Pseudo-isochronous cell forwarding

doi:10.1016/S0169-7552(98)00258-X

Computer Networks and ISDN Systems

Volume 30, Issue 24, 14 December 1998, Pages 2359-2372

Computer Networks an...

https://doi.org/10.1016/S0169-7552(98)00258-X Get rights and content

Abstract

This paper shows how to design a packet switched network, for real-time traffic, such that under full network load: (i) the end-to-end delay bound of a low-rate voice connection is minimized, (ii) the bound on the delay uncertainty or jitter is a fixed network parameter – independent of the network size and the connection rate, and (iii) the required buffer sizes (inside the network) to ensure congestion-free routing is minimized. In addition, this design can be generalized to accommodate either variable bit rate (VBR) traffic with statistical multiplexing or the integration of available bit rate (ABR) traffic 7, 12.

The isochronous timing information (can be provided by the global positioning system (GPS) [1]) is used for pacing the packet/cell forwarding inside the network. This means that a cell is forwarded from one switch to another not at a specific time but within a time frame of a relatively long duration as compared with the cell transmission time. This time frame is an independent network parameter, which determines the delay and jitter bounds inside the network.

A study of the blocking probability of this approach is presented. (Blocking is defined as the impossibility of allocating bandwidth for a new connection while capacity is still available, but not in the proper time frames.) The study includes both analytical and simulation results, which demonstrate an important trade-off between the blocking probability and the end-to-end delay bound.

Introduction

This work deals with means to ensure quality-of-service (QoS), such as end-to-end delay and jitter, of real-time connections, in a deterministic fashion. The proposed method also minimizes the buffering requirements inside the network switches, while ensuring no cell loss due to congestion.

The approach taken in this work can be viewed as a compromise between asynchronous packet switching and the traditional circuit switching. The idea is to increase the amount of timing information used for flow control inside the network, in order to achieve the quality-of-service objectives. However, in this method cells are not transmitted or forwarded in a specific time, but rather within a predefined time frame of a relatively long duration. Thus, this method is called pseudo-isochronous cell switching. The pseudo-isochronous method was originally proposed in the context of the Synchronous Optical Hypergraph 8, 9.

It has been shown that pure cell or packet switching, with well-controlled nominal rate at the boundary of the network, may become bursty inside the network. Burstyness increases the buffering requirements and can significantly increase the delay and jitter. This can result with cell loss probability that increases as the size of the network increases and as the capacity allocated to individual connections increases. The main consequence of the above is that the quality-of-service, as experienced by the end-user, may not be predictable. This may not be acceptable for applications, such as telephony, that are currently implemented over a circuit switched network, but we would like to implement them over packet switched networks.

The most notable property of the pseudo-isochronous cell switching is that the QoS parameters, delay, jitter and buffer space, are independent of the individual connection rate and the instantaneous traffic pattern. Furthermore, under full link and network load for time frame of 125 μs (8 KHz):

1.
The end-to-end delay is bounded and is equal to 250 μs times the number of hops plus the propagation delay,
2.
The bound on the delay uncertainty or jitter is a fixed network parameter of 125 μs, is independent of the network size, and
3.
The required buffer size for congestion-free routing is reduced significantly (in comparison with asynchronous methods), is equal to 250 μs times the link bandwidth, and is independent of the number and rates of the connections over that link.

In contrast to pseudo-isochronous cell switching with global timing, there are two types of traffic shaping schemes that are using only local timing: (i) without deadline scheduling, for example 4, 2, 13, and (ii) with deadline scheduling, for example 3, 5. In traffic shaping without deadline scheduling, in order to obtain high utilization with no loss, the delay and jitter are inversely proportional to the connection bandwidth and proportional to the number of hops, which means that low rate connections may experience large delay and jitter inside the network. (See Section 6for more details.) In traffic shaping with deadline scheduling the delay and jitter are controlled at the expense of either low utilization or possible congestion and loss.

The pseudo-isochronous cell switching is discussed in Section 2. At the end of Section 2we discuss the differences between pseudo-isochronous cell switching and circuit switching. Two scheduling algorithms are presented in Section 3. A study of the blocking probability of this novel cell forwarding approach is presented in Section 4. (Blocking is defined as the impossibility of allocating bandwidth for a new connection while capacity is still available, but not in the proper time frames.) The study includes also simulation results, in Section 5, which demonstrate an interesting trade-off between the blocking probability and the end-to-end delay bound. For example, it is shown how by increasing the end-to-end delay bound it is possible to decrease the blocking probability. The work is concluded in Section 6with a more detailed comparison with related schemes, statistical multiplexing and quality-of-service.

