Demonstrates that Ethernet LAN traffic is statistically self-similar, that none of the commonly used traffic models is able to capture this fractal-like behavior, that such behavior has serious implications for the design, control, and analysis of high-speed, cell-based networks, and that aggregating streams of such traffic typically intensifies the self-similarity ("burstiness") instead of smoothing it. These conclusions are supported by a rigorous statistical analysis of hundreds of millions of high quality Ethernet traffic measurements collected between 1989 and 1992, coupled with a discussion of the underlying mathematical and statistical properties of self-similarity and their relationship with actual network behavior. The authors also present traffic models based on self-similar stochastic processes that provide simple, accurate, and realistic descriptions of traffic scenarios expected during B-ISDN deployment. >

/pdf/on-the-self-similar-nature-of-ethernet-traffic-extended-3i4efiwsq9.pdf

On the self-similar nature of Ethernet traffic (extended version)

Recent advances in miniaturization and low-cost, low-power design have led to active research in large-scale networks of small, wireless, low-power sensors and actuators. Time synchronization is critical in sensor networks for diverse purposes including sensor data fusion, coordinated actuation, and power-efficient duty cycling. Though the clock accuracy and precision requirements are often stricter than in traditional distributed systems, strict energy constraints limit the resources available to meet these goals.We present Reference-Broadcast Synchronization, a scheme in which nodes send reference beacons to their neighbors using physical-layer broadcasts. A reference broadcast does not contain an explicit timestamp; instead, receivers use its arrival time as a point of reference for comparing their clocks. In this paper, we use measurements from two wireless implementations to show that removing the sender's nondeterminism from the critical path in this way produces high-precision clock agreement (1.85 ± 1.28μsec, using off-the-shelf 802.11 wireless Ethernet), while using minimal energy. We also describe a novel algorithm that uses this same broadcast property to federate clocks across broadcast domains with a slow decay in precision (3.68 ± 2.57μsec after 4 hops). RBS can be used without external references, forming a precise relative timescale, or can maintain microsecond-level synchronization to an external timescale such as UTC. We show a significant improvement over the Network Time Protocol (NTP) under similar conditions.

/pdf/fine-grained-network-time-synchronization-using-reference-3go0i1u94g.pdf

Fine-grained network time synchronization using reference broadcasts

A shared data structure is lock-free if its operations do not require mutual exclusion. If one process is interrupted in the middle of an operation, other processes will not be prevented from operating on that object. In highly concurrent systems, lock-free data structures avoid common problems associated with conventional locking techniques, including priority inversion, convoying, and difficulty of avoiding deadlock. This paper introduces transactional memory, a new multiprocessor architecture intended to make lock-free synchronization as efficient (and easy to use) as conventional techniques based on mutual exclusion. Transactional memory allows programmers to define customized read-modify-write operations that apply to multiple, independently-chosen words of memory. It is implemented by straightforward extensions to any multiprocessor cache-coherence protocol. Simulation results show that transactional memory matches or outperforms the best known locking techniques for simple benchmarks, even in the absence of priority inversion, convoying, and deadlock.

/pdf/transactional-memory-architectural-support-for-lock-free-4e7zw2v6tu.pdf

Transactional memory: architectural support for lock-free data structures

Virtual time is a new paradigm for organizing and synchronizing distributed systems which can be applied to such problems as distributed discrete event simulation and distributed database concurrency control. Virtual time provides a flexible abstraction of real time in much the same way that virtual memory provides an abstraction of real memory. It is implemented using the Time Warp mechanism, a synchronization protocol distinguished by its reliance on lookahead-rollback, and by its implementation of rollback via antimessages.

Virtual time

Remote procedure calls (RPC) appear to be a useful paradig m for providing communication across a network between programs written in a high-level language. This paper describes a package providing a remote procedure call facility, the options that face the designer of such a package, and the decisions ~we made. We describe the overall structure of our RPC mechanism, our facilities for binding RPC clients, the transport level communication protocol, and some performance measurements. We include descriptioro~ of some optimizations used to achieve high performance and to minimize the load on server machines that have many clients.

