In this paper, we study the performance of solid-state drives that employ flash technology as storage medium. Our prime objective is to understand how the scheduling of the user-generated read and write commands and the read, write, and erase operations induced by the garbage-collection process affect the basic performance measures throughput and latency. We demonstrate that the most straightforward scheduling that prioritizes the processing of garbage-collection-related commands over user-related commands suffers from severe latency deficiencies. These problems can be overcome by using a more sophisticated priority scheme that minimizes the user-perceived latency without throughput penalty or deadlock exposure. Using both analysis and simulation, we investigate how these schemes perform under a variety of system design parameters and workloads. Our results can be directly applied to the engineering of a performance-optimized solid-state-drive system.

Distributed-queue multiple-access (DQMA) and cyclic-reservation multiple-access (CRMA) access schemes for gigabit-per-second local and metropolitan area networks are described. These schemes are based on a slotted unidirectional bus structure, with both folded and dual-bus configurations. Like the distributed-queue dual-bus (DQDB) scheme, both the DQMA and the CRMA access schemes achieve full throughput independent of network speed and distance. They have two significant advantages over DQDB. They provide throughput fairness even at high speeds and large distances, where DQDB exhibits dramatic unfairness, and, by allowing reservation of multiple consecutive slots, they make segment labeling unnecessary and facilitate packet reassembly significantly.>

The major technical concepts underlying token-ring technology are examined, and performance issues arising in the design of such local-area networks (LANs) are detailed. Following a survey of analytical queuing models to describe the basic token-ring operation, three topics are discussed in detail: (1) the IEEE 802.5 token ring and its performance; (2) the ANSI fiber distributed data interface (FDDI) token ring and its performance; and (3) architecture and performance issues arising in the interconnection of token-ring networks with regard to the various components of a multiring architecture demonstrating which congestion-control problems can arise in such a network and how they can be overcome. The author concludes with a number of open questions concerning the understanding of the quantitative behavior of token-ring-based LANs.>

The fiber distributed data interface (FDDI) provides a 100 Mb/s communication system to interconnect computer and peripheral equipment using fiber optics as the transmission medium in a ring configuration. The performance of the FDDI media-access control (MAC) protocol, a timed-token protocol, is analyzed. The FDDI MAC priority mechanism supports two classes of traffic: synchronous and asynchronous. The focus is on the relationship between the FDDI MAC parameter settings, the ring configuration, and the performance of the asynchronous priority levels. A procedure to calculate estimates for the throughput of each asynchronous priority level over a range of frame-arrival rates is developed. Performance results are presented to demonstrate characteristics of the FDDI priority mechanism. A procedure is described that can be used to tune the FDDI parameters so that given performance objectives for the various priority levels are achieved.>

We investigate flow-control issues in local-area networks consisting of multiple token rings interconnected through bridges. To achieve high throughput, bridges perform only a very simple routing and store-and-forward function, but are not involved in error- or flowcontrol. In case of congestion, bridges discard arriving frames, which will be recovered through an appropriate end-to-end protocol between the communicating stations. The end-to-end protocol considered is the IEEE 802.2 type-2 logical-link-control (LLC) protocol. Extensive simulations show that performance can be severely degraded if, in such a network, the LLC protocol is employed as defined today. Therefore, we suggest an enhancement to this protocol in the form of a dynamic flow-control algorithm. As our results demonstrate, this enhancement guarantees close-to-optimal network performance under both normal traffic load and overload conditions.

A model of an adapter to a local area network is used to study the timing problem associated with the reception of back-to-back frames from a multiaccess transmission medium. The conditions under which frames are received successfully are presented. Analytic results are also obtained for the probabilities that frames are lost and a given number of back-to-back frames is successfully received. Numerical results illustrating the effect of the various design parameters on adapter performance are discussed.

Architecture, performance, transmission system, and wiring strategy of a token-ring local area network implemented at the IBM Zurich Research Laboratory are described. In the design of the system, particular emphasis was placed on high reliability, availability, and serviceability. To ensure robustness of the token-access protocol, we employ the concept of a monitor function which is responsible for fast recovery from access-related errors. Our protocol supports asynchronous transmission of data frames concurrently with full-duplex synchronous channels, e.g., for voice services or other applications requiring guaranteed delay. The delay-throughput performance of the token ring is shown to depend very little on data rate and distance. The transmission system of the ring is fully bit synchronous and allows insertion/removal of stations in/from the ring at any time. A mixed ring/star wiring strategy is used which provides the means for both fault detection and isolation, and system reconfiguration, and allows wiring of a building systematically.

The performance analysis of a local-area bus access scheme is presented, which is characterized by an alternating sequence of scheduling and data-transmission intervals. For different scheduling disciplines, viz., fixed and variable priorities, and "shortest-packet-first," we derive explicit results for the mean packet-transfer delay. Our results give insight into the question of fairness among the stations, and the potential for performance improvement by sophisticated scheduling.

A new analytic approach for the performance evaluation of buffer insertion rings is proposed which overcomes the basic limitations of former analyses. Key to our analysis is that the global performance model of a buffer insertion ring is decomposed into submodels in such a way that the basic interdependence of interarrival and transmission times of the frames transmitted is taken into account. The analysis itself is relatively straightforward and yields simple and explicit results for the mean delays of the frames. For special traffic patterns, the analysis is exact, for the general case, it is approximate but its accuracy turns out to be high, as comparison with simulation results shows.

This paper provides a comparative evaluation of the performance of ring and bus systems constituting subnetworks of local-area networks. Performance is measured in terms of the delaythroughput characteristics. Systems investigated include token-controlled and slotted rings as well as random-access buses (CSMA with collision detection) and ordered-access buses. The investigation is based on analytic models which describe the various topologies and access mechanisms to a sufficient level of detail. The paper includes a comprehensive discussion of how the performance of the different networks is affected by system parameters like transmission rate, cable length, packet lengths, and control overhead.

The prime objective of the present paper is to analyze the performance of HDLC Balanced Class of Procedures, i.e., to quantitatively study the interaction among a multiplicity of parameters which are procedure specific, characterize the properties of the transmission medium, and identify the operational characteristics and requirements. The approach taken is to consider two kinds of operation: a saturated case characterized by maximum throughput as the most suitable measure of performance, and a nonsaturated situation for which waiting and transfer times are the appropriate measures. The analysis is performed by means of both simulation and analytic techniques. Within the simulation model, the information transfer phase of the HDLC procedure was implemented in full detail. The key idea of the analytic approach is to use a so-called virtual transmission time, a quantity comprising both the real constant transmission time of a message and the duration of recovery actions in case of transmission errors. It allows the performance measures to be represented by explicit and easily-computable expressions. The results provide a fundamental insight into how the most relevant parameters interact and determine performance.

The prime objective of the present paper is to analyze the throughput characteristics of a multi-processor configuration for packet switching-suggested in the context of an integrated network for circuit and packet switching-and to identify the major parameters on which it depends. The results provide a fundamental insight into how the most relevant system and traffic parameters determine throughput as a function of the number of packet-handling modules. In particular, it is shown that the proposed partitioning of the packet-handling function does indeed permit high throughput before saturation is encountered.