Section snippets

System model and principles

We consider an arbitrary topology network with bidirectional links. Data units are sent over this network in fixed size cells, as in ATM networks. The payload size of each cell is k data bytes (e.g., k=48 in ATM networks).

The cells of a given connection are routed in the network, from the source station s₁ to the destination station s_h, where h≥2. The intermediate stations are $s_{2}, s_{3},…,s_{h−1}$ , and the routing links are $s_{1} →s_{2}, s_{2} →s_{3}, …, s_{h−1} →s_{h}$ .

Distributed scheduling algorithms

In this section we will present two distributed algorithms that can be used for determining the transmission schedules, from source to destination, of a given connection. The first algorithm is designed for the immediate forwarding method, and the second algorithm is designed for the 2-frame choice and arbitrary-frame choice methods.

The main operation criterion for selecting the schedules that is used in the three algorithms, for selecting the schedules is load balancing. This means that the

Blocking probabilities

In this section we show how to compute the call blocking probability for the various forwarding schemes. Blocking is defined as the impossibility of allocating bandwidth for a new connection while capacity is still available, but not in the proper time frames. It will be shown that the main gain in blocking probabilities is in the transition from the immediate forwarding to 2-frame forwarding.

The blocking probability is tightly coupled with the way the traffic load is distributed inside the

Numerical results

In this section we present some numerical examples in order to illustrate the analytical results shown in Table 1 (Section 4). The analytical computation of 2-frame forwarding requires very large memory for storing the elements of the transition probability matrix P (this matrix is of dimension 2^k×2^k), therefore, the exact numerical results are presented for k=5 (and h=5), as shown in Fig. 4.

Fig. 4 depicts the blocking probability versus p, the probability that a frame is nearly fully used.

Discussion

Extensive work has been done in asynchronous packet switching in the past decade, nevertheless many quality of service design issues have not been solved. In this work we use the concept of pseudo-isochronous cell forwarding, which provides deterministic quality of service guarantees for real-time applications, even if all the links in the network are almost fully utilized (see [7] on how this can be achieved over IP networks). This property is important since it is not realistic to assume that

Acknowledgements

We would like to thank Moti Yung for many useful discussions along the development of this new architectural concept.

References (14)

P.H. Dana, Global positioning system overview, http://www.utexas.edu/depts/grg/gcraft/notes/gps/gps.html,...
A. Demers, S. Keshav, S. Shenker, Analysis and simulation of a fair queueing algorithm, ACM Computer Communication...
D. Ferrari, D.C. Verma, A scheme for real-time channel establishment in wide-area networks, IEEE J. Selected Areas in...
S.J. Golestani, Congestion-free communication in high-speed packet networks, IEEE Trans. Comm. COM-39 (12) (December...
D.D. Kandlur, K.G. Shin, D. Ferrari, Real-time communication in multi-hop networks, IEEE Trans. Parallel and...
M.G.H. Katevenis, Fast switching and fair control of congested flow in broadband networks, IEEE J. Selected Areas in...
C-S. Li, Y. Ofek, M. Yung, Time-driven priority flow control for real-time heterogeneous internetworking, IEEE...

There are more references available in the full text version of this article.

Cited by (30)

