Download 3.1 Active network

2nd Assignment of ‘Research Planing Course’
A report of my research and related work
1. Project and research focus

Finished Project---the KKS-financed project "Energy- and EngineerEfficiency in Parallel Architectures and Application Development for
Future Embedded Signal Processing Systems (EEE)".
One goal is to identify which possible future processor and system
architectures are best suited to meet the performance, scalability,
viability and energy-efficiency demands in next-generation high-speed
signal processing systems. The other goal is to study how increased
functionality in the system- and communication-architecture and the
associated development environment can support efficiency and
usability in application development.
The system architecture and communication architecture are addressed
in this project, where my work focuses on studying new solutions to be
used in network architectures to improve functionality, for example,
add some intelligence into the Ethernet switch to get higher efficiency
for certain traffic patterns with real-time constraints.

Current Project---"Real-Time Multi-service System Area Networks (RTSAN)", financed by CERES research center.
Our approach is to obtain more general results in the area of real-time
multi-service system area networks, to find communication solutions
that meet the increasing demands in embedded applications.
My work is mainly about investigating network architectures, protocols
and methods to support the different services required in these system
area networks.

Future work---"Heterogeneous Communication Services in Embedded
Network"
We will study the support of many heterogeneous communication
services in embedded systems. I will focus on developing and analyzing
how methods to support heterogeneous real-time traffic can be
implemented in switches and network interface to get as much
functionality and performance as possible.
2. Methods and main results
Industry has strong interesting to build their systems on top of standard
Ethernet based networks. Therefore, in my research, I investigate how to form
method to extend Switched Ethernet networks to offer additional features for
parallel and distributed real-time processing.
Since provision of communication services with active networking technology
appears to be promising idea, the active Ethernet switch concept is used to
achieve our approach. The active networking is a novel concept, to provide
application-specific functionality within the network. The role of computations
in traditional data networks is limited, e.g., header processing in packetswitched networks and signaling in connection-oriented networks. The
networks are active in the sense that the nodes or the switches can perform
customized computations on, and modify, the packet content flowing through
them. Therefore, good range of features and abilities can be achieved by this
active device, sometimes called intelligent device.
We enhance the switch with some intelligence to provide efficient support for
different user services, including many-to-many communication and other
group communication services with high traffic-volumes of short messages.
Meanwhile, the real-time support is addressed for these special
communication patterns by incorporating deadline-based scheduling in the
switch and the end nodes.
Moreover, the deadline sorting in the switch brings high overhead, which
motivate us to find another solution to addresses real-time services. We
proposed an alternative solution, in which Earliest Deadline First (EDF)
algorithm is only used in the source nodes to support real-time traffic with
guaranteed bit rate and end-to-end worst-case delay bound. Meanwhile, the
differentiation of heterogeneous traffic is considered in the proposed system
by prioritization traffic into several classes with distinctly different QoS
levels.
The performance of the proposed methods is evaluated through simulations
and mathematical analysis. It is shown that our different Ethernet
extensions, in many cases, are efficient choices for distributed computing
systems.
3. Related work
The research questions stated in Section 2 have received attention in
networking and parallel and distributed computing research communities in
recent years. The key conferences include: IPDPS (International Parallel &
Distributed Processing Symposium), ECRTS (Euromicro Conference on Realtime Systems), RTSS (Real-time Systems Symposium). There are many active
research group in this area, for example, the real-time lab in the University of
Michigan (Prof. Kang Shin) and University of Virginia and some research
group in Europe.
There have been several attempts to achieve the suitability of the networks
for future parallel and distributed real-time systems. To our knowledge, most
of these networks, commercially available or reported in the research
literatures, fail to offer real-time services or do not discuss the concept of
multiple services.
3.1
Active network
The active networking is a novel concept, to provide application-specific
functionality within the network [Tennenhouse and Wetherall 2002]. The role
of computations in traditional data networks is limited, e.g., header processing
in packet-switched networks and signaling in connection-oriented networks.
The networks are active in the sense that the nodes or the switches can
perform customized computations on, and modify, the packet content flowing
through them. Therefore, good range of features and abilities can be achieved
by this active device, sometimes called intelligent device.
There are mainly two distinguished approaches to active networks, discrete
and integrated. The programmable switch is used in the discrete approach.
When a packet arrives, its header is examined and a program is dispatched to
operate on its contents. The program actively processes the packet, possibly
changing its contents. In the integrated approach, every message is a program,
probably followed by user data.