/pdf/implementing-remote-procedure-calls-2l74p5a3rf.pdf

Implementing remote procedure calls

Ethernet is a branching broadcast communication system for carrying digital data packets among locally distributed computing stations. The packet transport mechanism provided by Ethernet has been used to build systems which can be viewed as either local computer networks or loosely coupled multiprocessors. An Ethernet's shared communication facility, its Ether, is a passive broadcast medium with no central control. Coordination of access to the Ether for packet broadcasts is distributed among the contending transmitting stations using controlled statistical arbitration. Switching of packets to their destinations on the Ether is distributed among the receiving stations using packet address recognition. Design principles and implementation are described, based on experience with an operating Ethernet of 100 nodes along a kilometer of coaxial cable. A model for estimating performance under heavy loads and a packet protocol for error controlled communication are included for completeness.

/pdf/ethernet-distributed-packet-switching-for-local-computer-oet42clemw.pdf

Ethernet: distributed packet switching for local computer networks

A broadcast packet is for delivery to all nodes of a network. Algorithms for accomplishing this delivery through a store-and-forward packet switching computer network include (1) transmission of separately addressed packets, (2) multidestination addressing, (3) hot potato forwarding, (4) spanning tree forwarding, and (5) source based forwarding. To this list of algorithms we add (6) reverse path forwarding, a broadcast routing method which exploits routing procedures and data structures already available for packet switching. Reverse path forwarding is a practical algorithm for broadcast routing in store-and-forward packet switching computer networks. The algorithm is described as being practical because it is not optimal according to metrics developed for its analysis in this paper, and also because it can be implemented in existing networks with less complexity than that required for the known alternatives.

http://www.mathcs.emory.edu/~cheung/Courses/558/Syllabus/XX-Multicast/metcalfe.pdf

Reverse path forwarding of broadcast packets

Apparatus for enabling communications between two or more data processing stations comprising a communication cable arranged in branched segments including taps distributed thereover. Tied to each tap is a transceiver which on the other side connects to an associated interface stage. Each transceiver includes, in addition to the usual transmitter and receiver sections, a gate which compares the data from the interface stage with the data on the cable and indicates whether such are equal. Should such be unequal, an interference between the transceiver and the cable is indicated, disabling the associated transmitter section. Each interface stage tied to such transceiver also includes an input and an output buffer on the other end thereof interfacing with a using device, such input and output buffers storing both the incoming and outgoing data. The output buffer is connected to a clock-driven shift register which converts the buffered data to a serial stream, feeds such data to a phase encoder, which then connects to the transmitter or driver section of the transceiver. The input buffer is loaded by an input shift register which derives its clock from a phase decoder, the shift register and the phase decoder both connecting to the receiver section. When the station is to start transmitting, the phase decoder detects the presence of other transmissions on the cable and detains the output shift register until no other transmissions are sensed. Once a transmission has begun, if interference is detected and the transmitter section is disabled, a random number generator is used to select an interval of time at the completion of which the next attempted transmission will take place. Concurrently, a counter counts the number of interferences, or collisions, which recur in the attempted transmissions of one data packet and weights the mean of the random number generator accordingly. The input shift register is also connected to an address decoder which enables data transfer to the input buffer only during those times when the data is preceded by an appropriate address.

Multipoint data communication system with collision detection

Research in network interconnection techniques has been motivated by the desire to permit communication among diverse, geographically distributed computing resources and users interconnected by a wide variety of network technologies.

Robert M. Metcalfe

Papers

Ethernet: distributed packet switching for local computer networks

Reverse path forwarding of broadcast packets

Multipoint data communication system with collision detection

Ethernet: distributed packet switching for local computer networks

A Specific Internetwork Architecture (Pup)