Packet loss characteristics for M/G/1/N queueing systems

Fiems, Dieter; Vuyst, Stijn De; Wittevrongel, Sabine; Bruneel, Herwig

doi:10.1007/s10479-008-0436-9

Cited by 9 publications

(4 citation statements)

References 20 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Other related "loss metrics" include the loss period (the difference between the arrival times of the last and first in a series of consecutively lost packets); its behavior was addressed by Fiems et al [21], in an M/G/1/K queue, in terms of a joint transform with the number of losses within such a period. Another is the block loss probability (the fraction of lost packets within a block of consecutive arrivals); a recursive formula was obtained by Cidon et al [16] in an IPP/M/1/K queue, whereas an explicit expression was later derived by Gurewitz et al [25] using ballot theorems in the M/M/1/K case; for a discussion of applications of such results to FEC schemes see [21]. Lastly, we mention the number of lost packets in a busy period (loss as well as blocking periods are sub-intervals of busy periods); asymptotic properties (in K) were obtained by Abramov [1].…”

Section: Loss Distribution and Distance And Other Related Metricsmentioning

confidence: 99%

Queue and Loss Distributions in Finite-Buffer Queues

CiucuFlorin

PoloczekFelix²,

RizkAmr

2019

Proc. ACM Meas. Anal. Comput. Syst.

View full text Add to dashboard Cite

We derive simple bounds on the queue distribution in finite-buffer queues with Markovian arrivals. Our technique relies on a subtle equivalence between tail events and stopping times orderings. The bounds capture a truncated exponential behavior, involving joint horizontal and vertical shifts of an exponential function; this is fundamentally different than existing results capturing horizontal shifts only. Using the same technique, we obtain similar bounds on the loss distribution, which is a key metric to understand the impact of finite-buffer queues on real-time applications. Simulations show that the bounds are accurate in heavy-traffic regimes, and improve existing ones by orders of magnitude. In the limiting regime with utilization ρ=1 and iid arrivals, the bounds on the queue size distribution are insensitive to the arrivals distribution.

show abstract

Section: Loss Distribution and Distance And Other Related Metricsmentioning

confidence: 99%

Queue and Loss Distributions in Finite-Buffer Queues

CiucuFlorin

PoloczekFelix²,

RizkAmr

2019

Proc. ACM Meas. Anal. Comput. Syst.

View full text Add to dashboard Cite

show abstract

Section: Loss Distribution and Distance And Other Related Metricsmentioning

confidence: 99%

Queue and Loss Distributions in Finite-Buffer Queues

Ciucu

Poloczek²,

Rizk

2019

SIGMETRICS Perform. Eval. Rev.

View full text Add to dashboard Cite

We derive simple bounds on the queue distribution in finite-buffer queues with Markovian arrivals. The bounds capture a truncated exponential behavior, involving joint horizontal and vertical shifts of an exponential function; this is fundamentally different than existing results capturing horizontal shifts only.We also obtain similar bounds on the loss distribution, which is a key metric to understand the impact of finite-buffer queues on real-time applications. Simulations show that the bounds are accurate in heavy-traffic regimes, and improve existing ones by orders of magnitude.

show abstract

“…Since the number of newly incoming messages is larger than amount of processed and sent messages, the number of messages delayed in the receiving queue (waiting to be processed and forwarded) will continuously increase. Because a network system has finite buffer memory to hold the queue and cannot expand the buffer memory indefinitely, queue congestion occurs, and the queue becomes full as a result [37]. In this case, the system has no other option than to simply discard excess packets (i.e., packet loss).…”

Section: Throughput and Message Loss Ratiomentioning

confidence: 99%

A High-Performance Implementation of an IoT System Using DPDK

Pak

Park

2018

Applied Sciences

View full text Add to dashboard Cite

An IoT (Internet of Things) system typically encompasses a number of devices and sensors and is required to process a large number of messages at a high speed. To address this requirement, we propose a dual plane architecture, which consists of a control plane and a data plane. The control plane processes signaling messages and the data plane takes charge of processing user data messages. This allows the system to process messages separately and simultaneously in the different planes according to the type of incoming message. In this paper, we present the each plane's role and how messages are processed in the different planes. We also present the interworking method between both planes. To verify the proposed architecture, we implement and apply the architecture to our previous single plane IoT system. We also compare the performance of the proposed system with that of the single plane IoT system in terms of throughput and packet loss ratio. The results reveal that the performance of the proposed architecture is much higher than that of the previous single plane IoT systems. The results prove that the proposed architecture is highly appropriate for IoT environments.

show abstract

Packet loss characteristics for M/G/1/N queueing systems

Cited by 9 publications

References 20 publications

Queue and Loss Distributions in Finite-Buffer Queues

Queue and Loss Distributions in Finite-Buffer Queues

Queue and Loss Distributions in Finite-Buffer Queues

A High-Performance Implementation of an IoT System Using DPDK

Contact Info

Product

Resources

About