A lot of research effort is put in the development of active networking
[Wetherall et al. 1998] [Yemini and Silva1996] [Dharmalingam and Collier
2001]. An overview of research on active networking can be found in
[Tennenhouse et al. 1997]. However, most of these results have contributed in
the software technologies, such as development of languages, compiler and
platforms suitable for active network, or development of middleware services
to demonstrate active network capabilities, and have been mainly focused on
offering the application specific functions.
Provision of communication services with active networking technology
appears to be promising idea. Different research work shows how basic
multicast control mechanisms could benefit from the application of both
capsules [Wetherall et al. 1998] and programmable switches approaches
[Farber et al. 1996]. The personal multicast is implemented in SwitchWare by
a small program which moves itself from SwitchWare switch to switch [Shoch
and Hupp 1982], replicating itself selectively to output ports to create a perpacket multicast. In particular, sophisticated acknowledgement [Calderon et
al. 1998] and congestion control [Faber 1998] mechanisms for multicast
communication are proposed. Support of heterogeneous group communication
is considered in [Metzler et al. 1999]. An active networking for QoS support
was presented in this paper, the QoS support can even be adapted during runtime by having service modules dynamically loaded into the network
equipment.
There have been some efforts on active Ethernet switch in industry.
EtherCAT, an Ethernet network from Beckhoff for control automation
technology, is one of these attempts to use active switches to overcome the
system limitations of other Ethernet solutions [Jost http://www.beckhoff.com
add Ref. real-time Ethernet: ultra high speed right up to the I/O]. With
EtherCAT technology, switching is done in the network interfaces. A newly
developed FMMU (fieldbus memory management unit) in each I/O terminal
reads the data addressed to it, while the Ethernet frame continues through the
device. Similarly, input data are inserted while the Ethernet frame passes
through. EtherCAT has many good ideas, such as limiting the delay from the
Ethernet node to the actual I/O or drive controller by a few nanoseconds.
Some research has been done based on switched Ethernet with a similar idea
of enhancing the switch with some intelligence, e.g., supporting for guarantees
for real-time traffic without modifications to the Ethernet hardware [Hoang et
al. 2002] [Hoang et al. 2002 B].
The combining switches topic is a heave research area in parallel computing to avoid
system bottlenecks, such as hot spots contention [Dickey and Percus 1992] [Lee et al.
1986]. With message combining, the transmission overhead costs are spread over larger
combined payloads. Simulation results show that combining switches has a significant
improvement of avoiding the effect of hot spots under the distributed-shared-memory
paradigm where messages are relatively small. Even in larger systems and under a severe
hot spot, message combining is beneficial and succeeds in reducing average latency
significantly [Katsinis 2004].
3.2
Real-time support in packet-switched networks
Here I present the related work with focus on real-time support in packet
switched networks. Methods that offer real-time support in switched Ethernet
networks are especially relevant.
Since the CSMA/CD (Carrier Sense Multiple Access, Collision Detect) used in
shared medium Ethernet is not deterministic, Ethernet has no inherent
support for real-time communication. Today, Ethernet networks increasingly
moves toward switches as implementation technology thus replacing buses. In
contrast to CSMA/CD, there is no shared medium in the switched Ethernet.
Instead, end devices are assigned a full duplex connection to the switch. As a
result, there is no collision and contention.
Switched Ethernet relies on packet switch technology. In a packet-switched
network, each packet traverses a number of physical links towards the final
destination. After each hop, the packet is stored (queued) in a switch (or a
router). The choice of queuing architecture, traffic handling etc. is essential for
the QoS characteristics.
Several disciplines for real-time communication in packet-switched networks
can be classified into the following three classes, guaranteed deterministic
services (hard real-time), guaranteed statistical services (soft real-time) and
non-real-time traffic according to the service they offer [Zhang 1995]. If the
service offered is deterministic, it is normally offering both a guaranteed
minimum throughput and a bounded end-to-end delay. If the service offered is
probabilistic, it can still be the case that a guarantee is offered but only as a
guarantee to meet the specified QoS level at a certain probability. However,
the focus in [Zhang 1995] has been on wide area networks, e.g., the function of
each separate switch and the handling of each separate logical connection. A
great deal of work on establishing real-time channels in packet-switched
networks and using deadline sorting in the switch to gain real-time support
are founded in [Ferrari and Verma 1990] [Zhang and Shin 1994] [Rexford et al.