A scalable solution for engineering streaming traffic in the future Internet
2007, Computer Networks
Citation Excerpt :
However, results in [15] and [4] show that both TDP and TDS can achieve very high utilization with immediate forwarding; hence, only a small fraction of flows possibly needs to be subjected to the extra non-immediate forwarding delay. Buffer size requirements for non-immediate forwarding have been studied in [12]. Pipeline forwarding enables a scalable multi-service network able to provide guaranteed QoS and support to efficient traffic engineering, while providing full statistical multiplexing of elastic traffic.
As traffic on the Internet continues to grow exponentially, there is a real need to solve scalability and traffic engineering simultaneously – specifically, without using over-provisioning in order to accommodate streaming media traffic. One of the threats to the current operation stability of the Internet comes from UDP-based streaming media applications, such as Skype (which is currently doubling every 6-month) today and video services in the near future. This paper shows how the Internet can benefit from pipeline forwarding in order to: (i) construct ultra-scalable IP switches, (ii) provide predictable quality of service for UDP-based streaming applications, while (iii) preserving elastic TCP-based traffic as is, i.e., without affecting any existing best-effort applications.
Availability of end-to-end ideal QoS in IP packet networks
2005, Computer Communications
Citation Excerpt :
Although analysis for the homogeneous case is analogous to that used in the design of TDM digital telephony systems, it has rarely been applied in the context of packet switching networks, probably because many popular implementations use a connectionless architecture. Ofek and co-workers [3,4] discuss blocking probability in time-driven priority packet networks [5], where the granularity is at the level of a time interval (typically 125 μs). The model is analogous to the digital telephony switching model and the homogenous packet-sequencing model, albeit at different granularities.
Network processor technology has advanced to the point where high-precision time-based store-and-forward logic is readily incorporated into packet switches and routers. With appropriate scheduling, packets from multiple flows can be serviced without contending for link resources. Accordingly, packet flows traversing a network of switching elements can have both path and time determinacy attributes which support ideal end-to-end QoS (zero jitter, zero loss, acceptable end-to-end latency) for real-time UDP packet flows and guaranteed goodput for TCP flows. One approach to packing a network with a relatively large number of such deterministic flows, i.e. achieving high availability of the ideal QoS service in a network, uses precise buffering of packets at each switch, which introduces latency. This paper describes analysis methods for quantifying how much buffering may be necessary to achieve high (99.999%) availability. For typical network topologies the analysis shows that buffering latency requirements are very small compared to transport delays, even when the network is highly utilized with heterogeneous (e.g. voice, video, circuit emulation, and data) traffic. Actual physical implementations have empirically validated the analysis results as well as the scalability of the end-to-end, time-based forwarding approach and the end-to-end availability of ideal QoS services in IP packet networks.
Combined Asynchronous/Synchronous Packet Switching Architecture: QoS Guarantees for Integrated Parallel Computing and Real-Time Traffic
2000, Journal of Parallel and Distributed Computing
This paper focuses on the challenging problem of integrating real-time traffic with data traffic on the same network, while having predictable quality of service (QoS) guarantees. Predictable QoS guarantees mean both deterministic and probabilistic (with a known probability distribution) functions of loss, delay, and jitter. As of today there are only limited solutions to this problem. This work presents, to the best of our knowledge, the first coherent system solution, for combining, on one hand, bursty data traffic with deterministic no loss due to congestion, and on the other hand, periodic real-time traffic with deterministic bandwidth guarantees, constant jitter, and bounded delay. The principles of this architecture facilitate the implementation of a scalable multimedia system that will combine or integrate distributed/parallel computing (e.g., over a network of PC/workstations) with real-time applications (e.g., interactive video teleconferencing).
End-to-end delay analysis of videoconferencing over packet-switched networks
2000, IEEE/ACM Transactions on Networking
Efficient slice-aware H.264/AVC video transmission over time-driven priority networks
2014, International Journal of Communication Systems
Design of a deterministic service switch for avionics networks
2014, Journal of Aerospace Information Systems

View all citing articles on Scopus

Chung-Sheng Li received his B.S.E.E. degree from National Taiwan University, Taiwan, R.O.C. in 1984, and the M.S. and Ph.D. degree in electrical engineering and computer science from the University of California, Berkeley in 1989 and 1991, respectively. He has joined the computer science division of IBM T.J. Watson Research Center as a research staff member since September 1991, and manages the Image Information System Department since 1996. His research interests include (1) Broadband applications, which include digital library, knowledge discovery and data mining; (2) Broadband network and switching, which includes all-optical networks, storage area networks, and fiber channel; (3) Broadband technologies, which include optical chip interconnects, optoelectronics, and high-speed analog/digital VLSI circuit design. He has co-initiated several research activities in IBM on fast tunable receiver for all-optical networks and content-based retrieval in the compressed domain for large image/video databases. He is currently the principle investigator of a satellite image database project funded by NASA. Dr. Li has received a Research Division award from IBM in 1995 for his major contribution to the tunable receiver design for WDMA, and numerous invention and patent application awards. He is serving as the technical editor and feature editor for the IEEE Communication Magazine. He has authored or coauthored more than 70 journal and conference papers and received one of the best paper awards from the IEEE International Conference on Computer Design in 1992. He is a senior member of the IEEE Laser Electro-Optic Society, the Communication Society, the Computer Society, and the Circuit and System Society.

