] 15 RDFT Exponential 103 [87 ,140] 53 CT Exponential 14 [12,30] 18 ADFT Exponential 187 [160,210] 50 KM Exponential 23 [20,35] 15 AIDFT Exponential 187 [160,210] 50 I Q,EQ Normal 70 144 [56,107] 51 KM, AIDFT Exponential 210 [180 , 245] 50 Network Tasks Derived From E[t] ms) Var[t] Min,Max] ms) ....

Norival R. Figueira and Joseph Pasquale. Leave-in-Time: A New Service Discipline for Real-Time Communications in a Packet-Switching Network. In Proceedings of ACM SIGCOMM, pages 207--218. ACM Press, October 1995.

.... Sharing (or GPS) 13] and Rate Controlled Static Priority Queueing (or RCSP) 22] These models have also been extended to derive statistical guarantees; in particular, within the framework of RCSP (in [23] and for GPS (in [25] Related results for real time traffic sources can be found in [3] (using a policy like Virtual Clock [24] and in [21] using a modified FCFS method, with a variety of traffic distributions) In [6] statistical service quality is achieved via proportional share queueing, in conjunction with server guided backoff; i.e, servers dynamically adjust their rates ....

Norival R. Figueira and Joseph Pasquale. Leave-in-Time: A New Service Discipline for Real-Time Communications in a Packet-Switching Network. In Proceedings of ACM SIGCOMM, pages 207--218. ACM Press, October 1995.

....by implementing each communication channel as a guaranteed deadline (GD) server. At a GD server, each packet arrival from a guaranteed flow is given a deadline (also called priority) and the server ensures that the packet departs by the deadline. Many GD service disciplines have been proposed [1, 2, 3, 4, 9, 13]. For most applications, a flow is generated by its source as a sequence of messages, each of which is then segmented and transported as a sequence of packets by the network. 1 To these applications, the end to end delay of a message is a more important performance measure than the end to end ....

....f being the same for all j. If certain schedulablity conditions are met, then a Delay EDD server provides the guaranteed deadline in (1) with fi k = 0 [2] By induction, it is easy to show that for all j 1 P f k (j) Gamma V f k (j) d f k Gamma v f (11) For a leave in time server [3], the P values of packets are computed as follows 2 for j 1, P f k (j) maxfA f k (j) V f k (j Gamma 1)g d f k (j) 12) V f k (j) maxfA f k (j) V f k (j Gamma 1)g v f (j) 13) where V f k (0) 0, d f k (j) is the local delay bound of packet j, and v f (j) s f ....

[Article contains additional citation context not shown here]

Norival R. Figueira and Joseph Pasquale. Leave-in-time: A new service discipline for real-time communications in a packet-switching network. In Proceedings of ACM SIGCOMM '95, pages 207--218, August 1995.

....by implementing each communication channel as a guaranteed deadline (GD) server. At a GD server, each packet arrival from a guaranteed flow is given a deadline (also called priority) and the server ensures that the packet departs by the deadline. Many GD service disciplines have been proposed [1, 2, 3, 4, 9, 13]. For most applications, a flow is generated by its source as a sequence of messages, each of which is then segmented and transported as a sequence of packets by the network. 1 To these applications, the end to end delay of a message is a more important performance measure than the end to end ....

....f being the same for all j. If certain schedulablity conditions are met, then a Delay EDD server provides the guaranteed deadline in (1) with fi k = 0 [2] By induction, it is easy to show that for all j 1 P f k (j) Gamma V f k (j) d f k Gamma v f (11) For a leave in time server [3], the P values of packets are computed as follows 2 for j 1, P f k (j) maxfA f k (j) V f k (j Gamma 1)g d f k (j) 12) V f k (j) maxfA f k (j) V f k (j Gamma 1)g v f (j) 13) where V f k (0) 0, d f k (j) is the local delay bound of packet j, and v f (j) s f ....

[Article contains additional citation context not shown here]

Norival R. Figueira and Joseph Pasquale. Leave-in-time: A new service discipline for real-time communications in a packet-switching network. In Proceedings of ACM SIGCOMM '95, pages 207--218, August 1995.

No context found.

Figueira, N., and Pasquale, J. Leave-in-time: A new service discipline for real-time communications in a packet-switching network. In ACM SIGCOMM '95 (1995).

....is I O centric. In peer to peer I O, mechanisms are provided that allow the construction of in kernel data paths, directly connecting source and sink devices [48] Leave in Time, a new packet switching service discipline that provides distributional end toend performance (e.g. delay) bounds [50], building on new results for performance bounds of other service disciplines that were unknown, and in some cases, believed to not exist [51] 38 . Software implementation techniques for improving throughput and latency in TCP IP, based on one of the most detailed studies of where time is spent ....

N. Figueira and J. Pasquale, "Leave-in-time: a new service discipline for real-time communications in a packet-switching network," In Proc. ACM Comm. Archit. and Protocols Conf. (SIGCOMM), Cambridge, MA, Sept. 95.

No context found.

N.R.Figueira and J.Pasquale, "Leave-in-time: a new service discipline for real-time communications in a packet-switching network", Proc. of SIGCOMM'95, pp.207-18, 1995.

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