1998].
A number of protocols and schemes have been proposed to improve the realtime characteristics of switched Ethernet [Choi et al. 2000] [Song 2001] [Song
et al. 2002] [Jasperneite et al. 2002] [Georges et al. 2002] [Varadarajan and
Chiueh 1998] [Chiueh 2001] [Hoang 2002] [Hoang and Jonsson 2003] [Cheng
et al. 2002] [Loeser and Haertig 2004]. Some works evaluate the real-time
performance of the switched Ethernet by using stochastic modeling, and
estimate the queuing delay inside the switch [Choi et al. 2000] [Song 2001]
[Song et al. 2002] [Jasperneite et al. 2002]. Nevertheless, these analytic results
are based on some assumptions about traffic arrival, which are not exactly
representative to the messages sent by the applications.
A recent result of implementing traffic shaping for switched Ethernet to
achieve low-latency real-time communication was presented [Loeser and
Haertig 2004]. They reported the measurements with both Fast and Gigabit
Ethernet, which show it is able to guarantee sub-millisecond delays for a
network utilization of 93% and 49%, respectively. The limitation of this
approach is the achievable maximum transmission delay mainly depends on
the granularity of the traffic shaping, that is how often the traffic shapers are
run. In [Georges et al. 2002], the switched Ethernet network is modeled by
using a sequence of elementary components, and the properties of each
component are aggregated to obtain the maximum end-to-end delay. However,
Quality of Service (QoS) mechanisms, such as priorities, are not introduced in
these models.
The goal of the EtheReal project [Varadarajan and Chiueh 1998] [Chiueh
2001] was to build a scaleable real-time Ethernet switch, which supports
bandwidth reservation and guarantee without any hardware or operating
system modification. As an early work on switched real-time Ethernet,
EtheReal was only throughput-oriented, i.e., no explicit treatment of hard realtime communication. Another weakness of EtheReal is that the influence of
non-real-time traffic is not well controlled.
Another approach is to add a thin software layer between the Ethernet
protocols and the TCP/IP suite in the end nodes and the switch to provide
guarantees for both bit rates and delivery deadlines for periodic real-time
traffic [Hoang et al. 2002] [Hoang and Jonsson 2003]. The switch is responsible
for admission control where the feasibility analysis is made for each link and
direction between the end-nodes and the switch. EDF scheduling is used both
in the end nodes and the switch.
References
[Calderon et al. 1998] M. Calderon, M. Sedano, A. Azcorra and C. Alonso, "Active network
support for multicast applications", IEEE Network Magzine, vol. 12, no. 3, pp. 46-52, May
1998.
[Chen et al. 2002] J. Cheng, Z. Wang, and Y. Sun, "Real-time capability analysis for switch
industrial Ethernet traffic priority-based," Proc.of IEEE International Conference on Control
Applications, Glasgow, Scotland, U.K, Sept. 18-20, 2002.
[Choi et al. 2000] B.Y. Choi, S. Song, N. Birch, and J. Huang, "Probabilistic approach to
switched Ethernet for real-time control applications", Proc. of 7th International Conference on
Real-time Computing Systems and Applications, pp. 384-388, Dec. 12-14, 2000.
[Dharmalingam and Collier 2001]. Kalaiarul Dharmalingam and Martin Collier, "Netlets: a
new active network architecture", IEE/IEI Telecommunications Systems Research Symposium,
Dublin, Ireland 2001.
[Dickey and Percus 1992] S. R. Dickey and O. E. Percus, "Performance differences among
combining switch architectures", Proc. International Conference on Parallel Processing
(ICPP'1992), An Arbor, MI, USA, Aug. 17-21, 1992.
[Ferrari and Verma 1990] D. Ferrari and D. Verma, "A scheme for real-time channel
establishment in wide area network", IEEE Journal on Selected Areas in Communications,
vol. 8, no. 3, Apr. 1990, pp. 368-379.
[Georges 2002] J.P. Georges, E. Rondeau, and T. Divoux, "How to be sure that switched
Ethernet networks satisfy the real-time requirements of an industrial application?", Proc. of
2002 IEEE International Symposium on Industrial Electronics, pp. 158-163, July 8-11, 2002.
[Hoang and Jonsson 2003] H. Hoang, and M. Jonsson, "Switched real-time Ethernet in
industry applications," Proc. of the 9th Asian Pacific Conference on Communication, Penang,
Malaysia, Sept. 2003.