Y. Ofek is the president and CTO of Synchrodyne, Inc.. He received his B.Sc. degree in electrical engineering from the Technion, Israel Institute of Technology in 1979, and his M.Sc. and Ph.D. degrees in electrical engineering from the University of Illinois-Urbana in 1985 and 1987, respectively. From 1979 to 1982 he was research engineer at RAFAEL, Haifa, Israel. From 1983 to 1984 he was at Fermi National Accelerator Laboratory, Batavia, Ill., and from 1984 to 1986 he was with Gould Electronics, Urbana, Ill. From 1987 to 1998 he was with IBM T.J. Watson Research Center, Yorktown Heights, N.Y. Dr. Ofek was the program co-chairperson of the 6th and chair of the 7th IEEE Workshop on Local and Metropolitan Area Networks, he served on various program committees and was a guest editor in several journals. Dr. Ofek has invented, initiated, and managed the research activities of four novel network architectures: (1) Ring networks with spatial bandwidth reuse with a family of fairness algorithms, MetaRing, which is used as the underlying network for SSA (Serial Storage Architecture – ANSI Standard X3T10) and several IBM products, (2) Optical hypergraph for combining multiple passive optical stars with novel conservative code for bit synchronization, pseudo-isochronous flow control and global clock synchronization techniques, (3) Embedding of virtual rings in arbitrary topology network, MetaNet, for bursty data traffic with no packet loss, fairness and reliable/real-time broadcast/multicast, and (4) Global network solution for real-time traffic, which utilizes either GPS-based synchronization or a novel internal real-time clocks for providing deterministic quality of service (QoS) guarantees – as in circuit switching.

Adrian Segall was born in Rumania in 1944. He received the B.Sc. and M.Sc. degrees in Electrical Engineering from the Technion, Israel Institute of Technology in 1965 and 1971, respectively, and the Ph.D. degree in Electrical Engineering with a Minor in Statistics from Stanford University in 1973. After serving active duty in the Israel Defense Army, he joined in 1968 the Scientific Department of Israel's Ministry of Defense. From 1973 to 1974 he was a Research Engineer at System Control Inc., Palo Alto, CA and a Lecturer at Stanford University. From 1974 to 1976 he was an Assistant Professor of Electrical Engineering and Computer Science at the Massachusetts Institute of Technology. From 1987 to 1998 he was on the faculty of the Department of Computer Science at the Technion. He is presently Benjamin Professor of Computer-Communication Networks in the Department of Electrical Engineering, Technion, Israel Institute of Technology. From 1982 to 1984 he was on leave with the IBM T.J.Watson Research Center, Yorktown Heights, NY. His current research interests are in the area of high-speed communication networks, optical networks, TCP/IP protocols, ATM, wireless and Ad-Hoc Networks. Dr. Segall has served in the past as Editor for Computer Communication Theory of the IEEE Transactions on Communications and Editor for the IEEE Information Theory Society Newsletter. He was selected as an IEEE delegate to the 1975 IEEE–USSR Information Theory Workshop, and is the recipient of the 1981 Miriam and Ray Klein Award for Outstanding Research and of the 1990 Taub Award in Computer Science.

Khosrow Sohraby received B.Eng and M.Eng degrees from McGill University, Montreal, Canada in 1979 and 1981, respectively, and Ph.D degree in 1985 from the University of Toronto, Toronto, Canada, all in Electrical Engineering. His current research interests include design, analysis and control of high speed computer and communications networks, traffic management and analysis, multimedia networks, networking aspects of wireless and mobile communications, and large-scale computations applied to computer and communications networks. During 1984–1986 he was a research associate at the L'institute national de la recherche scientifique – INRS Telecommunications, Montreal, Canada. In 1986 he joined AT&T Bell Laboratories, Holmdel, NJ as a Member of Technical Staff in the Teletraffic Theory and System Performance Analysis Department. In 1989 he joined IBM T.J. Watson Research Center, Yorktown Heights, NY as a Research Staff Member in the Communications Systems Department. While at IBM research, he was an adjunct faculty at Polytechnic University and Columbia University in the Department of Electrical Engineering teaching undergraduate and graduate courses in telecommunications. He joined the Computer Science Telecommunications Porgram at the University of Missouri-Kansas City in 1994 as professor. He is an active member of IEEE Communications Society and has served as the guest editor of number of special issues of the Journal of Selected Areas in Communications and Communications Magazine. He is a member of editorial board of Computer Networks and ISDN Systems, Wireless Networks Journal, International Journal of Wireless Networks, Network Magazine and Mobile Computing and Communications Review. Dr. Sohraby has served on the technical program committee of many IEEE conferences and workshops. He served as the technical program vice-chair of IEEE INFOCOM '94 and the technical program chair of IEEE INFOCOM '95 and the chair of 10th Annual IEEE Workshop on Computer Communications. Dr. Sohraby was a consultant to Lucent Technology, Bell Labs during the summer of 1996.

View full text