[Hoang et al. 2002] H. Hoang, M. Jonsson, A. Larsson, R. Olsson, and C. Bergenhem,
"Deadline first scheduling in switched real-time Ethernet-deadline partitioning issues and
software implementation experiments". Proc. Workshop on Real-time LANs in the Internet
Age (RTLIA 2002), June 2002, Vienna, Austria.
[Hoang et al. 2002B] H. Hoang, M. Jonsson, U. Hagström, and A. Kallerdahl, "Switched realtime Ethernet with earliest deadline first scheduling - protocols and traffic handling," Proc.
Workshop on Parallel and Distributed Real-Time Systems (WPDRTS'2002) in conjunction
with International Parallel and Distributed Processing Symposium (IPDPS'02), Fort
Lauderdale, FL, USA, Apr. 15-16, 2002.
[Jasperneite et al. 2002] J. Jasperneite, P. Neumann, M. Theis, and K. Watson, "Deterministic
real-time communication with switched Ethernet", Proc. of 4th International Workshop on
Factory Communication Systems, pp. 11-18, Västerås, Sweden, Aug. 28-30, 2002.
[Jost http://www.beckhoff.com add Ref. real-time Ethernet: ultra high speed right up to the
I/O]
[Kaplan 2001] G. Kaplan, "Ethernet’s winning ways," IEEE Spectrum, vol. 38, Jan. 2001.
[Lee et al. 1986] G. H. Lee, C. P. Kruskal, and D. J. Kuck, "The effectiveness of combining in
shared memory parallel computers in the presence of ‘hot spots", Proc. International
Conference on Parallel Processing (ICPP'1986), University park, PA, USA, Aug. 17-21,
1986, pp. 35-41.
[Loeser and Haertig 2004] J. Loeser and H. Haertig "Low latency hard real-time
communication over switched Ethernet", Proc. of the 16th Euromicro Conference on Realtime Systems (ECRTC’04), Catania, Italy, June 2004.
[Song 2001] Y. Song, "Time constrained communication over switched Ethernet", Proc. of
International Conference on Fieldbus Systems and Their Applications(IFAC’2001), pp. 152169, Nov. 2001.
[Song et al. 2002] Y. Song, A. Koubaa, and F. Simonot, "Switched Ethernet for real-time
industrial communication modelling and message buffering delay evaluation", Proc. of 4th
International Workshop on Factory Communication Systems, pp. 27-35, Aug. 28-30, 2002.
[Tennenhouse and Wetherall 2002] D.L. Tennenhouse and D. J. Wetherall, "Towards an
active network architecture," Proceedings of the DAPPA Active Netowrks Conference and
Exposition (DANCE’02) , San Francisco, CA, USA, May 29-30, 2002.
[Tennenhouse et al. 1997] L. David, J. M. Smith, W. D. Sincoskie, D. J. Wetherall and G. J.
Minden, "A survey of active network research," IEEE Communication Magazine, vol. 35,
no.1, pp. 80-86, Jan. 1997.
[Varadarajan 2001] S. Varadarajan, " Experiences with Ethereal: a fault tolerant real-time
Ethernet switch," Proceedings of 8th IEEE International Conference on Emerging
Technologies and Factory Automation, pp. 183-194, vol. 1, 2001.
[Varadarajan and Chiueh 1998] S. Varadarajan and T. Chiueh, "Ethereal: a host-transparent
real-time Fast Ethernet switch," Proceedings of 6th IEEE International Conference on
Networks, Protocols, pp. 12-21, Oct. 1998.
[Wetherall et al. 1998] D. Wetherall, U. Legedza and J. guttag, "Introducing new Internet
services: why and how? ", IEEE Network Magazine, vol. 12, no. 3, pp. 12-19, May 1998.
[Yemini and Silva 1996] Y. Yemini, S. D. Silva, "Towards programmable networks''
IFIP/IEEE International Workshop on Distributed Systems: Operations and Management,
L’Aquila, Italy, Oct. 1996.
[Zhang and Shin 1994] Q. Zhang and K. G. Shin, "On the ability of establishing real-time
channels in point-to-point packet-switched networks", IEEE Transactions on
Communications, vol. 42, No. 2/3/4, pp. 1096-1105, Feb./Mar./Apr. 1994.